Feat: read web api testcases (#12383)

### What problem does this PR solve?

Web API testcase for list_messages, get_recent_message.

### Type of change

- [x] New Feature (non-breaking change which adds functionality)

---------

Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>

.github/workflows/release.yml

@@ -10,6 +10,12 @@ on:
     tags:
       - "v*.*.*"                  # normal release
 
+permissions:
+  contents: write
+  actions: read
+  checks: read
+  statuses: read
+
 # https://docs.github.com/en/actions/using-jobs/using-concurrency
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
@@ -76,6 +82,14 @@ jobs:
           # The body field does not support environment variable substitution directly.
           body_path: release_body.md
 
+      - name: Build and push image
+        run: |
+          sudo docker login --username infiniflow --password-stdin <<< ${{ secrets.DOCKERHUB_TOKEN }}
+          sudo docker build --build-arg NEED_MIRROR=1 --build-arg HTTPS_PROXY=${HTTPS_PROXY} --build-arg HTTP_PROXY=${HTTP_PROXY} -t infiniflow/ragflow:${RELEASE_TAG} -f Dockerfile .
+          sudo docker tag infiniflow/ragflow:${RELEASE_TAG} infiniflow/ragflow:latest
+          sudo docker push infiniflow/ragflow:${RELEASE_TAG}
+          sudo docker push infiniflow/ragflow:latest
+
       - name: Build and push ragflow-sdk
         if: startsWith(github.ref, 'refs/tags/v')
         run: |
@@ -85,11 +99,3 @@ jobs:
         if: startsWith(github.ref, 'refs/tags/v')
         run: |
           cd admin/client && uv build && uv publish --token ${{ secrets.PYPI_API_TOKEN }}
-
-      - name: Build and push image
-        run: |
-          sudo docker login --username infiniflow --password-stdin <<< ${{ secrets.DOCKERHUB_TOKEN }}
-          sudo docker build --build-arg NEED_MIRROR=1 --build-arg HTTPS_PROXY=${HTTPS_PROXY} --build-arg HTTP_PROXY=${HTTP_PROXY} -t infiniflow/ragflow:${RELEASE_TAG} -f Dockerfile .
-          sudo docker tag infiniflow/ragflow:${RELEASE_TAG} infiniflow/ragflow:latest
-          sudo docker push infiniflow/ragflow:${RELEASE_TAG}
-          sudo docker push infiniflow/ragflow:latest

.github/workflows/tests.yml

@@ -205,29 +205,29 @@ jobs:
       - name: Run sdk tests against Elasticsearch
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api
+          source .venv/bin/activate && set -o pipefail; pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api 2>&1 | tee es_sdk_test.log
 
       - name: Run frontend api tests against Elasticsearch
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py
+          source .venv/bin/activate && set -o pipefail; pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py 2>&1 | tee es_api_test.log
           
       - name: Run http api tests against Elasticsearch
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api
+          source .venv/bin/activate && set -o pipefail; pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api 2>&1 | tee es_http_api_test.log
 
       - name: Stop ragflow:nightly
         if: always()  # always run this step even if previous steps failed
@@ -243,29 +243,29 @@ jobs:
       - name: Run sdk tests against Infinity
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api
+          source .venv/bin/activate && set -o pipefail; DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api 2>&1 | tee infinity_sdk_test.log
 
       - name: Run frontend api tests against Infinity
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py
+          source .venv/bin/activate && set -o pipefail; DOC_ENGINE=infinity pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py 2>&1 | tee infinity_api_test.log
 
       - name: Run http api tests against Infinity
         run: |
           export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
-          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS} > /dev/null; do
+          until sudo docker exec ${RAGFLOW_CONTAINER} curl -s --connect-timeout 5 ${HOST_ADDRESS}/v1/system/ping > /dev/null; do
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api
+          source .venv/bin/activate && set -o pipefail; DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api 2>&1 | tee infinity_http_api_test.log
 
       - name: Stop ragflow:nightly
         if: always()  # always run this step even if previous steps failed

Dockerfile

@@ -19,17 +19,17 @@ RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/huggingface.co
 # This is the only way to run python-tika without internet access. Without this set, the default is to check the tika version and pull latest every time from Apache.
 RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/,target=/deps \
     cp -r /deps/nltk_data /root/ && \
-    cp /deps/tika-server-standard-3.0.0.jar /deps/tika-server-standard-3.0.0.jar.md5 /ragflow/ && \
+    cp /deps/tika-server-standard-3.2.3.jar /deps/tika-server-standard-3.2.3.jar.md5 /ragflow/ && \
     cp /deps/cl100k_base.tiktoken /ragflow/9b5ad71b2ce5302211f9c61530b329a4922fc6a4
 
-ENV TIKA_SERVER_JAR="file:///ragflow/tika-server-standard-3.0.0.jar"
+ENV TIKA_SERVER_JAR="file:///ragflow/tika-server-standard-3.2.3.jar"
 ENV DEBIAN_FRONTEND=noninteractive
 
 # Setup apt
 # Python package and implicit dependencies:
 # opencv-python: libglib2.0-0 libglx-mesa0 libgl1
 # aspose-slides: pkg-config libicu-dev libgdiplus         libssl1.1_1.1.1f-1ubuntu2_amd64.deb
-# python-pptx:   default-jdk                              tika-server-standard-3.0.0.jar
+# python-pptx:   default-jdk                              tika-server-standard-3.2.3.jar
 # selenium:      libatk-bridge2.0-0                       chrome-linux64-121-0-6167-85
 # Building C extensions: libpython3-dev libgtk-4-1 libnss3 xdg-utils libgbm-dev
 RUN --mount=type=cache,id=ragflow_apt,target=/var/cache/apt,sharing=locked \
@@ -192,6 +192,7 @@ COPY pyproject.toml uv.lock ./
 COPY mcp mcp
 COPY plugin plugin
 COPY common common
+COPY memory memory
 
 COPY docker/service_conf.yaml.template ./conf/service_conf.yaml.template
 COPY docker/entrypoint.sh ./

Dockerfile.deps

@@ -3,7 +3,7 @@
 FROM scratch
 
 # Copy resources downloaded via download_deps.py
-COPY chromedriver-linux64-121-0-6167-85 chrome-linux64-121-0-6167-85 cl100k_base.tiktoken libssl1.1_1.1.1f-1ubuntu2_amd64.deb libssl1.1_1.1.1f-1ubuntu2_arm64.deb tika-server-standard-3.0.0.jar tika-server-standard-3.0.0.jar.md5 libssl*.deb uv-x86_64-unknown-linux-gnu.tar.gz /
+COPY chromedriver-linux64-121-0-6167-85 chrome-linux64-121-0-6167-85 cl100k_base.tiktoken libssl1.1_1.1.1f-1ubuntu2_amd64.deb libssl1.1_1.1.1f-1ubuntu2_arm64.deb tika-server-standard-3.2.3.jar tika-server-standard-3.2.3.jar.md5 libssl*.deb uv-x86_64-unknown-linux-gnu.tar.gz /
 
 COPY nltk_data /nltk_data
 

README.md

@@ -22,7 +22,7 @@
         <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Document</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -72,7 +72,7 @@
 
 ## 💡 What is RAGFlow?
 
-[RAGFlow](https://ragflow.io/) is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs. It offers a streamlined RAG workflow adaptable to enterprises of any scale. Powered by a converged context engine and pre-built agent templates, RAGFlow enables developers to transform complex data into high-fidelity, production-ready AI systems with exceptional efficiency and precision.
+[RAGFlow](https://ragflow.io/) is a leading open-source Retrieval-Augmented Generation ([RAG](https://ragflow.io/basics/what-is-rag)) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs. It offers a streamlined RAG workflow adaptable to enterprises of any scale. Powered by a converged [context engine](https://ragflow.io/basics/what-is-agent-context-engine) and pre-built agent templates, RAGFlow enables developers to transform complex data into high-fidelity, production-ready AI systems with exceptional efficiency and precision.
 
 ## 🎮 Demo
 
@@ -85,6 +85,7 @@ Try our demo at [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Latest Updates
 
+- 2025-12-26 Supports 'Memory' for AI agent.
 - 2025-11-19 Supports Gemini 3 Pro.
 - 2025-11-12 Supports data synchronization from Confluence, S3, Notion, Discord, Google Drive.
 - 2025-10-23 Supports MinerU & Docling as document parsing methods.
@@ -187,15 +188,15 @@ releases! 🌟
 > All Docker images are built for x86 platforms. We don't currently offer Docker images for ARM64.
 > If you are on an ARM64 platform, follow [this guide](https://ragflow.io/docs/dev/build_docker_image) to build a Docker image compatible with your system.
 
-> The command below downloads the `v0.22.1` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.22.1`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server.
+> The command below downloads the `v0.23.1` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.23.1`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server.
 
 ```bash
    $ cd ragflow/docker
-  
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # Optional: use a stable tag (see releases: https://github.com/infiniflow/ragflow/releases)
    # This step ensures the **entrypoint.sh** file in the code matches the Docker image version.
-   
+
    # Use CPU for DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
 
@@ -206,10 +207,10 @@ releases! 🌟
 
 > Note: Prior to `v0.22.0`, we provided both images with embedding models and slim images without embedding models. Details as follows:
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > Starting with `v0.22.0`, we ship only the slim edition and no longer append the **-slim** suffix to the image tag.
 
@@ -232,7 +233,7 @@ releases! 🌟
     * Running on all addresses (0.0.0.0)
    ```
 
-   > If you skip this confirmation step and directly log in to RAGFlow, your browser may prompt a `network anormal`
+   > If you skip this confirmation step and directly log in to RAGFlow, your browser may prompt a `network abnormal`
    > error because, at that moment, your RAGFlow may not be fully initialized.
    >
 5. In your web browser, enter the IP address of your server and log in to RAGFlow.
@@ -302,6 +303,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+Or if you are behind a proxy, you can pass proxy arguments:
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 Launch service from source for development
 
 1. Install `uv` and `pre-commit`, or skip this step if they are already installed:
@@ -386,7 +396,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 Roadmap
 
-See the [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214)
+See the [RAGFlow Roadmap 2026](https://github.com/infiniflow/ragflow/issues/12241)
 
 ## 🏄 Community
 

README_id.md

@@ -22,7 +22,7 @@
         <img alt="Lencana Daring" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Rilis%20Terbaru" alt="Rilis Terbaru">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Dokumentasi</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Peta Jalan</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Peta Jalan</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -72,7 +72,7 @@
 
 ## 💡 Apa Itu RAGFlow?
 
-[RAGFlow](https://ragflow.io/) adalah mesin RAG (Retrieval-Augmented Generation) open-source terkemuka yang mengintegrasikan teknologi RAG mutakhir dengan kemampuan Agent untuk menciptakan lapisan kontekstual superior bagi LLM. Menyediakan alur kerja RAG yang efisien dan dapat diadaptasi untuk perusahaan segala skala. Didukung oleh mesin konteks terkonvergensi dan template Agent yang telah dipra-bangun, RAGFlow memungkinkan pengembang mengubah data kompleks menjadi sistem AI kesetiaan-tinggi dan siap-produksi dengan efisiensi dan presisi yang luar biasa.
+[RAGFlow](https://ragflow.io/) adalah mesin [RAG](https://ragflow.io/basics/what-is-rag) (Retrieval-Augmented Generation) open-source terkemuka yang mengintegrasikan teknologi RAG mutakhir dengan kemampuan Agent untuk menciptakan lapisan kontekstual superior bagi LLM. Menyediakan alur kerja RAG yang efisien dan dapat diadaptasi untuk perusahaan segala skala. Didukung oleh mesin konteks terkonvergensi dan template Agent yang telah dipra-bangun, RAGFlow memungkinkan pengembang mengubah data kompleks menjadi sistem AI kesetiaan-tinggi dan siap-produksi dengan efisiensi dan presisi yang luar biasa.
 
 ## 🎮 Demo
 
@@ -85,6 +85,7 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Pembaruan Terbaru
 
+- 2025-12-26 Mendukung 'Memori' untuk agen AI.
 - 2025-11-19 Mendukung Gemini 3 Pro.
 - 2025-11-12 Mendukung sinkronisasi data dari Confluence, S3, Notion, Discord, Google Drive.
 - 2025-10-23 Mendukung MinerU & Docling sebagai metode penguraian dokumen.
@@ -187,12 +188,12 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
 > Semua gambar Docker dibangun untuk platform x86. Saat ini, kami tidak menawarkan gambar Docker untuk ARM64.
 > Jika Anda menggunakan platform ARM64, [silakan gunakan panduan ini untuk membangun gambar Docker yang kompatibel dengan sistem Anda](https://ragflow.io/docs/dev/build_docker_image).
 
-> Perintah di bawah ini mengunduh edisi v0.22.1 dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.22.1, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server.
+> Perintah di bawah ini mengunduh edisi v0.23.1 dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.23.1, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server.
 
 ```bash
    $ cd ragflow/docker
-   
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # Opsional: gunakan tag stabil (lihat releases: https://github.com/infiniflow/ragflow/releases)
    # This steps ensures the **entrypoint.sh** file in the code matches the Docker image version.
 
@@ -206,10 +207,10 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
 
 > Catatan: Sebelum `v0.22.0`, kami menyediakan image dengan model embedding dan image slim tanpa model embedding. Detailnya sebagai berikut:
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > Mulai dari `v0.22.0`, kami hanya menyediakan edisi slim dan tidak lagi menambahkan akhiran **-slim** pada tag image.
 
@@ -232,7 +233,7 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
     * Running on all addresses (0.0.0.0)
    ```
 
-   > Jika Anda melewatkan langkah ini dan langsung login ke RAGFlow, browser Anda mungkin menampilkan error `network anormal`
+   > Jika Anda melewatkan langkah ini dan langsung login ke RAGFlow, browser Anda mungkin menampilkan error `network abnormal`
    > karena RAGFlow mungkin belum sepenuhnya siap.
    >
 2. Buka browser web Anda, masukkan alamat IP server Anda, dan login ke RAGFlow.
@@ -276,6 +277,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+Jika berada di belakang proxy, Anda dapat melewatkan argumen proxy:
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 Menjalankan Aplikasi dari untuk Pengembangan
 
 1. Instal `uv` dan `pre-commit`, atau lewati langkah ini jika sudah terinstal:
@@ -358,7 +368,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 Roadmap
 
-Lihat [Roadmap RAGFlow 2025](https://github.com/infiniflow/ragflow/issues/4214)
+Lihat [Roadmap RAGFlow 2026](https://github.com/infiniflow/ragflow/issues/12241)
 
 ## 🏄 Komunitas
 

README_ja.md

@@ -22,7 +22,7 @@
         <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Document</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -53,7 +53,7 @@
 
 ## 💡 RAGFlow とは？
 
-[RAGFlow](https://ragflow.io/) は、先進的なRAG（Retrieval-Augmented Generation）技術と Agent 機能を融合し、大規模言語モデル（LLM）に優れたコンテキスト層を構築する最先端のオープンソース RAG エンジンです。あらゆる規模の企業に対応可能な合理化された RAG ワークフローを提供し、統合型コンテキストエンジンと事前構築されたAgentテンプレートにより、開発者が複雑なデータを驚異的な効率性と精度で高精細なプロダクションレディAIシステムへ変換することを可能にします。
+[RAGFlow](https://ragflow.io/) は、先進的な[RAG](https://ragflow.io/basics/what-is-rag)（Retrieval-Augmented Generation）技術と Agent 機能を融合し、大規模言語モデル（LLM）に優れたコンテキスト層を構築する最先端のオープンソース RAG エンジンです。あらゆる規模の企業に対応可能な合理化された RAG ワークフローを提供し、統合型[コンテキストエンジン](https://ragflow.io/basics/what-is-agent-context-engine)と事前構築されたAgentテンプレートにより、開発者が複雑なデータを驚異的な効率性と精度で高精細なプロダクションレディAIシステムへ変換することを可能にします。
 
 ## 🎮 Demo
 
@@ -66,7 +66,8 @@
 
 ## 🔥 最新情報
 
-- 2025-11-19 Gemini 3 Proをサポートしています
+- 2025-12-26 AIエージェントの「メモリ」機能をサポート。
+- 2025-11-19 Gemini 3 Proをサポートしています。
 - 2025-11-12 Confluence、S3、Notion、Discord、Google Drive からのデータ同期をサポートします。
 - 2025-10-23 ドキュメント解析方法として MinerU と Docling をサポートします。
 - 2025-10-15 オーケストレーションされたデータパイプラインのサポート。
@@ -167,12 +168,12 @@
 > 現在、公式に提供されているすべての Docker イメージは x86 アーキテクチャ向けにビルドされており、ARM64 用の Docker イメージは提供されていません。
 > ARM64 アーキテクチャのオペレーティングシステムを使用している場合は、[このドキュメント](https://ragflow.io/docs/dev/build_docker_image)を参照して Docker イメージを自分でビルドしてください。
 
-> 以下のコマンドは、RAGFlow Docker イメージの v0.22.1 エディションをダウンロードします。異なる RAGFlow エディションの説明については、以下の表を参照してください。v0.22.1 とは異なるエディションをダウンロードするには、docker/.env ファイルの RAGFLOW_IMAGE 変数を適宜更新し、docker compose を使用してサーバーを起動してください。
+> 以下のコマンドは、RAGFlow Docker イメージの v0.23.1 エディションをダウンロードします。異なる RAGFlow エディションの説明については、以下の表を参照してください。v0.23.1 とは異なるエディションをダウンロードするには、docker/.env ファイルの RAGFLOW_IMAGE 変数を適宜更新し、docker compose を使用してサーバーを起動してください。
 
 ```bash
    $ cd ragflow/docker
 
-   # git checkout v0.22.1
+   # git checkout v0.23.1
    # 任意: 安定版タグを利用 (一覧: https://github.com/infiniflow/ragflow/releases)
    # この手順は、コード内の entrypoint.sh ファイルが Docker イメージのバージョンと一致していることを確認します。
 
@@ -186,15 +187,15 @@
 
 > 注意：`v0.22.0` より前のバージョンでは、embedding モデルを含むイメージと、embedding モデルを含まない slim イメージの両方を提供していました。詳細は以下の通りです：
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > `v0.22.0` 以降、当プロジェクトでは slim エディションのみを提供し、イメージタグに **-slim** サフィックスを付けなくなりました。
 
-   1. サーバーを立ち上げた後、サーバーの状態を確認する:   
-   
+   1. サーバーを立ち上げた後、サーバーの状態を確認する:
+
    ```bash
    $ docker logs -f docker-ragflow-cpu-1
    ```
@@ -276,6 +277,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+プロキシ環境下にいる場合は、プロキシ引数を指定できます：
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 ソースコードからサービスを起動する方法
 
 1. `uv` と `pre-commit` をインストールする。すでにインストールされている場合は、このステップをスキップしてください:
@@ -358,7 +368,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 ロードマップ
 
-[RAGFlow ロードマップ 2025](https://github.com/infiniflow/ragflow/issues/4214) を参照
+[RAGFlow ロードマップ 2026](https://github.com/infiniflow/ragflow/issues/12241) を参照
 
 ## 🏄 コミュニティ
 

README_ko.md

@@ -22,7 +22,7 @@
         <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Document</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -54,7 +54,7 @@
 
 ## 💡 RAGFlow란?
 
-[RAGFlow](https://ragflow.io/) 는 최첨단 RAG(Retrieval-Augmented Generation)와 Agent 기능을 융합하여 대규모 언어 모델(LLM)을 위한 우수한 컨텍스트 계층을 생성하는 선도적인 오픈소스 RAG 엔진입니다. 모든 규모의 기업에 적용 가능한 효율적인 RAG 워크플로를 제공하며, 통합 컨텍스트 엔진과 사전 구축된 Agent 템플릿을 통해 개발자들이 복잡한 데이터를 예외적인 효율성과 정밀도로 고급 구현도의 프로덕션 준비 완료 AI 시스템으로 변환할 수 있도록 지원합니다.
+[RAGFlow](https://ragflow.io/) 는 최첨단 [RAG](https://ragflow.io/basics/what-is-rag)(Retrieval-Augmented Generation)와 Agent 기능을 융합하여 대규모 언어 모델(LLM)을 위한 우수한 컨텍스트 계층을 생성하는 선도적인 오픈소스 RAG 엔진입니다. 모든 규모의 기업에 적용 가능한 효율적인 RAG 워크플로를 제공하며, 통합 [컨텍스트 엔진](https://ragflow.io/basics/what-is-agent-context-engine)과 사전 구축된 Agent 템플릿을 통해 개발자들이 복잡한 데이터를 예외적인 효율성과 정밀도로 고급 구현도의 프로덕션 준비 완료 AI 시스템으로 변환할 수 있도록 지원합니다.
 
 ## 🎮 데모
 
@@ -67,6 +67,7 @@
 
 ## 🔥 업데이트
 
+- 2025-12-26 AI 에이전트의 '메모리' 기능 지원.
 - 2025-11-19 Gemini 3 Pro를 지원합니다.
 - 2025-11-12 Confluence, S3, Notion, Discord, Google Drive에서 데이터 동기화를 지원합니다.
 - 2025-10-23 문서 파싱 방법으로 MinerU 및 Docling을 지원합니다.
@@ -169,12 +170,12 @@
 > 모든 Docker 이미지는 x86 플랫폼을 위해 빌드되었습니다. 우리는 현재 ARM64 플랫폼을 위한 Docker 이미지를 제공하지 않습니다.
 > ARM64 플랫폼을 사용 중이라면, [시스템과 호환되는 Docker 이미지를 빌드하려면 이 가이드를 사용해 주세요](https://ragflow.io/docs/dev/build_docker_image).
 
-   > 아래 명령어는 RAGFlow Docker 이미지의 v0.22.1 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.22.1과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오.
+   > 아래 명령어는 RAGFlow Docker 이미지의 v0.23.1 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.23.1과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오.
 
    ```bash
    $ cd ragflow/docker
-   
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # Optional: use a stable tag (see releases: https://github.com/infiniflow/ragflow/releases)
    # 이 단계는 코드의 entrypoint.sh 파일이 Docker 이미지 버전과 일치하도록 보장합니다.
 
@@ -188,10 +189,10 @@
 
 > 참고: `v0.22.0` 이전 버전에서는 embedding 모델이 포함된 이미지와 embedding 모델이 포함되지 않은 slim 이미지를 모두 제공했습니다. 자세한 내용은 다음과 같습니다:
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > `v0.22.0`부터는 slim 에디션만 배포하며 이미지 태그에 **-slim** 접미사를 더 이상 붙이지 않습니다.
 
@@ -213,7 +214,7 @@
     * Running on all addresses (0.0.0.0)
    ```
 
-   > 만약 확인 단계를 건너뛰고 바로 RAGFlow에 로그인하면, RAGFlow가 완전히 초기화되지 않았기 때문에 브라우저에서 `network anormal` 오류가 발생할 수 있습니다.
+   > 만약 확인 단계를 건너뛰고 바로 RAGFlow에 로그인하면, RAGFlow가 완전히 초기화되지 않았기 때문에 브라우저에서 `network abnormal` 오류가 발생할 수 있습니다.
 
 2. 웹 브라우저에 서버의 IP 주소를 입력하고 RAGFlow에 로그인하세요.
    > 기본 설정을 사용할 경우, `http://IP_OF_YOUR_MACHINE`만 입력하면 됩니다 (포트 번호는 제외). 기본 HTTP 서비스 포트 `80`은 기본 구성으로 사용할 때 생략할 수 있습니다.
@@ -270,6 +271,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+프록시 환경인 경우, 프록시 인수를 전달할 수 있습니다：
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 소스 코드로 서비스를 시작합니다.
 
 1. `uv` 와 `pre-commit` 을 설치하거나, 이미 설치된 경우 이 단계를 건너뜁니다:
@@ -362,7 +372,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 로드맵
 
-[RAGFlow 로드맵 2025](https://github.com/infiniflow/ragflow/issues/4214)을 확인하세요.
+[RAGFlow 로드맵 2026](https://github.com/infiniflow/ragflow/issues/12241)을 확인하세요.
 
 ## 🏄 커뮤니티
 

README_pt_br.md

@@ -22,7 +22,7 @@
         <img alt="Badge Estático" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Última%20Relese" alt="Última Versão">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Documentação</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -73,7 +73,7 @@
 
 ## 💡 O que é o RAGFlow?
 
-[RAGFlow](https://ragflow.io/) é um mecanismo de RAG (Retrieval-Augmented Generation) open-source líder que fusiona tecnologias RAG de ponta com funcionalidades Agent para criar uma camada contextual superior para LLMs. Oferece um fluxo de trabalho RAG otimizado adaptável a empresas de qualquer escala. Alimentado por um motor de contexto convergente e modelos Agent pré-construídos, o RAGFlow permite que desenvolvedores transformem dados complexos em sistemas de IA de alta fidelidade e pronto para produção com excepcional eficiência e precisão.
+[RAGFlow](https://ragflow.io/) é um mecanismo de [RAG](https://ragflow.io/basics/what-is-rag) (Retrieval-Augmented Generation) open-source líder que fusiona tecnologias RAG de ponta com funcionalidades Agent para criar uma camada contextual superior para LLMs. Oferece um fluxo de trabalho RAG otimizado adaptável a empresas de qualquer escala. Alimentado por [um motor de contexto](https://ragflow.io/basics/what-is-agent-context-engine) convergente e modelos Agent pré-construídos, o RAGFlow permite que desenvolvedores transformem dados complexos em sistemas de IA de alta fidelidade e pronto para produção com excepcional eficiência e precisão.
 
 ## 🎮 Demo
 
@@ -86,6 +86,7 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Últimas Atualizações
 
+- 26-12-2025 Suporte à função 'Memória' para agentes de IA.
 - 19-11-2025 Suporta Gemini 3 Pro.
 - 12-11-2025 Suporta a sincronização de dados do Confluence, S3, Notion, Discord e Google Drive.
 - 23-10-2025 Suporta MinerU e Docling como métodos de análise de documentos.
@@ -187,12 +188,12 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
 > Todas as imagens Docker são construídas para plataformas x86. Atualmente, não oferecemos imagens Docker para ARM64.
 > Se você estiver usando uma plataforma ARM64, por favor, utilize [este guia](https://ragflow.io/docs/dev/build_docker_image) para construir uma imagem Docker compatível com o seu sistema.
 
-    > O comando abaixo baixa a edição`v0.22.1` da imagem Docker do RAGFlow. Consulte a tabela a seguir para descrições de diferentes edições do RAGFlow. Para baixar uma edição do RAGFlow diferente da `v0.22.1`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor.
+    > O comando abaixo baixa a edição`v0.23.1` da imagem Docker do RAGFlow. Consulte a tabela a seguir para descrições de diferentes edições do RAGFlow. Para baixar uma edição do RAGFlow diferente da `v0.23.1`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor.
 
 ```bash
    $ cd ragflow/docker
-   
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # Opcional: use uma tag estável (veja releases: https://github.com/infiniflow/ragflow/releases)
    # Esta etapa garante que o arquivo entrypoint.sh no código corresponda à versão da imagem do Docker.
 
@@ -206,10 +207,10 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
 
 > Nota: Antes da `v0.22.0`, fornecíamos imagens com modelos de embedding e imagens slim sem modelos de embedding. Detalhes a seguir:
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > A partir da `v0.22.0`, distribuímos apenas a edição slim e não adicionamos mais o sufixo **-slim** às tags das imagens.
 
@@ -231,7 +232,7 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
     * Rodando em todos os endereços (0.0.0.0)
    ```
 
-   > Se você pular essa etapa de confirmação e acessar diretamente o RAGFlow, seu navegador pode exibir um erro `network anormal`, pois, nesse momento, seu RAGFlow pode não estar totalmente inicializado.
+   > Se você pular essa etapa de confirmação e acessar diretamente o RAGFlow, seu navegador pode exibir um erro `network abnormal`, pois, nesse momento, seu RAGFlow pode não estar totalmente inicializado.
    >
 5. No seu navegador, insira o endereço IP do seu servidor e faça login no RAGFlow.
 
@@ -293,6 +294,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+Se você estiver atrás de um proxy, pode passar argumentos de proxy:
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 Lançar o serviço a partir do código-fonte para desenvolvimento
 
 1. Instale o `uv` e o `pre-commit`, ou pule esta etapa se eles já estiverem instalados:
@@ -375,7 +385,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 Roadmap
 
-Veja o [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214)
+Veja o [RAGFlow Roadmap 2026](https://github.com/infiniflow/ragflow/issues/12241)
 
 ## 🏄 Comunidade
 

README_tzh.md

@@ -22,7 +22,7 @@
         <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Document</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -72,7 +72,7 @@
 
 ## 💡 RAGFlow 是什麼？
 
-[RAGFlow](https://ragflow.io/) 是一款領先的開源 RAG（Retrieval-Augmented Generation）引擎，通過融合前沿的 RAG 技術與 Agent 能力，為大型語言模型提供卓越的上下文層。它提供可適配任意規模企業的端到端 RAG 工作流，憑藉融合式上下文引擎與預置的 Agent 模板，助力開發者以極致效率與精度將複雜數據轉化為高可信、生產級的人工智能系統。
+[RAGFlow](https://ragflow.io/) 是一款領先的開源 [RAG](https://ragflow.io/basics/what-is-rag)（Retrieval-Augmented Generation）引擎，通過融合前沿的 RAG 技術與 Agent 能力，為大型語言模型提供卓越的上下文層。它提供可適配任意規模企業的端到端 RAG 工作流，憑藉融合式[上下文引擎](https://ragflow.io/basics/what-is-agent-context-engine)與預置的 Agent 模板，助力開發者以極致效率與精度將複雜數據轉化為高可信、生產級的人工智能系統。
 
 ## 🎮 Demo 試用
 
@@ -85,15 +85,16 @@
 
 ## 🔥 近期更新
 
-- 2025-11-19 支援 Gemini 3 Pro.
+- 2025-12-26 支援AI代理的「記憶」功能。
+- 2025-11-19 支援 Gemini 3 Pro。
 - 2025-11-12 支援從 Confluence、S3、Notion、Discord、Google Drive 進行資料同步。
 - 2025-10-23 支援 MinerU 和 Docling 作為文件解析方法。
 - 2025-10-15 支援可編排的資料管道。
 - 2025-08-08 支援 OpenAI 最新的 GPT-5 系列模型。
-- 2025-08-01 支援 agentic workflow 和 MCP
+- 2025-08-01 支援 agentic workflow 和 MCP。
 - 2025-05-23 為 Agent 新增 Python/JS 程式碼執行器元件。
 - 2025-05-05 支援跨語言查詢。
-- 2025-03-19 PDF和DOCX中的圖支持用多模態大模型去解析得到描述.
+- 2025-03-19 PDF和DOCX中的圖支持用多模態大模型去解析得到描述。
 - 2024-12-18 升級了 DeepDoc 的文檔佈局分析模型。
 - 2024-08-22 支援用 RAG 技術實現從自然語言到 SQL 語句的轉換。
 
@@ -124,7 +125,7 @@
 
 ### 🍔 **相容各類異質資料來源**
 
-- 支援豐富的文件類型，包括 Word 文件、PPT、excel 表格、txt 檔案、圖片、PDF、影印件、影印件、結構化資料、網頁等。
+- 支援豐富的文件類型，包括 Word 文件、PPT、excel 表格、txt 檔案、圖片、PDF、影印件、複印件、結構化資料、網頁等。
 
 ### 🛀 **全程無憂、自動化的 RAG 工作流程**
 
@@ -186,12 +187,12 @@
 > 所有 Docker 映像檔都是為 x86 平台建置的。目前，我們不提供 ARM64 平台的 Docker 映像檔。
 > 如果您使用的是 ARM64 平台，請使用 [這份指南](https://ragflow.io/docs/dev/build_docker_image) 來建置適合您系統的 Docker 映像檔。
 
-> 執行以下指令會自動下載 RAGFlow Docker 映像 `v0.22.1`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.22.1` 的 Docker 映像，請在執行 `docker compose` 啟動服務之前先更新 **docker/.env** 檔案內的 `RAGFLOW_IMAGE` 變數。
+> 執行以下指令會自動下載 RAGFlow Docker 映像 `v0.23.1`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.23.1` 的 Docker 映像，請在執行 `docker compose` 啟動服務之前先更新 **docker/.env** 檔案內的 `RAGFLOW_IMAGE` 變數。
 
 ```bash
    $ cd ragflow/docker
-   
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # 可選：使用穩定版標籤（查看發佈：https://github.com/infiniflow/ragflow/releases）
    # 此步驟確保程式碼中的 entrypoint.sh 檔案與 Docker 映像版本一致。
 
@@ -205,10 +206,10 @@
 
 > 注意：在 `v0.22.0` 之前的版本，我們會同時提供包含 embedding 模型的映像和不含 embedding 模型的 slim 映像。具體如下：
 
-| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-| ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+|-------------------|-----------------|-----------------------|----------------|
+| v0.21.1           | &approx;9       | ✔️                    | Stable release |
+| v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
 > 從 `v0.22.0` 開始，我們只發佈 slim 版本，並且不再在映像標籤後附加 **-slim** 後綴。
 
@@ -236,7 +237,7 @@
     * Running on all addresses (0.0.0.0)
    ```
 
-   > 如果您跳過這一步驟系統確認步驟就登入 RAGFlow，你的瀏覽器有可能會提示 `network anormal` 或 `網路異常`，因為 RAGFlow 可能並未完全啟動成功。
+   > 如果您跳過這一步驟系統確認步驟就登入 RAGFlow，你的瀏覽器有可能會提示 `network abnormal` 或 `網路異常`，因為 RAGFlow 可能並未完全啟動成功。
    >
 5. 在你的瀏覽器中輸入你的伺服器對應的 IP 位址並登入 RAGFlow。
 
@@ -302,6 +303,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+若您位於代理環境，可傳遞代理參數：
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 以原始碼啟動服務
 
 1. 安裝 `uv` 和 `pre-commit`。如已安裝，可跳過此步驟：
@@ -389,7 +399,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 路線圖
 
-詳見 [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214) 。
+詳見 [RAGFlow Roadmap 2026](https://github.com/infiniflow/ragflow/issues/12241) 。
 
 ## 🏄 開源社群
 

README_zh.md

@@ -22,7 +22,7 @@
         <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
     </a>
     <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.22.1">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.23.1">
     </a>
     <a href="https://github.com/infiniflow/ragflow/releases/latest">
         <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@@ -37,7 +37,7 @@
 
 <h4 align="center">
   <a href="https://ragflow.io/docs/dev/">Document</a> |
-  <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/12241">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
   <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
@@ -72,7 +72,7 @@
 
 ## 💡 RAGFlow 是什么？
 
-[RAGFlow](https://ragflow.io/) 是一款领先的开源检索增强生成（RAG）引擎，通过融合前沿的 RAG 技术与 Agent 能力，为大型语言模型提供卓越的上下文层。它提供可适配任意规模企业的端到端 RAG 工作流，凭借融合式上下文引擎与预置的 Agent 模板，助力开发者以极致效率与精度将复杂数据转化为高可信、生产级的人工智能系统。
+[RAGFlow](https://ragflow.io/) 是一款领先的开源检索增强生成（[RAG](https://ragflow.io/basics/what-is-rag)）引擎，通过融合前沿的 RAG 技术与 Agent 能力，为大型语言模型提供卓越的上下文层。它提供可适配任意规模企业的端到端 RAG 工作流，凭借融合式[上下文引擎](https://ragflow.io/basics/what-is-agent-context-engine)与预置的 Agent 模板，助力开发者以极致效率与精度将复杂数据转化为高可信、生产级的人工智能系统。
 
 ## 🎮 Demo 试用
 
@@ -85,7 +85,8 @@
 
 ## 🔥 近期更新
 
-- 2025-11-19 支持 Gemini 3 Pro.
+- 2025-12-26 支持AI代理的“记忆”功能。
+- 2025-11-19 支持 Gemini 3 Pro。
 - 2025-11-12 支持从 Confluence、S3、Notion、Discord、Google Drive 进行数据同步。
 - 2025-10-23 支持 MinerU 和 Docling 作为文档解析方法。
 - 2025-10-15 支持可编排的数据管道。
@@ -93,7 +94,7 @@
 - 2025-08-01 支持 agentic workflow 和 MCP。
 - 2025-05-23 Agent 新增 Python/JS 代码执行器组件。
 - 2025-05-05 支持跨语言查询。
-- 2025-03-19 PDF 和 DOCX 中的图支持用多模态大模型去解析得到描述.
+- 2025-03-19 PDF 和 DOCX 中的图支持用多模态大模型去解析得到描述。
 - 2024-12-18 升级了 DeepDoc 的文档布局分析模型。
 - 2024-08-22 支持用 RAG 技术实现从自然语言到 SQL 语句的转换。
 
@@ -187,12 +188,12 @@
 > 请注意，目前官方提供的所有 Docker 镜像均基于 x86 架构构建，并不提供基于 ARM64 的 Docker 镜像。
 > 如果你的操作系统是 ARM64 架构，请参考[这篇文档](https://ragflow.io/docs/dev/build_docker_image)自行构建 Docker 镜像。
 
-   > 运行以下命令会自动下载 RAGFlow Docker 镜像 `v0.22.1`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.22.1` 的 Docker 镜像，请在运行 `docker compose` 启动服务之前先更新 **docker/.env** 文件内的 `RAGFLOW_IMAGE` 变量。
+   > 运行以下命令会自动下载 RAGFlow Docker 镜像 `v0.23.1`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.23.1` 的 Docker 镜像，请在运行 `docker compose` 启动服务之前先更新 **docker/.env** 文件内的 `RAGFLOW_IMAGE` 变量。
 
    ```bash
    $ cd ragflow/docker
-   
-   # git checkout v0.22.1
+
+   # git checkout v0.23.1
    # 可选：使用稳定版本标签（查看发布：https://github.com/infiniflow/ragflow/releases）
    # 这一步确保代码中的 entrypoint.sh 文件与 Docker 镜像的版本保持一致。
 
@@ -203,13 +204,13 @@
    # sed -i '1i DEVICE=gpu' .env
    # docker compose -f docker-compose.yml up -d
    ```
-   
+
    > 注意：在 `v0.22.0` 之前的版本，我们会同时提供包含 embedding 模型的镜像和不含 embedding 模型的 slim 镜像。具体如下：
 
-   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.21.1           | &approx;9       | ✔️                    | Stable release           |
-   | v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
+   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?        |
+   |-------------------|-----------------|-----------------------|----------------|
+   | v0.21.1           | &approx;9       | ✔️                    | Stable release |
+   | v0.21.1-slim      | &approx;2       | ❌                     | Stable release |
 
    > 从 `v0.22.0` 开始，我们只发布 slim 版本，并且不再在镜像标签后附加 **-slim** 后缀。
 
@@ -237,7 +238,7 @@
     * Running on all addresses (0.0.0.0)
    ```
 
-   > 如果您在没有看到上面的提示信息出来之前，就尝试登录 RAGFlow，你的浏览器有可能会提示 `network anormal` 或 `网络异常`。
+   > 如果您在没有看到上面的提示信息出来之前，就尝试登录 RAGFlow，你的浏览器有可能会提示 `network abnormal` 或 `网络异常`。
 
 5. 在你的浏览器中输入你的服务器对应的 IP 地址并登录 RAGFlow。
    > 上面这个例子中，您只需输入 http://IP_OF_YOUR_MACHINE 即可：未改动过配置则无需输入端口（默认的 HTTP 服务端口 80）。
@@ -301,6 +302,15 @@ cd ragflow/
 docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
+如果您处在代理环境下，可以传递代理参数：
+
+```bash
+docker build --platform linux/amd64 \
+  --build-arg http_proxy=http://YOUR_PROXY:PORT \
+  --build-arg https_proxy=http://YOUR_PROXY:PORT \
+  -f Dockerfile -t infiniflow/ragflow:nightly .
+```
+
 ## 🔨 以源代码启动服务
 
 1. 安装 `uv` 和 `pre-commit`。如已经安装，可跳过本步骤：
@@ -392,7 +402,7 @@ docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly
 
 ## 📜 路线图
 
-详见 [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214) 。
+详见 [RAGFlow Roadmap 2026](https://github.com/infiniflow/ragflow/issues/12241) 。
 
 ## 🏄 开源社区
 

SECURITY.md

@@ -6,8 +6,8 @@ Use this section to tell people about which versions of your project are
 currently being supported with security updates.
 
 | Version | Supported          |
-| ------- | ------------------ |
-| <=0.7.0   | :white_check_mark: |
+|---------|--------------------|
+| <=0.7.0 | :white_check_mark: |
 
 ## Reporting a Vulnerability
 

admin/client/README.md

@@ -48,7 +48,7 @@ It consists of a server-side Service and a command-line client (CLI), both imple
 1.  Ensure the Admin Service is running.
 2.  Install ragflow-cli.
     ```bash
-    pip install ragflow-cli==0.22.1
+    pip install ragflow-cli==0.23.1
     ```
 3.  Launch the CLI client:
     ```bash

admin/client/admin_client.py

@@ -16,14 +16,14 @@
 
 import argparse
 import base64
-from cmd import Cmd
-
-from Cryptodome.PublicKey import RSA
-from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
-from typing import Dict, List, Any
-from lark import Lark, Transformer, Tree
-import requests
 import getpass
+from cmd import Cmd
+from typing import Any, Dict, List
+
+import requests
+from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
+from Cryptodome.PublicKey import RSA
+from lark import Lark, Transformer, Tree
 
 GRAMMAR = r"""
 start: command
@@ -53,6 +53,8 @@ sql_command: list_services
            | alter_user_role
            | show_user_permission
            | show_version
+           | grant_admin
+           | revoke_admin
 
 // meta command definition
 meta_command: "\\" meta_command_name [meta_args]
@@ -77,6 +79,7 @@ DROP: "DROP"i
 USER: "USER"i
 ALTER: "ALTER"i
 ACTIVE: "ACTIVE"i
+ADMIN: "ADMIN"i
 PASSWORD: "PASSWORD"i
 DATASETS: "DATASETS"i
 OF: "OF"i
@@ -123,6 +126,9 @@ revoke_permission: REVOKE action_list ON identifier FROM ROLE identifier ";"
 alter_user_role: ALTER USER quoted_string SET ROLE identifier ";"
 show_user_permission: SHOW USER PERMISSION quoted_string ";"
 
+grant_admin: GRANT ADMIN quoted_string ";"
+revoke_admin: REVOKE ADMIN quoted_string ";"
+
 show_version: SHOW VERSION ";"
 
 action_list: identifier ("," identifier)*
@@ -141,7 +147,6 @@ NUMBER: /[0-9]+/
 
 
 class AdminTransformer(Transformer):
-
     def start(self, items):
         return items[0]
 
@@ -149,7 +154,7 @@ class AdminTransformer(Transformer):
         return items[0]
 
     def list_services(self, items):
-        result = {'type': 'list_services'}
+        result = {"type": "list_services"}
         return result
 
     def show_service(self, items):
@@ -236,11 +241,7 @@ class AdminTransformer(Transformer):
         action_list = items[1]
         resource = items[3]
         role_name = items[6]
-        return {
-            "type": "revoke_permission",
-            "role_name": role_name,
-            "resource": resource, "actions": action_list
-        }
+        return {"type": "revoke_permission", "role_name": role_name, "resource": resource, "actions": action_list}
 
     def alter_user_role(self, items):
         user_name = items[2]
@@ -254,6 +255,14 @@ class AdminTransformer(Transformer):
     def show_version(self, items):
         return {"type": "show_version"}
 
+    def grant_admin(self, items):
+        user_name = items[2]
+        return {"type": "grant_admin", "user_name": user_name}
+
+    def revoke_admin(self, items):
+        user_name = items[2]
+        return {"type": "revoke_admin", "user_name": user_name}
+
     def action_list(self, items):
         return items
 
@@ -264,12 +273,12 @@ class AdminTransformer(Transformer):
         # handle quoted parameter
         parsed_args = []
         for arg in args:
-            if hasattr(arg, 'value'):
+            if hasattr(arg, "value"):
                 parsed_args.append(arg.value)
             else:
                 parsed_args.append(str(arg))
 
-        return {'type': 'meta', 'command': command_name, 'args': parsed_args}
+        return {"type": "meta", "command": command_name, "args": parsed_args}
 
     def meta_command_name(self, items):
         return items[0]
@@ -279,22 +288,59 @@ class AdminTransformer(Transformer):
 
 
 def encrypt(input_string):
-    pub = '-----BEGIN PUBLIC KEY-----\nMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEArq9XTUSeYr2+N1h3Afl/z8Dse/2yD0ZGrKwx+EEEcdsBLca9Ynmx3nIB5obmLlSfmskLpBo0UACBmB5rEjBp2Q2f3AG3Hjd4B+gNCG6BDaawuDlgANIhGnaTLrIqWrrcm4EMzJOnAOI1fgzJRsOOUEfaS318Eq9OVO3apEyCCt0lOQK6PuksduOjVxtltDav+guVAA068NrPYmRNabVKRNLJpL8w4D44sfth5RvZ3q9t+6RTArpEtc5sh5ChzvqPOzKGMXW83C95TxmXqpbK6olN4RevSfVjEAgCydH6HN6OhtOQEcnrU97r9H0iZOWwbw3pVrZiUkuRD1R56Wzs2wIDAQAB\n-----END PUBLIC KEY-----'
+    pub = "-----BEGIN PUBLIC KEY-----\nMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEArq9XTUSeYr2+N1h3Afl/z8Dse/2yD0ZGrKwx+EEEcdsBLca9Ynmx3nIB5obmLlSfmskLpBo0UACBmB5rEjBp2Q2f3AG3Hjd4B+gNCG6BDaawuDlgANIhGnaTLrIqWrrcm4EMzJOnAOI1fgzJRsOOUEfaS318Eq9OVO3apEyCCt0lOQK6PuksduOjVxtltDav+guVAA068NrPYmRNabVKRNLJpL8w4D44sfth5RvZ3q9t+6RTArpEtc5sh5ChzvqPOzKGMXW83C95TxmXqpbK6olN4RevSfVjEAgCydH6HN6OhtOQEcnrU97r9H0iZOWwbw3pVrZiUkuRD1R56Wzs2wIDAQAB\n-----END PUBLIC KEY-----"
     pub_key = RSA.importKey(pub)
     cipher = Cipher_pkcs1_v1_5.new(pub_key)
-    cipher_text = cipher.encrypt(base64.b64encode(input_string.encode('utf-8')))
+    cipher_text = cipher.encrypt(base64.b64encode(input_string.encode("utf-8")))
     return base64.b64encode(cipher_text).decode("utf-8")
 
 
 def encode_to_base64(input_string):
-    base64_encoded = base64.b64encode(input_string.encode('utf-8'))
-    return base64_encoded.decode('utf-8')
+    base64_encoded = base64.b64encode(input_string.encode("utf-8"))
+    return base64_encoded.decode("utf-8")
+
+
+def show_help():
+    """Help info"""
+    help_text = """
+Commands:
+LIST SERVICES
+SHOW SERVICE <service>
+STARTUP SERVICE <service>
+SHUTDOWN SERVICE <service>
+RESTART SERVICE <service>
+LIST USERS
+SHOW USER <user>
+DROP USER <user>
+CREATE USER <user> <password>
+ALTER USER PASSWORD <user> <new_password>
+ALTER USER ACTIVE <user> <on/off>
+LIST DATASETS OF <user>
+LIST AGENTS OF <user>
+CREATE ROLE <role>
+DROP ROLE <role>
+ALTER ROLE <role> SET DESCRIPTION <description>
+LIST ROLES
+SHOW ROLE <role>
+GRANT <action_list> ON <function> TO ROLE <role>
+REVOKE <action_list> ON <function> TO ROLE <role>
+ALTER USER <user> SET ROLE <role>
+SHOW USER PERMISSION <user>
+SHOW VERSION
+GRANT ADMIN <user>
+REVOKE ADMIN <user>
+
+Meta Commands:
+\\?, \\h, \\help     Show this help
+\\q, \\quit, \\exit   Quit the CLI
+    """
+    print(help_text)
 
 
 class AdminCLI(Cmd):
     def __init__(self):
         super().__init__()
-        self.parser = Lark(GRAMMAR, start='start', parser='lalr', transformer=AdminTransformer())
+        self.parser = Lark(GRAMMAR, start="start", parser="lalr", transformer=AdminTransformer())
         self.command_history = []
         self.is_interactive = False
         self.admin_account = "admin@ragflow.io"
@@ -312,7 +358,7 @@ class AdminCLI(Cmd):
             result = self.parse_command(command)
 
             if isinstance(result, dict):
-                if 'type' in result and result.get('type') == 'empty':
+                if "type" in result and result.get("type") == "empty":
                     return False
 
             self.execute_command(result)
@@ -320,7 +366,7 @@ class AdminCLI(Cmd):
             if isinstance(result, Tree):
                 return False
 
-            if result.get('type') == 'meta' and result.get('command') in ['q', 'quit', 'exit']:
+            if result.get("type") == "meta" and result.get("command") in ["q", "quit", "exit"]:
                 return True
 
         except KeyboardInterrupt:
@@ -338,7 +384,7 @@ class AdminCLI(Cmd):
 
     def parse_command(self, command_str: str) -> dict[str, str]:
         if not command_str.strip():
-            return {'type': 'empty'}
+            return {"type": "empty"}
 
         self.command_history.append(command_str)
 
@@ -346,11 +392,11 @@ class AdminCLI(Cmd):
             result = self.parser.parse(command_str)
             return result
         except Exception as e:
-            return {'type': 'error', 'message': f'Parse error: {str(e)}'}
+            return {"type": "error", "message": f"Parse error: {str(e)}"}
 
     def verify_admin(self, arguments: dict, single_command: bool):
-        self.host = arguments['host']
-        self.port = arguments['port']
+        self.host = arguments["host"]
+        self.port = arguments["port"]
         print("Attempt to access server for admin login")
         url = f"http://{self.host}:{self.port}/api/v1/admin/login"
 
@@ -365,25 +411,21 @@ class AdminCLI(Cmd):
                 return False
 
             if single_command:
-                admin_passwd = arguments['password']
+                admin_passwd = arguments["password"]
             else:
                 admin_passwd = getpass.getpass(f"password for {self.admin_account}: ").strip()
             try:
                 self.admin_password = encrypt(admin_passwd)
-                response = self.session.post(url, json={'email': self.admin_account, 'password': self.admin_password})
+                response = self.session.post(url, json={"email": self.admin_account, "password": self.admin_password})
                 if response.status_code == 200:
                     res_json = response.json()
-                    error_code = res_json.get('code', -1)
+                    error_code = res_json.get("code", -1)
                     if error_code == 0:
-                        self.session.headers.update({
-                            'Content-Type': 'application/json',
-                            'Authorization': response.headers['Authorization'],
-                            'User-Agent': 'RAGFlow-CLI/0.22.1'
-                        })
+                        self.session.headers.update({"Content-Type": "application/json", "Authorization": response.headers["Authorization"], "User-Agent": "RAGFlow-CLI/0.23.1"})
                         print("Authentication successful.")
                         return True
                     else:
-                        error_message = res_json.get('message', 'Unknown error')
+                        error_message = res_json.get("message", "Unknown error")
                         print(f"Authentication failed: {error_message}, try again")
                         continue
                 else:
@@ -403,10 +445,14 @@ class AdminCLI(Cmd):
         for k, v in data.items():
             # display latest status
             heartbeats = sorted(v, key=lambda x: x["now"], reverse=True)
-            task_executor_list.append({
-                "task_executor_name": k,
-                **heartbeats[0],
-            } if heartbeats else {"task_executor_name": k})
+            task_executor_list.append(
+                {
+                    "task_executor_name": k,
+                    **heartbeats[0],
+                }
+                if heartbeats
+                else {"task_executor_name": k}
+            )
         return task_executor_list
 
     def _print_table_simple(self, data):
@@ -422,12 +468,7 @@ class AdminCLI(Cmd):
         col_widths = {}
 
         def get_string_width(text):
-            half_width_chars = (
-                " !\"#$%&'()*+,-./0123456789:;<=>?@"
-                "ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`"
-                "abcdefghijklmnopqrstuvwxyz{|}~"
-                "\t\n\r"
-            )
+            half_width_chars = " !\"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`abcdefghijklmnopqrstuvwxyz{|}~\t\n\r"
             width = 0
             for char in text:
                 if char in half_width_chars:
@@ -439,7 +480,7 @@ class AdminCLI(Cmd):
         for col in columns:
             max_width = get_string_width(str(col))
             for item in data:
-                value_len = get_string_width(str(item.get(col, '')))
+                value_len = get_string_width(str(item.get(col, "")))
                 if value_len > max_width:
                     max_width = value_len
             col_widths[col] = max(2, max_width)
@@ -457,16 +498,15 @@ class AdminCLI(Cmd):
         for item in data:
             row = "|"
             for col in columns:
-                value = str(item.get(col, ''))
+                value = str(item.get(col, ""))
                 if get_string_width(value) > col_widths[col]:
-                    value = value[:col_widths[col] - 3] + "..."
+                    value = value[: col_widths[col] - 3] + "..."
                 row += f" {value:<{col_widths[col] - (get_string_width(value) - len(value))}} |"
             print(row)
 
         print(separator)
 
     def run_interactive(self):
-
         self.is_interactive = True
         print("RAGFlow Admin command line interface - Type '\\?' for help, '\\q' to quit")
 
@@ -483,7 +523,7 @@ class AdminCLI(Cmd):
                 if isinstance(result, Tree):
                     continue
 
-                if result.get('type') == 'meta' and result.get('command') in ['q', 'quit', 'exit']:
+                if result.get("type") == "meta" and result.get("command") in ["q", "quit", "exit"]:
                     break
 
             except KeyboardInterrupt:
@@ -497,36 +537,30 @@ class AdminCLI(Cmd):
         self.execute_command(result)
 
     def parse_connection_args(self, args: List[str]) -> Dict[str, Any]:
-        parser = argparse.ArgumentParser(description='Admin CLI Client', add_help=False)
-        parser.add_argument('-h', '--host', default='localhost', help='Admin service host')
-        parser.add_argument('-p', '--port', type=int, default=9381, help='Admin service port')
-        parser.add_argument('-w', '--password', default='admin', type=str, help='Superuser password')
-        parser.add_argument('command', nargs='?', help='Single command')
+        parser = argparse.ArgumentParser(description="Admin CLI Client", add_help=False)
+        parser.add_argument("-h", "--host", default="localhost", help="Admin service host")
+        parser.add_argument("-p", "--port", type=int, default=9381, help="Admin service port")
+        parser.add_argument("-w", "--password", default="admin", type=str, help="Superuser password")
+        parser.add_argument("command", nargs="?", help="Single command")
         try:
             parsed_args, remaining_args = parser.parse_known_args(args)
             if remaining_args:
                 command = remaining_args[0]
-                return {
-                    'host': parsed_args.host,
-                    'port': parsed_args.port,
-                    'password': parsed_args.password,
-                    'command': command
-                }
+                return {"host": parsed_args.host, "port": parsed_args.port, "password": parsed_args.password, "command": command}
             else:
                 return {
-                    'host': parsed_args.host,
-                    'port': parsed_args.port,
+                    "host": parsed_args.host,
+                    "port": parsed_args.port,
                 }
         except SystemExit:
-            return {'error': 'Invalid connection arguments'}
+            return {"error": "Invalid connection arguments"}
 
     def execute_command(self, parsed_command: Dict[str, Any]):
-
         command_dict: dict
         if isinstance(parsed_command, Tree):
             command_dict = parsed_command.children[0]
         else:
-            if parsed_command['type'] == 'error':
+            if parsed_command["type"] == "error":
                 print(f"Error: {parsed_command['message']}")
                 return
             else:
@@ -534,56 +568,60 @@ class AdminCLI(Cmd):
 
         # print(f"Parsed command: {command_dict}")
 
-        command_type = command_dict['type']
+        command_type = command_dict["type"]
 
         match command_type:
-            case 'list_services':
+            case "list_services":
                 self._handle_list_services(command_dict)
-            case 'show_service':
+            case "show_service":
                 self._handle_show_service(command_dict)
-            case 'restart_service':
+            case "restart_service":
                 self._handle_restart_service(command_dict)
-            case 'shutdown_service':
+            case "shutdown_service":
                 self._handle_shutdown_service(command_dict)
-            case 'startup_service':
+            case "startup_service":
                 self._handle_startup_service(command_dict)
-            case 'list_users':
+            case "list_users":
                 self._handle_list_users(command_dict)
-            case 'show_user':
+            case "show_user":
                 self._handle_show_user(command_dict)
-            case 'drop_user':
+            case "drop_user":
                 self._handle_drop_user(command_dict)
-            case 'alter_user':
+            case "alter_user":
                 self._handle_alter_user(command_dict)
-            case 'create_user':
+            case "create_user":
                 self._handle_create_user(command_dict)
-            case 'activate_user':
+            case "activate_user":
                 self._handle_activate_user(command_dict)
-            case 'list_datasets':
+            case "list_datasets":
                 self._handle_list_datasets(command_dict)
-            case 'list_agents':
+            case "list_agents":
                 self._handle_list_agents(command_dict)
-            case 'create_role':
+            case "create_role":
                 self._create_role(command_dict)
-            case 'drop_role':
+            case "drop_role":
                 self._drop_role(command_dict)
-            case 'alter_role':
+            case "alter_role":
                 self._alter_role(command_dict)
-            case 'list_roles':
+            case "list_roles":
                 self._list_roles(command_dict)
-            case 'show_role':
+            case "show_role":
                 self._show_role(command_dict)
-            case 'grant_permission':
+            case "grant_permission":
                 self._grant_permission(command_dict)
-            case 'revoke_permission':
+            case "revoke_permission":
                 self._revoke_permission(command_dict)
-            case 'alter_user_role':
+            case "alter_user_role":
                 self._alter_user_role(command_dict)
-            case 'show_user_permission':
+            case "show_user_permission":
                 self._show_user_permission(command_dict)
-            case 'show_version':
+            case "show_version":
                 self._show_version(command_dict)
-            case 'meta':
+            case "grant_admin":
+                self._grant_admin(command_dict)
+            case "revoke_admin":
+                self._revoke_admin(command_dict)
+            case "meta":
                 self._handle_meta_command(command_dict)
             case _:
                 print(f"Command '{command_type}' would be executed with API")
@@ -591,29 +629,29 @@ class AdminCLI(Cmd):
     def _handle_list_services(self, command):
         print("Listing all services")
 
-        url = f'http://{self.host}:{self.port}/api/v1/admin/services'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/services"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to get all services, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_show_service(self, command):
-        service_id: int = command['number']
+        service_id: int = command["number"]
         print(f"Showing service: {service_id}")
 
-        url = f'http://{self.host}:{self.port}/api/v1/admin/services/{service_id}'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/services/{service_id}"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            res_data = res_json['data']
-            if 'status' in res_data and res_data['status'] == 'alive':
+            res_data = res_json["data"]
+            if "status" in res_data and res_data["status"] == "alive":
                 print(f"Service {res_data['service_name']} is alive, ")
-                if isinstance(res_data['message'], str):
-                    print(res_data['message'])
+                if isinstance(res_data["message"], str):
+                    print(res_data["message"])
                 else:
-                    data = self._format_service_detail_table(res_data['message'])
+                    data = self._format_service_detail_table(res_data["message"])
                     self._print_table_simple(data)
             else:
                 print(f"Service {res_data['service_name']} is down, {res_data['message']}")
@@ -621,47 +659,47 @@ class AdminCLI(Cmd):
             print(f"Fail to show service, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_restart_service(self, command):
-        service_id: int = command['number']
+        service_id: int = command["number"]
         print(f"Restart service {service_id}")
 
     def _handle_shutdown_service(self, command):
-        service_id: int = command['number']
+        service_id: int = command["number"]
         print(f"Shutdown service {service_id}")
 
     def _handle_startup_service(self, command):
-        service_id: int = command['number']
+        service_id: int = command["number"]
         print(f"Startup service {service_id}")
 
     def _handle_list_users(self, command):
         print("Listing all users")
 
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to get all users, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_show_user(self, command):
-        username_tree: Tree = command['user_name']
+        username_tree: Tree = command["user_name"]
         user_name: str = username_tree.children[0].strip("'\"")
         print(f"Showing user: {user_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            table_data = res_json['data']
-            table_data.pop('avatar')
+            table_data = res_json["data"]
+            table_data.pop("avatar")
             self._print_table_simple(table_data)
         else:
             print(f"Fail to get user {user_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_drop_user(self, command):
-        username_tree: Tree = command['user_name']
+        username_tree: Tree = command["user_name"]
         user_name: str = username_tree.children[0].strip("'\"")
         print(f"Drop user: {user_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}"
         response = self.session.delete(url)
         res_json = response.json()
         if response.status_code == 200:
@@ -670,13 +708,13 @@ class AdminCLI(Cmd):
             print(f"Fail to drop user, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_alter_user(self, command):
-        user_name_tree: Tree = command['user_name']
+        user_name_tree: Tree = command["user_name"]
         user_name: str = user_name_tree.children[0].strip("'\"")
-        password_tree: Tree = command['password']
+        password_tree: Tree = command["password"]
         password: str = password_tree.children[0].strip("'\"")
         print(f"Alter user: {user_name}, password: ******")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/password'
-        response = self.session.put(url, json={'new_password': encrypt(password)})
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/password"
+        response = self.session.put(url, json={"new_password": encrypt(password)})
         res_json = response.json()
         if response.status_code == 200:
             print(res_json["message"])
@@ -684,32 +722,29 @@ class AdminCLI(Cmd):
             print(f"Fail to alter password, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_create_user(self, command):
-        user_name_tree: Tree = command['user_name']
+        user_name_tree: Tree = command["user_name"]
         user_name: str = user_name_tree.children[0].strip("'\"")
-        password_tree: Tree = command['password']
+        password_tree: Tree = command["password"]
         password: str = password_tree.children[0].strip("'\"")
-        role: str = command['role']
+        role: str = command["role"]
         print(f"Create user: {user_name}, password: ******, role: {role}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users'
-        response = self.session.post(
-            url,
-            json={'user_name': user_name, 'password': encrypt(password), 'role': role}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users"
+        response = self.session.post(url, json={"user_name": user_name, "password": encrypt(password), "role": role})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to create user {user_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_activate_user(self, command):
-        user_name_tree: Tree = command['user_name']
+        user_name_tree: Tree = command["user_name"]
         user_name: str = user_name_tree.children[0].strip("'\"")
-        activate_tree: Tree = command['activate_status']
+        activate_tree: Tree = command["activate_status"]
         activate_status: str = activate_tree.children[0].strip("'\"")
-        if activate_status.lower() in ['on', 'off']:
+        if activate_status.lower() in ["on", "off"]:
             print(f"Alter user {user_name} activate status, turn {activate_status.lower()}.")
-            url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/activate'
-            response = self.session.put(url, json={'activate_status': activate_status})
+            url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/activate"
+            response = self.session.put(url, json={"activate_status": activate_status})
             res_json = response.json()
             if response.status_code == 200:
                 print(res_json["message"])
@@ -718,231 +753,214 @@ class AdminCLI(Cmd):
         else:
             print(f"Unknown activate status: {activate_status}.")
 
+
+    def _grant_admin(self, command):
+        user_name_tree: Tree = command["user_name"]
+        user_name: str = user_name_tree.children[0].strip("'\"")
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/admin"
+        # print(f"Grant admin: {url}")
+        # return
+        response = self.session.put(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            print(res_json["message"])
+        else:
+            print(f"Fail to grant {user_name} admin authorization, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _revoke_admin(self, command):
+        user_name_tree: Tree = command["user_name"]
+        user_name: str = user_name_tree.children[0].strip("'\"")
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/admin"
+        # print(f"Revoke admin: {url}")
+        # return
+        response = self.session.delete(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            print(res_json["message"])
+        else:
+            print(f"Fail to revoke {user_name} admin authorization, code: {res_json['code']}, message: {res_json['message']}")
+
     def _handle_list_datasets(self, command):
-        username_tree: Tree = command['user_name']
+        username_tree: Tree = command["user_name"]
         user_name: str = username_tree.children[0].strip("'\"")
         print(f"Listing all datasets of user: {user_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/datasets'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/datasets"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            table_data = res_json['data']
+            table_data = res_json["data"]
             for t in table_data:
-                t.pop('avatar')
+                t.pop("avatar")
             self._print_table_simple(table_data)
         else:
             print(f"Fail to get all datasets of {user_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_list_agents(self, command):
-        username_tree: Tree = command['user_name']
+        username_tree: Tree = command["user_name"]
         user_name: str = username_tree.children[0].strip("'\"")
         print(f"Listing all agents of user: {user_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/agents'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/agents"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            table_data = res_json['data']
+            table_data = res_json["data"]
             for t in table_data:
-                t.pop('avatar')
+                t.pop("avatar")
             self._print_table_simple(table_data)
         else:
             print(f"Fail to get all agents of {user_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _create_role(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name: str = role_name_tree.children[0].strip("'\"")
-        desc_str: str = ''
-        if 'description' in command:
-            desc_tree: Tree = command['description']
+        desc_str: str = ""
+        if "description" in command:
+            desc_tree: Tree = command["description"]
             desc_str = desc_tree.children[0].strip("'\"")
 
         print(f"create role name: {role_name}, description: {desc_str}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles'
-        response = self.session.post(
-            url,
-            json={'role_name': role_name, 'description': desc_str}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles"
+        response = self.session.post(url, json={"role_name": role_name, "description": desc_str})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to create role {role_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _drop_role(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name: str = role_name_tree.children[0].strip("'\"")
         print(f"drop role name: {role_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}"
         response = self.session.delete(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to drop role {role_name}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _alter_role(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name: str = role_name_tree.children[0].strip("'\"")
-        desc_tree: Tree = command['description']
+        desc_tree: Tree = command["description"]
         desc_str: str = desc_tree.children[0].strip("'\"")
 
         print(f"alter role name: {role_name}, description: {desc_str}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}'
-        response = self.session.put(
-            url,
-            json={'description': desc_str}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}"
+        response = self.session.put(url, json={"description": desc_str})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
-            print(
-                f"Fail to update role {role_name} with description: {desc_str}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to update role {role_name} with description: {desc_str}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _list_roles(self, command):
         print("Listing all roles")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to list roles, code: {res_json['code']}, message: {res_json['message']}")
 
     def _show_role(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name: str = role_name_tree.children[0].strip("'\"")
         print(f"show role: {role_name}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}/permission'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}/permission"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to list roles, code: {res_json['code']}, message: {res_json['message']}")
 
     def _grant_permission(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name_str: str = role_name_tree.children[0].strip("'\"")
-        resource_tree: Tree = command['resource']
+        resource_tree: Tree = command["resource"]
         resource_str: str = resource_tree.children[0].strip("'\"")
-        action_tree_list: list = command['actions']
+        action_tree_list: list = command["actions"]
         actions: list = []
         for action_tree in action_tree_list:
             action_str: str = action_tree.children[0].strip("'\"")
             actions.append(action_str)
         print(f"grant role_name: {role_name_str}, resource: {resource_str}, actions: {actions}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission'
-        response = self.session.post(
-            url,
-            json={'actions': actions, 'resource': resource_str}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission"
+        response = self.session.post(url, json={"actions": actions, "resource": resource_str})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
-            print(
-                f"Fail to grant role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to grant role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _revoke_permission(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name_str: str = role_name_tree.children[0].strip("'\"")
-        resource_tree: Tree = command['resource']
+        resource_tree: Tree = command["resource"]
         resource_str: str = resource_tree.children[0].strip("'\"")
-        action_tree_list: list = command['actions']
+        action_tree_list: list = command["actions"]
         actions: list = []
         for action_tree in action_tree_list:
             action_str: str = action_tree.children[0].strip("'\"")
             actions.append(action_str)
         print(f"revoke role_name: {role_name_str}, resource: {resource_str}, actions: {actions}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission'
-        response = self.session.delete(
-            url,
-            json={'actions': actions, 'resource': resource_str}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission"
+        response = self.session.delete(url, json={"actions": actions, "resource": resource_str})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
-            print(
-                f"Fail to revoke role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to revoke role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _alter_user_role(self, command):
-        role_name_tree: Tree = command['role_name']
+        role_name_tree: Tree = command["role_name"]
         role_name_str: str = role_name_tree.children[0].strip("'\"")
-        user_name_tree: Tree = command['user_name']
+        user_name_tree: Tree = command["user_name"]
         user_name_str: str = user_name_tree.children[0].strip("'\"")
         print(f"alter_user_role user_name: {user_name_str}, role_name: {role_name_str}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/role'
-        response = self.session.put(
-            url,
-            json={'role_name': role_name_str}
-        )
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/role"
+        response = self.session.put(url, json={"role_name": role_name_str})
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
-            print(
-                f"Fail to alter user: {user_name_str} to role {role_name_str}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to alter user: {user_name_str} to role {role_name_str}, code: {res_json['code']}, message: {res_json['message']}")
 
     def _show_user_permission(self, command):
-        user_name_tree: Tree = command['user_name']
+        user_name_tree: Tree = command["user_name"]
         user_name_str: str = user_name_tree.children[0].strip("'\"")
         print(f"show_user_permission user_name: {user_name_str}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/permission'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/permission"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
-            print(
-                f"Fail to show user: {user_name_str} permission, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to show user: {user_name_str} permission, code: {res_json['code']}, message: {res_json['message']}")
 
     def _show_version(self, command):
         print("show_version")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/version'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/version"
         response = self.session.get(url)
         res_json = response.json()
         if response.status_code == 200:
-            self._print_table_simple(res_json['data'])
+            self._print_table_simple(res_json["data"])
         else:
             print(f"Fail to show version, code: {res_json['code']}, message: {res_json['message']}")
 
     def _handle_meta_command(self, command):
-        meta_command = command['command']
-        args = command.get('args', [])
+        meta_command = command["command"]
+        args = command.get("args", [])
 
-        if meta_command in ['?', 'h', 'help']:
-            self.show_help()
-        elif meta_command in ['q', 'quit', 'exit']:
+        if meta_command in ["?", "h", "help"]:
+            show_help()
+        elif meta_command in ["q", "quit", "exit"]:
             print("Goodbye!")
         else:
             print(f"Meta command '{meta_command}' with args {args}")
 
-    def show_help(self):
-        """Help info"""
-        help_text = """
-Commands:
-  LIST SERVICES
-  SHOW SERVICE <service>
-  STARTUP SERVICE <service>
-  SHUTDOWN SERVICE <service>
-  RESTART SERVICE <service>
-  LIST USERS
-  SHOW USER <user>
-  DROP USER <user>
-  CREATE USER <user> <password>
-  ALTER USER PASSWORD <user> <new_password>
-  ALTER USER ACTIVE <user> <on/off>
-  LIST DATASETS OF <user>
-  LIST AGENTS OF <user>
-
-Meta Commands:
-  \\?, \\h, \\help     Show this help
-  \\q, \\quit, \\exit   Quit the CLI
-        """
-        print(help_text)
-
 
 def main():
     import sys
@@ -950,16 +968,16 @@ def main():
     cli = AdminCLI()
 
     args = cli.parse_connection_args(sys.argv)
-    if 'error' in args:
+    if "error" in args:
         print("Error: Invalid connection arguments")
         return
 
-    if 'command' in args:
-        if 'password' not in args:
+    if "command" in args:
+        if "password" not in args:
             print("Error: password is missing")
             return
         if cli.verify_admin(args, single_command=True):
-            command: str = args['command']
+            command: str = args["command"]
             # print(f"Run single command: {command}")
             cli.run_single_command(command)
     else:
@@ -974,5 +992,5 @@ def main():
             cli.cmdloop()
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     main()

admin/client/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ragflow-cli"
-version = "0.22.1"
+version = "0.23.1"
 description = "Admin Service's client of [RAGFlow](https://github.com/infiniflow/ragflow). The Admin Service provides user management and system monitoring. "
 authors = [{ name = "Lynn", email = "lynn_inf@hotmail.com" }]
 license = { text = "Apache License, Version 2.0" }

admin/client/uv.lock

@@ -196,7 +196,7 @@ wheels = [
 
 [[package]]
 name = "ragflow-cli"
-version = "0.22.1"
+version = "0.23.1"
 source = { virtual = "." }
 dependencies = [
     { name = "beartype" },

admin/server/routes.py

@@ -29,6 +29,11 @@ from common.versions import get_ragflow_version
 admin_bp = Blueprint('admin', __name__, url_prefix='/api/v1/admin')
 
 
+@admin_bp.route('/ping', methods=['GET'])
+def ping():
+    return success_response('PONG')
+
+
 @admin_bp.route('/login', methods=['POST'])
 def login():
     if not request.json:
@@ -153,6 +158,36 @@ def alter_user_activate_status(username):
         return error_response(str(e), 500)
 
 
+@admin_bp.route('/users/<username>/admin', methods=['PUT'])
+@login_required
+@check_admin_auth
+def grant_admin(username):
+    try:
+        if current_user.email == username:
+            return error_response(f"can't grant current user: {username}", 409)
+        msg = UserMgr.grant_admin(username)
+        return success_response(None, msg)
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+@admin_bp.route('/users/<username>/admin', methods=['DELETE'])
+@login_required
+@check_admin_auth
+def revoke_admin(username):
+    try:
+        if current_user.email == username:
+            return error_response(f"can't grant current user: {username}", 409)
+        msg = UserMgr.revoke_admin(username)
+        return success_response(None, msg)
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
 @admin_bp.route('/users/<username>', methods=['GET'])
 @login_required
 @check_admin_auth

admin/server/services.py

@@ -13,6 +13,8 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+
+import os
 import logging
 import re
 from werkzeug.security import check_password_hash
@@ -135,6 +137,38 @@ class UserMgr:
         UserService.update_user(usr.id, {"is_active": target_status})
         return f"Turn {_activate_status} user activate status successfully!"
 
+    @staticmethod
+    def grant_admin(username: str):
+        # use email to find user. check exist and unique.
+        user_list = UserService.query_user_by_email(username)
+        if not user_list:
+            raise UserNotFoundError(username)
+        elif len(user_list) > 1:
+            raise AdminException(f"Exist more than 1 user: {username}!")
+        # check activate status different from new
+        usr = user_list[0]
+        if usr.is_superuser:
+            return f"{usr} is already superuser!"
+        # update is_active
+        UserService.update_user(usr.id, {"is_superuser": True})
+        return "Grant successfully!"
+
+    @staticmethod
+    def revoke_admin(username: str):
+        # use email to find user. check exist and unique.
+        user_list = UserService.query_user_by_email(username)
+        if not user_list:
+            raise UserNotFoundError(username)
+        elif len(user_list) > 1:
+            raise AdminException(f"Exist more than 1 user: {username}!")
+        # check activate status different from new
+        usr = user_list[0]
+        if not usr.is_superuser:
+            return f"{usr} isn't superuser, yet!"
+        # update is_active
+        UserService.update_user(usr.id, {"is_superuser": False})
+        return "Revoke successfully!"
+
 
 class UserServiceMgr:
 
@@ -179,10 +213,14 @@ class ServiceMgr:
 
     @staticmethod
     def get_all_services():
+        doc_engine = os.getenv('DOC_ENGINE', 'elasticsearch')
         result = []
         configs = SERVICE_CONFIGS.configs
         for service_id, config in enumerate(configs):
             config_dict = config.to_dict()
+            if config_dict['service_type'] == 'retrieval':
+                if config_dict['extra']['retrieval_type'] != doc_engine:
+                    continue
             try:
                 service_detail = ServiceMgr.get_service_details(service_id)
                 if "status" in service_detail:

agent/canvas.py

@@ -160,7 +160,7 @@ class Graph:
         return self._tenant_id
 
     def get_value_with_variable(self,value: str) -> Any:
-        pat = re.compile(r"\{* *\{([a-zA-Z:0-9]+@[A-Za-z0-9_.]+|sys\.[A-Za-z0-9_.]+|env\.[A-Za-z0-9_.]+)\} *\}*")
+        pat = re.compile(r"\{* *\{([a-zA-Z:0-9]+@[A-Za-z0-9_.-]+|sys\.[A-Za-z0-9_.]+|env\.[A-Za-z0-9_.]+)\} *\}*")
         out_parts = []
         last = 0
 
@@ -278,7 +278,7 @@ class Graph:
 
 class Canvas(Graph):
 
-    def __init__(self, dsl: str, tenant_id=None, task_id=None):
+    def __init__(self, dsl: str, tenant_id=None, task_id=None, canvas_id=None):
         self.globals = {
             "sys.query": "",
             "sys.user_id": tenant_id,
@@ -287,6 +287,7 @@ class Canvas(Graph):
         }
         self.variables = {}
         super().__init__(dsl, tenant_id, task_id)
+        self._id = canvas_id
 
     def load(self):
         super().load()
@@ -368,8 +369,13 @@ class Canvas(Graph):
 
         if kwargs.get("webhook_payload"):
             for k, cpn in self.components.items():
-                if self.components[k]["obj"].component_name.lower() == "webhook":
-                    for kk, vv in kwargs["webhook_payload"].items():
+                if self.components[k]["obj"].component_name.lower() == "begin"  and self.components[k]["obj"]._param.mode == "Webhook":
+                    payload = kwargs.get("webhook_payload", {})
+                    if "input" in payload:
+                        self.components[k]["obj"].set_input_value("request", payload["input"])
+                    for kk, vv in payload.items():
+                        if kk == "input":
+                            continue
                         self.components[k]["obj"].set_output(kk, vv)
 
         for k in kwargs.keys():
@@ -535,6 +541,8 @@ class Canvas(Graph):
                         cite = re.search(r"\[ID:[ 0-9]+\]",  cpn_obj.output("content"))
 
                     message_end = {}
+                    if cpn_obj.get_param("status"):
+                        message_end["status"] = cpn_obj.get_param("status")
                     if isinstance(cpn_obj.output("attachment"), dict):
                         message_end["attachment"] = cpn_obj.output("attachment")
                     if cite:
@@ -714,6 +722,9 @@ class Canvas(Graph):
     def get_mode(self):
         return self.components["begin"]["obj"]._param.mode
 
+    def get_sys_query(self):
+        return self.globals.get("sys.query", "")
+
     def set_global_param(self, **kwargs):
         self.globals.update(kwargs)
 

agent/component/agent_with_tools.py

@@ -29,8 +29,8 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.mcp_server_service import MCPServerService
 from common.connection_utils import timeout
-from rag.prompts.generator import next_step_async, COMPLETE_TASK, analyze_task_async, \
-    citation_prompt, reflect_async, kb_prompt, citation_plus, full_question, message_fit_in, structured_output_prompt
+from rag.prompts.generator import next_step_async, COMPLETE_TASK, \
+    citation_prompt, kb_prompt, citation_plus, full_question, message_fit_in, structured_output_prompt
 from common.mcp_tool_call_conn import MCPToolCallSession, mcp_tool_metadata_to_openai_tool
 from agent.component.llm import LLMParam, LLM
 
@@ -84,9 +84,11 @@ class Agent(LLM, ToolBase):
     def __init__(self, canvas, id, param: LLMParam):
         LLM.__init__(self, canvas, id, param)
         self.tools = {}
-        for cpn in self._param.tools:
+        for idx, cpn in enumerate(self._param.tools):
             cpn = self._load_tool_obj(cpn)
-            self.tools[cpn.get_meta()["function"]["name"]] = cpn
+            original_name = cpn.get_meta()["function"]["name"]
+            indexed_name = f"{original_name}_{idx}"
+            self.tools[indexed_name] = cpn
 
         self.chat_mdl = LLMBundle(self._canvas.get_tenant_id(), TenantLLMService.llm_id2llm_type(self._param.llm_id), self._param.llm_id,
                                   max_retries=self._param.max_retries,
@@ -94,7 +96,12 @@ class Agent(LLM, ToolBase):
                                   max_rounds=self._param.max_rounds,
                                   verbose_tool_use=True
                                   )
-        self.tool_meta = [v.get_meta() for _,v in self.tools.items()]
+        self.tool_meta = []
+        for indexed_name, tool_obj in self.tools.items():
+            original_meta = tool_obj.get_meta()
+            indexed_meta = deepcopy(original_meta)
+            indexed_meta["function"]["name"] = indexed_name
+            self.tool_meta.append(indexed_meta)
 
         for mcp in self._param.mcp:
             _, mcp_server = MCPServerService.get_by_id(mcp["mcp_id"])
@@ -108,7 +115,8 @@ class Agent(LLM, ToolBase):
 
     def _load_tool_obj(self, cpn: dict) -> object:
         from agent.component import component_class
-        param = component_class(cpn["component_name"] + "Param")()
+        tool_name = cpn["component_name"]
+        param = component_class(tool_name + "Param")()
         param.update(cpn["params"])
         try:
             param.check()
@@ -202,7 +210,7 @@ class Agent(LLM, ToolBase):
         _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
         use_tools = []
         ans = ""
-        async for delta_ans, _tk in self._react_with_tools_streamly_async(prompt, msg, use_tools, user_defined_prompt,schema_prompt=schema_prompt):
+        async for delta_ans, _tk in self._react_with_tools_streamly_async_simple(prompt, msg, use_tools, user_defined_prompt,schema_prompt=schema_prompt):
             if self.check_if_canceled("Agent processing"):
                 return
             ans += delta_ans
@@ -246,7 +254,7 @@ class Agent(LLM, ToolBase):
         _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
         answer_without_toolcall = ""
         use_tools = []
-        async for delta_ans, _ in self._react_with_tools_streamly_async(prompt, msg, use_tools, user_defined_prompt):
+        async for delta_ans, _ in self._react_with_tools_streamly_async_simple(prompt, msg, use_tools, user_defined_prompt):
             if self.check_if_canceled("Agent streaming"):
                 return
 
@@ -264,7 +272,7 @@ class Agent(LLM, ToolBase):
         if use_tools:
             self.set_output("use_tools", use_tools)
 
-    async def _react_with_tools_streamly_async(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
+    async def _react_with_tools_streamly_async_simple(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
         token_count = 0
         tool_metas = self.tool_meta
         hist = deepcopy(history)
@@ -276,6 +284,24 @@ class Agent(LLM, ToolBase):
         else:
             user_request = history[-1]["content"]
 
+        def build_task_desc(prompt: str, user_request: str, user_defined_prompt: dict | None = None) -> str:
+            """Build a minimal task_desc by concatenating prompt, query, and tool schemas."""
+            user_defined_prompt = user_defined_prompt or {}
+
+            task_desc = (
+                "### Agent Prompt\n"
+                f"{prompt}\n\n"
+                "### User Request\n"
+                f"{user_request}\n\n"
+            )
+
+            if user_defined_prompt:
+                udp_json = json.dumps(user_defined_prompt, ensure_ascii=False, indent=2)
+                task_desc += "\n### User Defined Prompts\n" + udp_json + "\n"
+
+            return task_desc
+
+
         async def use_tool_async(name, args):
             nonlocal hist, use_tools, last_calling
             logging.info(f"{last_calling=} == {name=}")
@@ -286,9 +312,6 @@ class Agent(LLM, ToolBase):
                 "arguments": args,
                 "results": tool_response
             })
-            # self.callback("add_memory", {}, "...")
-            #self.add_memory(hist[-2]["content"], hist[-1]["content"], name, args, str(tool_response), user_defined_prompt)
-
             return name, tool_response
 
         async def complete():
@@ -326,6 +349,21 @@ class Agent(LLM, ToolBase):
 
             self.callback("gen_citations", {}, txt, elapsed_time=timer()-st)
 
+        def build_observation(tool_call_res: list[tuple]) -> str:
+            """
+            Build a Observation from tool call results.
+            No LLM involved.
+            """
+            if not tool_call_res:
+                return ""
+
+            lines = ["Observation:"]
+            for name, result in tool_call_res:
+                lines.append(f"[{name} result]")
+                lines.append(str(result))
+
+            return "\n".join(lines)
+
         def append_user_content(hist, content):
             if hist[-1]["role"] == "user":
                 hist[-1]["content"] += content
@@ -333,7 +371,7 @@ class Agent(LLM, ToolBase):
                 hist.append({"role": "user", "content": content})
 
         st = timer()
-        task_desc = await analyze_task_async(self.chat_mdl, prompt, user_request, tool_metas, user_defined_prompt)
+        task_desc = build_task_desc(prompt, user_request, user_defined_prompt)
         self.callback("analyze_task", {}, task_desc, elapsed_time=timer()-st)
         for _ in range(self._param.max_rounds + 1):
             if self.check_if_canceled("Agent streaming"):
@@ -364,7 +402,7 @@ class Agent(LLM, ToolBase):
 
                 results = await asyncio.gather(*tool_tasks) if tool_tasks else []
                 st = timer()
-                reflection = await reflect_async(self.chat_mdl, hist, results, user_defined_prompt)
+                reflection = build_observation(results)
                 append_user_content(hist, reflection)
                 self.callback("reflection", {}, str(reflection), elapsed_time=timer()-st)
 
@@ -393,6 +431,135 @@ Respond immediately with your final comprehensive answer.
         async for txt, tkcnt in complete():
             yield txt, tkcnt
 
+#     async def _react_with_tools_streamly_async(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
+#         token_count = 0
+#         tool_metas = self.tool_meta
+#         hist = deepcopy(history)
+#         last_calling = ""
+#         if len(hist) > 3:
+#             st = timer()
+#             user_request = await full_question(messages=history, chat_mdl=self.chat_mdl)
+#             self.callback("Multi-turn conversation optimization", {}, user_request, elapsed_time=timer()-st)
+#         else:
+#             user_request = history[-1]["content"]
+
+#         async def use_tool_async(name, args):
+#             nonlocal hist, use_tools, last_calling
+#             logging.info(f"{last_calling=} == {name=}")
+#             last_calling = name
+#             tool_response = await self.toolcall_session.tool_call_async(name, args)
+#             use_tools.append({
+#                 "name": name,
+#                 "arguments": args,
+#                 "results": tool_response
+#             })
+#             # self.callback("add_memory", {}, "...")
+#             #self.add_memory(hist[-2]["content"], hist[-1]["content"], name, args, str(tool_response), user_defined_prompt)
+
+#             return name, tool_response
+
+#         async def complete():
+#             nonlocal hist
+#             need2cite = self._param.cite and self._canvas.get_reference()["chunks"] and self._id.find("-->") < 0
+#             if schema_prompt:
+#                 need2cite = False
+#             cited = False
+#             if hist and hist[0]["role"] == "system":
+#                 if schema_prompt:
+#                     hist[0]["content"] += "\n" + schema_prompt
+#                 if need2cite and len(hist) < 7:
+#                     hist[0]["content"] += citation_prompt()
+#                     cited = True
+#             yield "", token_count
+
+#             _hist = hist
+#             if len(hist) > 12:
+#                 _hist = [hist[0], hist[1], *hist[-10:]]
+#             entire_txt = ""
+#             async for delta_ans in self._generate_streamly(_hist):
+#                 if not need2cite or cited:
+#                     yield delta_ans, 0
+#                 entire_txt += delta_ans
+#             if not need2cite or cited:
+#                 return
+
+#             st = timer()
+#             txt = ""
+#             async for delta_ans in self._gen_citations_async(entire_txt):
+#                 if self.check_if_canceled("Agent streaming"):
+#                     return
+#                 yield delta_ans, 0
+#                 txt += delta_ans
+
+#             self.callback("gen_citations", {}, txt, elapsed_time=timer()-st)
+
+#         def append_user_content(hist, content):
+#             if hist[-1]["role"] == "user":
+#                 hist[-1]["content"] += content
+#             else:
+#                 hist.append({"role": "user", "content": content})
+
+#         st = timer()
+#         task_desc = await analyze_task_async(self.chat_mdl, prompt, user_request, tool_metas, user_defined_prompt)
+#         self.callback("analyze_task", {}, task_desc, elapsed_time=timer()-st)
+#         for _ in range(self._param.max_rounds + 1):
+#             if self.check_if_canceled("Agent streaming"):
+#                 return
+#             response, tk = await next_step_async(self.chat_mdl, hist, tool_metas, task_desc, user_defined_prompt)
+#             # self.callback("next_step", {}, str(response)[:256]+"...")
+#             token_count += tk or 0
+#             hist.append({"role": "assistant", "content": response})
+#             try:
+#                 functions = json_repair.loads(re.sub(r"```.*", "", response))
+#                 if not isinstance(functions, list):
+#                     raise TypeError(f"List should be returned, but `{functions}`")
+#                 for f in functions:
+#                     if not isinstance(f, dict):
+#                         raise TypeError(f"An object type should be returned, but `{f}`")
+
+#                 tool_tasks = []
+#                 for func in functions:
+#                     name = func["name"]
+#                     args = func["arguments"]
+#                     if name == COMPLETE_TASK:
+#                         append_user_content(hist, f"Respond with a formal answer. FORGET(DO NOT mention) about `{COMPLETE_TASK}`. The language for the response MUST be as the same as the first user request.\n")
+#                         async for txt, tkcnt in complete():
+#                             yield txt, tkcnt
+#                         return
+
+#                     tool_tasks.append(asyncio.create_task(use_tool_async(name, args)))
+
+#                 results = await asyncio.gather(*tool_tasks) if tool_tasks else []
+#                 st = timer()
+#                 reflection = await reflect_async(self.chat_mdl, hist, results, user_defined_prompt)
+#                 append_user_content(hist, reflection)
+#                 self.callback("reflection", {}, str(reflection), elapsed_time=timer()-st)
+
+#             except Exception as e:
+#                 logging.exception(msg=f"Wrong JSON argument format in LLM ReAct response: {e}")
+#                 e = f"\nTool call error, please correct the input parameter of response format and call it again.\n *** Exception ***\n{e}"
+#                 append_user_content(hist, str(e))
+
+#         logging.warning( f"Exceed max rounds: {self._param.max_rounds}")
+#         final_instruction = f"""
+# {user_request}
+# IMPORTANT: You have reached the conversation limit. Based on ALL the information and research you have gathered so far, please provide a DIRECT and COMPREHENSIVE final answer to the original request.
+# Instructions:
+# 1. SYNTHESIZE all information collected during this conversation
+# 2. Provide a COMPLETE response using existing data - do not suggest additional research
+# 3. Structure your response as a FINAL DELIVERABLE, not a plan
+# 4. If information is incomplete, state what you found and provide the best analysis possible with available data
+# 5. DO NOT mention conversation limits or suggest further steps
+# 6. Focus on delivering VALUE with the information already gathered
+# Respond immediately with your final comprehensive answer.
+#         """
+#         if self.check_if_canceled("Agent final instruction"):
+#             return
+#         append_user_content(hist, final_instruction)
+
+#         async for txt, tkcnt in complete():
+#             yield txt, tkcnt
+
     async def _gen_citations_async(self, text):
         retrievals = self._canvas.get_reference()
         retrievals = {"chunks": list(retrievals["chunks"].values()), "doc_aggs": list(retrievals["doc_aggs"].values())}

agent/component/base.py

@@ -361,7 +361,7 @@ class ComponentParamBase(ABC):
 class ComponentBase(ABC):
     component_name: str
     thread_limiter = asyncio.Semaphore(int(os.environ.get("MAX_CONCURRENT_CHATS", 10)))
-    variable_ref_patt = r"\{* *\{([a-zA-Z_:0-9]+@[A-Za-z0-9_.]+|sys\.[A-Za-z0-9_.]+|env\.[A-Za-z0-9_.]+)\} *\}*"
+    variable_ref_patt = r"\{* *\{([a-zA-Z:0-9]+@[A-Za-z0-9_.-]+|sys\.[A-Za-z0-9_.]+|env\.[A-Za-z0-9_.]+)\} *\}*"
 
     def __str__(self):
         """

agent/component/begin.py

@@ -28,7 +28,7 @@ class BeginParam(UserFillUpParam):
         self.prologue = "Hi! I'm your smart assistant. What can I do for you?"
 
     def check(self):
-        self.check_valid_value(self.mode, "The 'mode' should be either `conversational` or `task`", ["conversational", "task"])
+        self.check_valid_value(self.mode, "The 'mode' should be either `conversational` or `task`", ["conversational", "task","Webhook"])
 
     def get_input_form(self) -> dict[str, dict]:
         return getattr(self, "inputs")

agent/component/llm.py

@@ -56,7 +56,6 @@ class LLMParam(ComponentParamBase):
         self.check_nonnegative_number(int(self.max_tokens), "[Agent] Max tokens")
         self.check_decimal_float(float(self.top_p), "[Agent] Top P")
         self.check_empty(self.llm_id, "[Agent] LLM")
-        self.check_empty(self.sys_prompt, "[Agent] System prompt")
         self.check_empty(self.prompts, "[Agent] User prompt")
 
     def gen_conf(self):

agent/component/loopitem.py

@@ -113,6 +113,10 @@ class LoopItem(ComponentBase, ABC):
                 return len(var) == 0
             elif operator == "not empty":
                 return len(var) > 0
+        elif var is None:
+            if operator == "empty":
+                return True
+            return False
 
         raise Exception(f"Invalid operator: {operator}")
 

agent/component/message.py

@@ -33,6 +33,8 @@ from common.connection_utils import timeout
 from common.misc_utils import get_uuid
 from common import settings
 
+from api.db.joint_services.memory_message_service import queue_save_to_memory_task
+
 
 class MessageParam(ComponentParamBase):
     """
@@ -166,6 +168,7 @@ class Message(ComponentBase):
 
         self.set_output("content", all_content)
         self._convert_content(all_content)
+        await self._save_to_memory(all_content)
 
     def _is_jinjia2(self, content:str) -> bool:
         patt = [
@@ -198,6 +201,7 @@ class Message(ComponentBase):
 
         self.set_output("content", content)
         self._convert_content(content)
+        self._save_to_memory(content)
 
     def thoughts(self) -> str:
         return ""
@@ -421,3 +425,16 @@ class Message(ComponentBase):
 
         except Exception as e:
             logging.error(f"Error converting content to {self._param.output_format}: {e}")
+
+    async def _save_to_memory(self, content):
+        if not hasattr(self._param, "memory_ids") or not self._param.memory_ids:
+            return True, "No memory selected."
+
+        message_dict = {
+            "user_id": self._canvas._tenant_id,
+            "agent_id": self._canvas._id,
+            "session_id": self._canvas.task_id,
+            "user_input": self._canvas.get_sys_query(),
+            "agent_response": content
+        }
+        return await queue_save_to_memory_task(self._param.memory_ids, message_dict)

agent/tools/retrieval.py

@@ -25,10 +25,12 @@ from api.db.services.document_service import DocumentService
 from common.metadata_utils import apply_meta_data_filter
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import LLMBundle
+from api.db.services.memory_service import MemoryService
+from api.db.joint_services import memory_message_service
 from common import settings
 from common.connection_utils import timeout
 from rag.app.tag import label_question
-from rag.prompts.generator import cross_languages, kb_prompt
+from rag.prompts.generator import cross_languages, kb_prompt, memory_prompt
 
 
 class RetrievalParam(ToolParamBase):
@@ -57,6 +59,7 @@ class RetrievalParam(ToolParamBase):
         self.top_n = 8
         self.top_k = 1024
         self.kb_ids = []
+        self.memory_ids = []
         self.kb_vars = []
         self.rerank_id = ""
         self.empty_response = ""
@@ -81,15 +84,7 @@ class RetrievalParam(ToolParamBase):
 class Retrieval(ToolBase, ABC):
     component_name = "Retrieval"
 
-    @timeout(int(os.environ.get("COMPONENT_EXEC_TIMEOUT", 12)))
-    async def _invoke_async(self, **kwargs):
-        if self.check_if_canceled("Retrieval processing"):
-            return
-
-        if not kwargs.get("query"):
-            self.set_output("formalized_content", self._param.empty_response)
-            return
-
+    async def _retrieve_kb(self, query_text: str):
         kb_ids: list[str] = []
         for id in self._param.kb_ids:
             if id.find("@") < 0:
@@ -124,12 +119,12 @@ class Retrieval(ToolBase, ABC):
         if self._param.rerank_id:
             rerank_mdl = LLMBundle(kbs[0].tenant_id, LLMType.RERANK, self._param.rerank_id)
 
-        vars = self.get_input_elements_from_text(kwargs["query"])
-        vars = {k:o["value"] for k,o in vars.items()}
-        query = self.string_format(kwargs["query"], vars)
+        vars = self.get_input_elements_from_text(query_text)
+        vars = {k: o["value"] for k, o in vars.items()}
+        query = self.string_format(query_text, vars)
 
-        doc_ids=[]
-        if self._param.meta_data_filter!={}:
+        doc_ids = []
+        if self._param.meta_data_filter != {}:
             metas = DocumentService.get_meta_by_kbs(kb_ids)
 
             def _resolve_manual_filter(flt: dict) -> dict:
@@ -198,18 +193,20 @@ class Retrieval(ToolBase, ABC):
 
             if self._param.toc_enhance:
                 chat_mdl = LLMBundle(self._canvas._tenant_id, LLMType.CHAT)
-                cks = settings.retriever.retrieval_by_toc(query, kbinfos["chunks"], [kb.tenant_id for kb in kbs], chat_mdl, self._param.top_n)
+                cks = settings.retriever.retrieval_by_toc(query, kbinfos["chunks"], [kb.tenant_id for kb in kbs],
+                                                          chat_mdl, self._param.top_n)
                 if self.check_if_canceled("Retrieval processing"):
                     return
                 if cks:
                     kbinfos["chunks"] = cks
-            kbinfos["chunks"] = settings.retriever.retrieval_by_children(kbinfos["chunks"], [kb.tenant_id for kb in kbs])
+            kbinfos["chunks"] = settings.retriever.retrieval_by_children(kbinfos["chunks"],
+                                                                         [kb.tenant_id for kb in kbs])
             if self._param.use_kg:
-                ck = settings.kg_retriever.retrieval(query,
-                                                       [kb.tenant_id for kb in kbs],
-                                                       kb_ids,
-                                                       embd_mdl,
-                                                       LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT))
+                ck = await settings.kg_retriever.retrieval(query,
+                                                     [kb.tenant_id for kb in kbs],
+                                                     kb_ids,
+                                                     embd_mdl,
+                                                     LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT))
                 if self.check_if_canceled("Retrieval processing"):
                     return
                 if ck["content_with_weight"]:
@@ -218,7 +215,8 @@ class Retrieval(ToolBase, ABC):
             kbinfos = {"chunks": [], "doc_aggs": []}
 
         if self._param.use_kg and kbs:
-            ck = settings.kg_retriever.retrieval(query, [kb.tenant_id for kb in kbs], filtered_kb_ids, embd_mdl, LLMBundle(kbs[0].tenant_id, LLMType.CHAT))
+            ck = await settings.kg_retriever.retrieval(query, [kb.tenant_id for kb in kbs], filtered_kb_ids, embd_mdl,
+                                                 LLMBundle(kbs[0].tenant_id, LLMType.CHAT))
             if self.check_if_canceled("Retrieval processing"):
                 return
             if ck["content_with_weight"]:
@@ -248,6 +246,54 @@ class Retrieval(ToolBase, ABC):
 
         return form_cnt
 
+    async def _retrieve_memory(self, query_text: str):
+        memory_ids: list[str] = [memory_id for memory_id in self._param.memory_ids]
+        memory_list = MemoryService.get_by_ids(memory_ids)
+        if not memory_list:
+            raise Exception("No memory is selected.")
+
+        embd_names = list({memory.embd_id for memory in memory_list})
+        assert len(embd_names) == 1, "Memory use different embedding models."
+
+        vars = self.get_input_elements_from_text(query_text)
+        vars = {k: o["value"] for k, o in vars.items()}
+        query = self.string_format(query_text, vars)
+        # query message
+        message_list = memory_message_service.query_message({"memory_id": memory_ids}, {
+            "query": query,
+            "similarity_threshold": self._param.similarity_threshold,
+            "keywords_similarity_weight": self._param.keywords_similarity_weight,
+            "top_n": self._param.top_n
+        })
+        if not message_list:
+            self.set_output("formalized_content", self._param.empty_response)
+            return ""
+        formated_content = "\n".join(memory_prompt(message_list, 200000))
+        # set formalized_content output
+        self.set_output("formalized_content", formated_content)
+
+        return formated_content
+
+    @timeout(int(os.environ.get("COMPONENT_EXEC_TIMEOUT", 12)))
+    async def _invoke_async(self, **kwargs):
+        if self.check_if_canceled("Retrieval processing"):
+            return
+        if not kwargs.get("query"):
+            self.set_output("formalized_content", self._param.empty_response)
+            return
+
+        if hasattr(self._param, "retrieval_from") and self._param.retrieval_from == "dataset":
+            return await self._retrieve_kb(kwargs["query"])
+        elif hasattr(self._param, "retrieval_from") and self._param.retrieval_from == "memory":
+            return await self._retrieve_memory(kwargs["query"])
+        elif self._param.kb_ids:
+            return await self._retrieve_kb(kwargs["query"])
+        elif hasattr(self._param, "memory_ids") and self._param.memory_ids:
+            return await self._retrieve_memory(kwargs["query"])
+        else:
+            self.set_output("formalized_content", self._param.empty_response)
+            return
+
     @timeout(int(os.environ.get("COMPONENT_EXEC_TIMEOUT", 12)))
     def _invoke(self, **kwargs):
         return asyncio.run(self._invoke_async(**kwargs))

api/apps/__init__.py

@@ -38,7 +38,6 @@ settings.init_settings()
 
 __all__ = ["app"]
 
-
 app = Quart(__name__)
 app = cors(app, allow_origin="*")
 
@@ -103,12 +102,13 @@ from werkzeug.local import LocalProxy
 T = TypeVar("T")
 P = ParamSpec("P")
 
+
 def _load_user():
     jwt = Serializer(secret_key=settings.SECRET_KEY)
     authorization = request.headers.get("Authorization")
     g.user = None
     if not authorization:
-        return
+        return None
 
     try:
         access_token = str(jwt.loads(authorization))
@@ -164,7 +164,7 @@ def login_required(func: Callable[P, Awaitable[T]]) -> Callable[P, Awaitable[T]]
 
     @wraps(func)
     async def wrapper(*args: P.args, **kwargs: P.kwargs) -> T:
-        if not current_user:# or not session.get("_user_id"):
+        if not current_user:  # or not session.get("_user_id"):
             raise Unauthorized()
         else:
             return await current_app.ensure_async(func)(*args, **kwargs)
@@ -228,6 +228,7 @@ def logout_user():
 
     return True
 
+
 def search_pages_path(page_path):
     app_path_list = [
         path for path in page_path.glob("*_app.py") if not path.name.startswith(".")
@@ -274,6 +275,16 @@ client_urls_prefix = [
 ]
 
 
+@app.errorhandler(404)
+async def not_found(error):
+    error_msg: str = f"The requested URL {request.path} was not found"
+    logging.error(error_msg)
+    return {
+        "error": "Not Found",
+        "message": error_msg,
+    }, 404
+
+
 @app.teardown_request
 def _db_close(exception):
     if exception:

api/apps/canvas_app.py

@@ -153,7 +153,7 @@ async def run():
         return get_json_result(data={"message_id": task_id})
 
     try:
-        canvas = Canvas(cvs.dsl, current_user.id)
+        canvas = Canvas(cvs.dsl, current_user.id, canvas_id=cvs.id)
     except Exception as e:
         return server_error_response(e)
 
@@ -192,7 +192,7 @@ async def rerun():
     if 0 < doc["progress"] < 1:
         return get_data_error_result(message=f"`{doc['name']}` is processing...")
 
-    if settings.docStoreConn.indexExist(search.index_name(current_user.id), doc["kb_id"]):
+    if settings.docStoreConn.index_exist(search.index_name(current_user.id), doc["kb_id"]):
         settings.docStoreConn.delete({"doc_id": doc["id"]}, search.index_name(current_user.id), doc["kb_id"])
     doc["progress_msg"] = ""
     doc["chunk_num"] = 0
@@ -232,7 +232,7 @@ async def reset():
         if not e:
             return get_data_error_result(message="canvas not found.")
 
-        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id)
+        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id, canvas_id=user_canvas.id)
         canvas.reset()
         req["dsl"] = json.loads(str(canvas))
         UserCanvasService.update_by_id(req["id"], {"dsl": req["dsl"]})
@@ -270,7 +270,7 @@ def input_form():
                 data=False, message='Only owner of canvas authorized for this operation.',
                 code=RetCode.OPERATING_ERROR)
 
-        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id)
+        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id, canvas_id=user_canvas.id)
         return get_json_result(data=canvas.get_component_input_form(cpn_id))
     except Exception as e:
         return server_error_response(e)
@@ -287,7 +287,7 @@ async def debug():
             code=RetCode.OPERATING_ERROR)
     try:
         e, user_canvas = UserCanvasService.get_by_id(req["id"])
-        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id)
+        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id, canvas_id=user_canvas.id)
         canvas.reset()
         canvas.message_id = get_uuid()
         component = canvas.get_component(req["component_id"])["obj"]

api/apps/chunk_app.py

@@ -76,6 +76,7 @@ async def list_chunk():
                 "image_id": sres.field[id].get("img_id", ""),
                 "available_int": int(sres.field[id].get("available_int", 1)),
                 "positions": sres.field[id].get("position_int", []),
+                "doc_type_kwd": sres.field[id].get("doc_type_kwd")
             }
             assert isinstance(d["positions"], list)
             assert len(d["positions"]) == 0 or (isinstance(d["positions"][0], list) and len(d["positions"][0]) == 5)
@@ -178,8 +179,9 @@ async def set():
             # update image
             image_base64 = req.get("image_base64", None)
             if image_base64:
+                bkt, name = req.get("img_id", "-").split("-")
                 image_binary = base64.b64decode(image_base64)
-                settings.STORAGE_IMPL.put(req["doc_id"], req["chunk_id"], image_binary)
+                settings.STORAGE_IMPL.put(bkt, name, image_binary)
             return get_json_result(data=True)
 
         return await asyncio.to_thread(_set_sync)
@@ -379,7 +381,7 @@ async def retrieval_test():
                                rank_feature=labels
                                )
         if use_kg:
-            ck = settings.kg_retriever.retrieval(_question,
+            ck = await settings.kg_retriever.retrieval(_question,
                                                    tenant_ids,
                                                    kb_ids,
                                                    embd_mdl,

api/apps/document_app.py

@@ -234,6 +234,10 @@ async def list_docs():
 
     req = await get_request_json()
 
+    return_empty_metadata = req.get("return_empty_metadata", False)
+    if isinstance(return_empty_metadata, str):
+        return_empty_metadata = return_empty_metadata.lower() == "true"
+
     run_status = req.get("run_status", [])
     if run_status:
         invalid_status = {s for s in run_status if s not in VALID_TASK_STATUS}
@@ -248,18 +252,73 @@ async def list_docs():
 
     suffix = req.get("suffix", [])
     metadata_condition = req.get("metadata_condition", {}) or {}
-    if metadata_condition and not isinstance(metadata_condition, dict):
-        return get_data_error_result(message="metadata_condition must be an object.")
+    metadata = req.get("metadata", {}) or {}
+    if isinstance(metadata, dict) and metadata.get("empty_metadata"):
+        return_empty_metadata = True
+        metadata = {k: v for k, v in metadata.items() if k != "empty_metadata"}
+    if return_empty_metadata:
+        metadata_condition = {}
+        metadata = {}
+    else:
+        if metadata_condition and not isinstance(metadata_condition, dict):
+            return get_data_error_result(message="metadata_condition must be an object.")
+        if metadata and not isinstance(metadata, dict):
+            return get_data_error_result(message="metadata must be an object.")
 
     doc_ids_filter = None
-    if metadata_condition:
+    metas = None
+    if metadata_condition or metadata:
         metas = DocumentService.get_flatted_meta_by_kbs([kb_id])
-        doc_ids_filter = meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and"))
+
+    if metadata_condition:
+        doc_ids_filter = set(meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and")))
         if metadata_condition.get("conditions") and not doc_ids_filter:
             return get_json_result(data={"total": 0, "docs": []})
 
+    if metadata:
+        metadata_doc_ids = None
+        for key, values in metadata.items():
+            if not values:
+                continue
+            if not isinstance(values, list):
+                values = [values]
+            values = [str(v) for v in values if v is not None and str(v).strip()]
+            if not values:
+                continue
+            key_doc_ids = set()
+            for value in values:
+                key_doc_ids.update(metas.get(key, {}).get(value, []))
+            if metadata_doc_ids is None:
+                metadata_doc_ids = key_doc_ids
+            else:
+                metadata_doc_ids &= key_doc_ids
+            if not metadata_doc_ids:
+                return get_json_result(data={"total": 0, "docs": []})
+        if metadata_doc_ids is not None:
+            if doc_ids_filter is None:
+                doc_ids_filter = metadata_doc_ids
+            else:
+                doc_ids_filter &= metadata_doc_ids
+            if not doc_ids_filter:
+                return get_json_result(data={"total": 0, "docs": []})
+
+    if doc_ids_filter is not None:
+        doc_ids_filter = list(doc_ids_filter)
+
     try:
-        docs, tol = DocumentService.get_by_kb_id(kb_id, page_number, items_per_page, orderby, desc, keywords, run_status, types, suffix, doc_ids_filter)
+        docs, tol = DocumentService.get_by_kb_id(
+            kb_id,
+            page_number,
+            items_per_page,
+            orderby,
+            desc,
+            keywords,
+            run_status,
+            types,
+            suffix,
+            doc_ids_filter,
+            return_empty_metadata=return_empty_metadata,
+        )
 
         if create_time_from or create_time_to:
             filtered_docs = []
@@ -411,6 +470,26 @@ async def metadata_update():
     return get_json_result(data={"updated": updated, "matched_docs": len(target_doc_ids)})
 
 
+@manager.route("/update_metadata_setting", methods=["POST"])  # noqa: F821
+@login_required
+@validate_request("doc_id", "metadata")
+async def update_metadata_setting():
+    req = await get_request_json()
+    if not DocumentService.accessible(req["doc_id"], current_user.id):
+        return get_json_result(data=False, message="No authorization.", code=RetCode.AUTHENTICATION_ERROR)
+
+    e, doc = DocumentService.get_by_id(req["doc_id"])
+    if not e:
+        return get_data_error_result(message="Document not found!")
+
+    DocumentService.update_parser_config(doc.id, {"metadata": req["metadata"]})
+    e, doc = DocumentService.get_by_id(doc.id)
+    if not e:
+        return get_data_error_result(message="Document not found!")
+
+    return get_json_result(data=doc.to_dict())
+
+
 @manager.route("/thumbnails", methods=["GET"])  # noqa: F821
 # @login_required
 def thumbnails():
@@ -528,10 +607,17 @@ async def run():
                 DocumentService.update_by_id(id, info)
                 if req.get("delete", False):
                     TaskService.filter_delete([Task.doc_id == id])
-                    if settings.docStoreConn.indexExist(search.index_name(tenant_id), doc.kb_id):
+                    if settings.docStoreConn.index_exist(search.index_name(tenant_id), doc.kb_id):
                         settings.docStoreConn.delete({"doc_id": id}, search.index_name(tenant_id), doc.kb_id)
 
                 if str(req["run"]) == TaskStatus.RUNNING.value:
+                    if req.get("apply_kb"):
+                        e, kb = KnowledgebaseService.get_by_id(doc.kb_id)
+                        if not e:
+                            raise LookupError("Can't find this dataset!")
+                        doc.parser_config["enable_metadata"] = kb.parser_config.get("enable_metadata", False)
+                        doc.parser_config["metadata"] = kb.parser_config.get("metadata", {})
+                        DocumentService.update_parser_config(doc.id, doc.parser_config)
                     doc_dict = doc.to_dict()
                     DocumentService.run(tenant_id, doc_dict, kb_table_num_map)
 
@@ -579,7 +665,7 @@ async def rename():
                 "title_tks": title_tks,
                 "title_sm_tks": rag_tokenizer.fine_grained_tokenize(title_tks),
             }
-            if settings.docStoreConn.indexExist(search.index_name(tenant_id), doc.kb_id):
+            if settings.docStoreConn.index_exist(search.index_name(tenant_id), doc.kb_id):
                 settings.docStoreConn.update(
                     {"doc_id": req["doc_id"]},
                     es_body,
@@ -660,7 +746,8 @@ async def change_parser():
             tenant_id = DocumentService.get_tenant_id(req["doc_id"])
             if not tenant_id:
                 return get_data_error_result(message="Tenant not found!")
-            if settings.docStoreConn.indexExist(search.index_name(tenant_id), doc.kb_id):
+            DocumentService.delete_chunk_images(doc, tenant_id)
+            if settings.docStoreConn.index_exist(search.index_name(tenant_id), doc.kb_id):
                 settings.docStoreConn.delete({"doc_id": doc.id}, search.index_name(tenant_id), doc.kb_id)
         return None
 

api/apps/kb_app.py

@@ -39,9 +39,9 @@ from api.utils.api_utils import get_json_result
 from rag.nlp import search
 from api.constants import DATASET_NAME_LIMIT
 from rag.utils.redis_conn import REDIS_CONN
-from rag.utils.doc_store_conn import OrderByExpr
 from common.constants import RetCode, PipelineTaskType, StatusEnum, VALID_TASK_STATUS, FileSource, LLMType, PAGERANK_FLD
 from common import settings
+from common.doc_store.doc_store_base import OrderByExpr
 from api.apps import login_required, current_user
 
 
@@ -97,6 +97,19 @@ async def update():
                 code=RetCode.OPERATING_ERROR)
 
         e, kb = KnowledgebaseService.get_by_id(req["kb_id"])
+
+        # Rename folder in FileService
+        if e and req["name"].lower() != kb.name.lower():
+            FileService.filter_update(
+                [
+                    File.tenant_id == kb.tenant_id,
+                    File.source_type == FileSource.KNOWLEDGEBASE,
+                    File.type == "folder",
+                    File.name == kb.name,
+                ],
+                {"name": req["name"]},
+            )
+
         if not e:
             return get_data_error_result(
                 message="Can't find this dataset!")
@@ -150,6 +163,21 @@ async def update():
         return server_error_response(e)
 
 
+@manager.route('/update_metadata_setting', methods=['post'])  # noqa: F821
+@login_required
+@validate_request("kb_id", "metadata")
+async def update_metadata_setting():
+    req = await get_request_json()
+    e, kb = KnowledgebaseService.get_by_id(req["kb_id"])
+    if not e:
+        return get_data_error_result(
+            message="Database error (Knowledgebase rename)!")
+    kb = kb.to_dict()
+    kb["parser_config"]["metadata"] = req["metadata"]
+    KnowledgebaseService.update_by_id(kb["id"], kb)
+    return get_json_result(data=kb)
+
+
 @manager.route('/detail', methods=['GET'])  # noqa: F821
 @login_required
 def detail():
@@ -245,13 +273,19 @@ async def rm():
                     FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == f2d[0].file_id])
                 File2DocumentService.delete_by_document_id(doc.id)
             FileService.filter_delete(
-                [File.source_type == FileSource.KNOWLEDGEBASE, File.type == "folder", File.name == kbs[0].name])
+                [
+                    File.tenant_id == kbs[0].tenant_id,
+                    File.source_type == FileSource.KNOWLEDGEBASE,
+                    File.type == "folder",
+                    File.name == kbs[0].name,
+                ]
+            )
             if not KnowledgebaseService.delete_by_id(req["kb_id"]):
                 return get_data_error_result(
                     message="Database error (Knowledgebase removal)!")
             for kb in kbs:
                 settings.docStoreConn.delete({"kb_id": kb.id}, search.index_name(kb.tenant_id), kb.id)
-                settings.docStoreConn.deleteIdx(search.index_name(kb.tenant_id), kb.id)
+                settings.docStoreConn.delete_idx(search.index_name(kb.tenant_id), kb.id)
                 if hasattr(settings.STORAGE_IMPL, 'remove_bucket'):
                     settings.STORAGE_IMPL.remove_bucket(kb.id)
             return get_json_result(data=True)
@@ -352,7 +386,7 @@ def knowledge_graph(kb_id):
     }
 
     obj = {"graph": {}, "mind_map": {}}
-    if not settings.docStoreConn.indexExist(search.index_name(kb.tenant_id), kb_id):
+    if not settings.docStoreConn.index_exist(search.index_name(kb.tenant_id), kb_id):
         return get_json_result(data=obj)
     sres = settings.retriever.search(req, search.index_name(kb.tenant_id), [kb_id])
     if not len(sres.ids):
@@ -824,11 +858,11 @@ async def check_embedding():
         index_nm = search.index_name(tenant_id)
 
         res0 = docStoreConn.search(
-            selectFields=[], highlightFields=[],
+            select_fields=[], highlight_fields=[],
             condition={"kb_id": kb_id, "available_int": 1},
-            matchExprs=[], orderBy=OrderByExpr(),
+            match_expressions=[], order_by=OrderByExpr(),
             offset=0, limit=1,
-            indexNames=index_nm, knowledgebaseIds=[kb_id]
+            index_names=index_nm, knowledgebase_ids=[kb_id]
         )
         total = docStoreConn.get_total(res0)
         if total <= 0:
@@ -840,14 +874,14 @@ async def check_embedding():
 
         for off in offsets:
             res1 = docStoreConn.search(
-                selectFields=list(base_fields),
-                highlightFields=[],
+                select_fields=list(base_fields),
+                highlight_fields=[],
                 condition={"kb_id": kb_id, "available_int": 1},
-                matchExprs=[], orderBy=OrderByExpr(),
+                match_expressions=[], order_by=OrderByExpr(),
                 offset=off, limit=1,
-                indexNames=index_nm, knowledgebaseIds=[kb_id]
+                index_names=index_nm, knowledgebase_ids=[kb_id]
             )
-            ids = docStoreConn.get_chunk_ids(res1)
+            ids = docStoreConn.get_doc_ids(res1)
             if not ids:
                 continue
 

api/apps/llm_app.py

@@ -25,7 +25,7 @@ from api.utils.api_utils import get_allowed_llm_factories, get_data_error_result
 from common.constants import StatusEnum, LLMType
 from api.db.db_models import TenantLLM
 from rag.utils.base64_image import test_image
-from rag.llm import EmbeddingModel, ChatModel, RerankModel, CvModel, TTSModel, OcrModel
+from rag.llm import EmbeddingModel, ChatModel, RerankModel, CvModel, TTSModel, OcrModel, Seq2txtModel
 
 
 @manager.route("/factories", methods=["GET"])  # noqa: F821
@@ -157,7 +157,7 @@ async def add_llm():
     elif factory == "Bedrock":
         # For Bedrock, due to its special authentication method
         # Assemble bedrock_ak, bedrock_sk, bedrock_region
-        api_key = apikey_json(["bedrock_ak", "bedrock_sk", "bedrock_region"])
+        api_key = apikey_json(["auth_mode", "bedrock_ak", "bedrock_sk", "bedrock_region", "aws_role_arn"])
 
     elif factory == "LocalAI":
         llm_name += "___LocalAI"
@@ -208,70 +208,83 @@ async def add_llm():
     msg = ""
     mdl_nm = llm["llm_name"].split("___")[0]
     extra = {"provider": factory}
-    if llm["model_type"] == LLMType.EMBEDDING.value:
-        assert factory in EmbeddingModel, f"Embedding model from {factory} is not supported yet."
-        mdl = EmbeddingModel[factory](key=llm["api_key"], model_name=mdl_nm, base_url=llm["api_base"])
-        try:
-            arr, tc = mdl.encode(["Test if the api key is available"])
-            if len(arr[0]) == 0:
-                raise Exception("Fail")
-        except Exception as e:
-            msg += f"\nFail to access embedding model({mdl_nm})." + str(e)
-    elif llm["model_type"] == LLMType.CHAT.value:
-        assert factory in ChatModel, f"Chat model from {factory} is not supported yet."
-        mdl = ChatModel[factory](
-            key=llm["api_key"],
-            model_name=mdl_nm,
-            base_url=llm["api_base"],
-            **extra,
-        )
-        try:
-            m, tc = await mdl.async_chat(None, [{"role": "user", "content": "Hello! How are you doing!"}], {"temperature": 0.9})
-            if not tc and m.find("**ERROR**:") >= 0:
-                raise Exception(m)
-        except Exception as e:
-            msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
-    elif llm["model_type"] == LLMType.RERANK:
-        assert factory in RerankModel, f"RE-rank model from {factory} is not supported yet."
-        try:
-            mdl = RerankModel[factory](key=llm["api_key"], model_name=mdl_nm, base_url=llm["api_base"])
-            arr, tc = mdl.similarity("Hello~ RAGFlower!", ["Hi, there!", "Ohh, my friend!"])
-            if len(arr) == 0:
-                raise Exception("Not known.")
-        except KeyError:
-            msg += f"{factory} dose not support this model({factory}/{mdl_nm})"
-        except Exception as e:
-            msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
-    elif llm["model_type"] == LLMType.IMAGE2TEXT.value:
-        assert factory in CvModel, f"Image to text model from {factory} is not supported yet."
-        mdl = CvModel[factory](key=llm["api_key"], model_name=mdl_nm, base_url=llm["api_base"])
-        try:
-            image_data = test_image
-            m, tc = mdl.describe(image_data)
-            if not tc and m.find("**ERROR**:") >= 0:
-                raise Exception(m)
-        except Exception as e:
-            msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
-    elif llm["model_type"] == LLMType.TTS:
-        assert factory in TTSModel, f"TTS model from {factory} is not supported yet."
-        mdl = TTSModel[factory](key=llm["api_key"], model_name=mdl_nm, base_url=llm["api_base"])
-        try:
-            for resp in mdl.tts("Hello~ RAGFlower!"):
-                pass
-        except RuntimeError as e:
-            msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
-    elif llm["model_type"] == LLMType.OCR.value:
-        assert factory in OcrModel, f"OCR model from {factory} is not supported yet."
-        try:
-            mdl = OcrModel[factory](key=llm["api_key"], model_name=mdl_nm, base_url=llm.get("api_base", ""))
-            ok, reason = mdl.check_available()
-            if not ok:
-                raise RuntimeError(reason or "Model not available")
-        except Exception as e:
-            msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
-    else:
-        # TODO: check other type of models
-        pass
+    model_type = llm["model_type"]
+    model_api_key = llm["api_key"]
+    model_base_url = llm.get("api_base", "")
+    match model_type:
+        case LLMType.EMBEDDING.value:
+            assert factory in EmbeddingModel, f"Embedding model from {factory} is not supported yet."
+            mdl = EmbeddingModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+            try:
+                arr, tc = mdl.encode(["Test if the api key is available"])
+                if len(arr[0]) == 0:
+                    raise Exception("Fail")
+            except Exception as e:
+                msg += f"\nFail to access embedding model({mdl_nm})." + str(e)
+        case LLMType.CHAT.value:
+            assert factory in ChatModel, f"Chat model from {factory} is not supported yet."
+            mdl = ChatModel[factory](
+                key=model_api_key,
+                model_name=mdl_nm,
+                base_url=model_base_url,
+                **extra,
+            )
+            try:
+                m, tc = await mdl.async_chat(None, [{"role": "user", "content": "Hello! How are you doing!"}],
+                                             {"temperature": 0.9})
+                if not tc and m.find("**ERROR**:") >= 0:
+                    raise Exception(m)
+            except Exception as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+
+        case LLMType.RERANK.value:
+            assert factory in RerankModel, f"RE-rank model from {factory} is not supported yet."
+            try:
+                mdl = RerankModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+                arr, tc = mdl.similarity("Hello~ RAGFlower!", ["Hi, there!", "Ohh, my friend!"])
+                if len(arr) == 0:
+                    raise Exception("Not known.")
+            except KeyError:
+                msg += f"{factory} dose not support this model({factory}/{mdl_nm})"
+            except Exception as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+
+        case LLMType.IMAGE2TEXT.value:
+            assert factory in CvModel, f"Image to text model from {factory} is not supported yet."
+            mdl = CvModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+            try:
+                image_data = test_image
+                m, tc = mdl.describe(image_data)
+                if not tc and m.find("**ERROR**:") >= 0:
+                    raise Exception(m)
+            except Exception as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+        case LLMType.TTS.value:
+            assert factory in TTSModel, f"TTS model from {factory} is not supported yet."
+            mdl = TTSModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+            try:
+                for resp in mdl.tts("Hello~ RAGFlower!"):
+                    pass
+            except RuntimeError as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+        case LLMType.OCR.value:
+            assert factory in OcrModel, f"OCR model from {factory} is not supported yet."
+            try:
+                mdl = OcrModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+                ok, reason = mdl.check_available()
+                if not ok:
+                    raise RuntimeError(reason or "Model not available")
+            except Exception as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+        case LLMType.SPEECH2TEXT:
+            assert factory in Seq2txtModel, f"Speech model from {factory} is not supported yet."
+            try:
+                mdl = Seq2txtModel[factory](key=model_api_key, model_name=mdl_nm, base_url=model_base_url)
+                # TODO: check the availability
+            except Exception as e:
+                msg += f"\nFail to access model({factory}/{mdl_nm})." + str(e)
+        case _:
+            raise RuntimeError(f"Unknown model type: {model_type}")
 
     if msg:
         return get_data_error_result(message=msg)

api/apps/mcp_server_app.py

@@ -13,6 +13,8 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import asyncio
+
 from quart import Response, request
 from api.apps import current_user, login_required
 
@@ -106,7 +108,7 @@ async def create() -> Response:
             return get_data_error_result(message="Tenant not found.")
 
         mcp_server = MCPServer(id=server_name, name=server_name, url=url, server_type=server_type, variables=variables, headers=headers)
-        server_tools, err_message = get_mcp_tools([mcp_server], timeout)
+        server_tools, err_message = await asyncio.to_thread(get_mcp_tools, [mcp_server], timeout)
         if err_message:
             return get_data_error_result(err_message)
 
@@ -158,7 +160,7 @@ async def update() -> Response:
         req["id"] = mcp_id
 
         mcp_server = MCPServer(id=server_name, name=server_name, url=url, server_type=server_type, variables=variables, headers=headers)
-        server_tools, err_message = get_mcp_tools([mcp_server], timeout)
+        server_tools, err_message = await asyncio.to_thread(get_mcp_tools, [mcp_server], timeout)
         if err_message:
             return get_data_error_result(err_message)
 
@@ -242,7 +244,7 @@ async def import_multiple() -> Response:
             headers = {"authorization_token": config["authorization_token"]} if "authorization_token" in config else {}
             variables = {k: v for k, v in config.items() if k not in {"type", "url", "headers"}}
             mcp_server = MCPServer(id=new_name, name=new_name, url=config["url"], server_type=config["type"], variables=variables, headers=headers)
-            server_tools, err_message = get_mcp_tools([mcp_server], timeout)
+            server_tools, err_message = await asyncio.to_thread(get_mcp_tools, [mcp_server], timeout)
             if err_message:
                 results.append({"server": base_name, "success": False, "message": err_message})
                 continue
@@ -322,9 +324,8 @@ async def list_tools() -> Response:
                 tool_call_sessions.append(tool_call_session)
 
                 try:
-                    tools = tool_call_session.get_tools(timeout)
+                    tools = await asyncio.to_thread(tool_call_session.get_tools, timeout)
                 except Exception as e:
-                    tools = []
                     return get_data_error_result(message=f"MCP list tools error: {e}")
 
                 results[server_key] = []
@@ -340,7 +341,7 @@ async def list_tools() -> Response:
         return server_error_response(e)
     finally:
         # PERF: blocking call to close sessions — consider moving to background thread or task queue
-        close_multiple_mcp_toolcall_sessions(tool_call_sessions)
+        await asyncio.to_thread(close_multiple_mcp_toolcall_sessions, tool_call_sessions)
 
 
 @manager.route("/test_tool", methods=["POST"])  # noqa: F821
@@ -367,10 +368,10 @@ async def test_tool() -> Response:
 
         tool_call_session = MCPToolCallSession(mcp_server, mcp_server.variables)
         tool_call_sessions.append(tool_call_session)
-        result = tool_call_session.tool_call(tool_name, arguments, timeout)
+        result = await asyncio.to_thread(tool_call_session.tool_call, tool_name, arguments, timeout)
 
         # PERF: blocking call to close sessions — consider moving to background thread or task queue
-        close_multiple_mcp_toolcall_sessions(tool_call_sessions)
+        await asyncio.to_thread(close_multiple_mcp_toolcall_sessions, tool_call_sessions)
         return get_json_result(data=result)
     except Exception as e:
         return server_error_response(e)
@@ -424,13 +425,12 @@ async def test_mcp() -> Response:
         tool_call_session = MCPToolCallSession(mcp_server, mcp_server.variables)
 
         try:
-            tools = tool_call_session.get_tools(timeout)
+            tools = await asyncio.to_thread(tool_call_session.get_tools, timeout)
         except Exception as e:
-            tools = []
             return get_data_error_result(message=f"Test MCP error: {e}")
         finally:
             # PERF: blocking call to close sessions — consider moving to background thread or task queue
-            close_multiple_mcp_toolcall_sessions([tool_call_session])
+            await asyncio.to_thread(close_multiple_mcp_toolcall_sessions, [tool_call_session])
 
         for tool in tools:
             tool_dict = tool.model_dump()

api/apps/memories_app.py

@@ -20,10 +20,13 @@ from api.apps import login_required, current_user
 from api.db import TenantPermission
 from api.db.services.memory_service import MemoryService
 from api.db.services.user_service import UserTenantService
-from api.utils.api_utils import validate_request, get_request_json, get_error_argument_result, get_json_result, \
-    not_allowed_parameters
+from api.db.services.canvas_service import UserCanvasService
+from api.db.joint_services.memory_message_service import get_memory_size_cache, judge_system_prompt_is_default
+from api.utils.api_utils import validate_request, get_request_json, get_error_argument_result, get_json_result
 from api.utils.memory_utils import format_ret_data_from_memory, get_memory_type_human
 from api.constants import MEMORY_NAME_LIMIT, MEMORY_SIZE_LIMIT
+from memory.services.messages import MessageService
+from memory.utils.prompt_util import PromptAssembler
 from common.constants import MemoryType, RetCode, ForgettingPolicy
 
 
@@ -57,7 +60,6 @@ async def create_memory():
 
         if res:
             return get_json_result(message=True, data=format_ret_data_from_memory(memory))
-
         else:
             return get_json_result(message=memory, code=RetCode.SERVER_ERROR)
 
@@ -67,7 +69,6 @@ async def create_memory():
 
 @manager.route("/<memory_id>", methods=["PUT"])  # noqa: F821
 @login_required
-@not_allowed_parameters("id", "tenant_id", "memory_type", "storage_type", "embd_id")
 async def update_memory(memory_id):
     req = await get_request_json()
     update_dict = {}
@@ -87,6 +88,14 @@ async def update_memory(memory_id):
         update_dict["permissions"] = req["permissions"]
     if req.get("llm_id"):
         update_dict["llm_id"] = req["llm_id"]
+    if req.get("embd_id"):
+        update_dict["embd_id"] = req["embd_id"]
+    if req.get("memory_type"):
+        memory_type = set(req["memory_type"])
+        invalid_type = memory_type - {e.name.lower() for e in MemoryType}
+        if invalid_type:
+            return get_error_argument_result(f"Memory type '{invalid_type}' is not supported.")
+        update_dict["memory_type"] = list(memory_type)
     # check memory_size valid
     if req.get("memory_size"):
         if not 0 < int(req["memory_size"]) <= MEMORY_SIZE_LIMIT:
@@ -122,9 +131,18 @@ async def update_memory(memory_id):
 
     if not to_update:
         return get_json_result(message=True, data=memory_dict)
+    # check memory empty when update embd_id, memory_type
+    memory_size = get_memory_size_cache(memory_id, current_memory.tenant_id)
+    not_allowed_update = [f for f in ["embd_id", "memory_type"] if f in to_update and memory_size > 0]
+    if not_allowed_update:
+        return get_error_argument_result(f"Can't update {not_allowed_update} when memory isn't empty.")
+    if "memory_type" in to_update:
+        if "system_prompt" not in to_update and judge_system_prompt_is_default(current_memory.system_prompt, current_memory.memory_type):
+            # update old default prompt, assemble a new one
+            to_update["system_prompt"] = PromptAssembler.assemble_system_prompt({"memory_type": to_update["memory_type"]})
 
     try:
-        MemoryService.update_memory(memory_id, to_update)
+        MemoryService.update_memory(current_memory.tenant_id, memory_id, to_update)
         updated_memory = MemoryService.get_by_memory_id(memory_id)
         return get_json_result(message=True, data=format_ret_data_from_memory(updated_memory))
 
@@ -133,7 +151,7 @@ async def update_memory(memory_id):
         return get_json_result(message=str(e), code=RetCode.SERVER_ERROR)
 
 
-@manager.route("/<memory_id>", methods=["DELETE"]) # noqa: F821
+@manager.route("/<memory_id>", methods=["DELETE"])  # noqa: F821
 @login_required
 async def delete_memory(memory_id):
     memory = MemoryService.get_by_memory_id(memory_id)
@@ -141,13 +159,15 @@ async def delete_memory(memory_id):
         return get_json_result(message=True, code=RetCode.NOT_FOUND)
     try:
         MemoryService.delete_memory(memory_id)
+        if MessageService.has_index(memory.tenant_id, memory_id):
+            MessageService.delete_message({"memory_id": memory_id}, memory.tenant_id, memory_id)
         return get_json_result(message=True)
     except Exception as e:
         logging.error(e)
         return get_json_result(message=str(e), code=RetCode.SERVER_ERROR)
 
 
-@manager.route("", methods=["GET"]) # noqa: F821
+@manager.route("", methods=["GET"])  # noqa: F821
 @login_required
 async def list_memory():
     args = request.args
@@ -183,3 +203,26 @@ async def get_memory_config(memory_id):
     if not memory:
         return get_json_result(code=RetCode.NOT_FOUND, message=f"Memory '{memory_id}' not found.")
     return get_json_result(message=True, data=format_ret_data_from_memory(memory))
+
+
+@manager.route("/<memory_id>", methods=["GET"])  # noqa: F821
+@login_required
+async def get_memory_detail(memory_id):
+    args = request.args
+    agent_ids = args.getlist("agent_id")
+    keywords = args.get("keywords", "")
+    keywords = keywords.strip()
+    page = int(args.get("page", 1))
+    page_size = int(args.get("page_size", 50))
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return get_json_result(code=RetCode.NOT_FOUND, message=f"Memory '{memory_id}' not found.")
+    messages = MessageService.list_message(
+        memory.tenant_id, memory_id, agent_ids, keywords, page, page_size)
+    agent_name_mapping = {}
+    if messages["message_list"]:
+        agent_list = UserCanvasService.get_basic_info_by_canvas_ids([message["agent_id"] for message in messages["message_list"]])
+        agent_name_mapping = {agent["id"]: agent["title"] for agent in agent_list}
+    for message in messages["message_list"]:
+        message["agent_name"] = agent_name_mapping.get(message["agent_id"], "Unknown")
+    return get_json_result(data={"messages": messages, "storage_type": memory.storage_type}, message=True)

api/apps/messages_app.py

@@ -0,0 +1,168 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+from quart import request
+from api.apps import login_required
+from api.db.services.memory_service import MemoryService
+from common.time_utils import current_timestamp, timestamp_to_date
+
+from memory.services.messages import MessageService
+from api.db.joint_services import memory_message_service
+from api.utils.api_utils import validate_request, get_request_json, get_error_argument_result, get_json_result
+from common.constants import RetCode
+
+
+@manager.route("", methods=["POST"]) # noqa: F821
+@login_required
+@validate_request("memory_id", "agent_id", "session_id", "user_input", "agent_response")
+async def add_message():
+
+    req = await get_request_json()
+    memory_ids = req["memory_id"]
+    agent_id = req["agent_id"]
+    session_id = req["session_id"]
+    user_id = req["user_id"] if req.get("user_id") else ""
+    user_input = req["user_input"]
+    agent_response = req["agent_response"]
+
+    res = []
+    for memory_id in memory_ids:
+        success, msg = await memory_message_service.save_to_memory(
+            memory_id,
+            {
+                "user_id": user_id,
+                "agent_id": agent_id,
+                "session_id": session_id,
+                "user_input": user_input,
+                "agent_response": agent_response
+            }
+        )
+        res.append({
+            "memory_id": memory_id,
+            "success": success,
+            "message": msg
+        })
+
+    if all([r["success"] for r in res]):
+        return get_json_result(message="Successfully added to memories.")
+
+    return get_json_result(code=RetCode.SERVER_ERROR, message="Some messages failed to add.", data=res)
+
+
+@manager.route("/<memory_id>:<message_id>", methods=["DELETE"]) # noqa: F821
+@login_required
+async def forget_message(memory_id: str, message_id: int):
+
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return get_json_result(code=RetCode.NOT_FOUND, message=f"Memory '{memory_id}' not found.")
+
+    forget_time = timestamp_to_date(current_timestamp())
+    update_succeed = MessageService.update_message(
+        {"memory_id": memory_id, "message_id": int(message_id)},
+        {"forget_at": forget_time},
+        memory.tenant_id, memory_id)
+    if update_succeed:
+        return get_json_result(message=update_succeed)
+    else:
+        return get_json_result(code=RetCode.SERVER_ERROR, message=f"Failed to forget message '{message_id}' in memory '{memory_id}'.")
+
+
+@manager.route("/<memory_id>:<message_id>", methods=["PUT"]) # noqa: F821
+@login_required
+@validate_request("status")
+async def update_message(memory_id: str, message_id: int):
+    req = await get_request_json()
+    status = req["status"]
+    if not isinstance(status, bool):
+        return get_error_argument_result("Status must be a boolean.")
+
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return get_json_result(code=RetCode.NOT_FOUND, message=f"Memory '{memory_id}' not found.")
+
+    update_succeed = MessageService.update_message({"memory_id": memory_id, "message_id": int(message_id)}, {"status": status}, memory.tenant_id, memory_id)
+    if update_succeed:
+        return get_json_result(message=update_succeed)
+    else:
+        return get_json_result(code=RetCode.SERVER_ERROR, message=f"Failed to set status for message '{message_id}' in memory '{memory_id}'.")
+
+
+@manager.route("/search", methods=["GET"]) # noqa: F821
+@login_required
+async def search_message():
+    args = request.args
+    print(args, flush=True)
+    empty_fields = [f for f in ["memory_id", "query"] if not args.get(f)]
+    if empty_fields:
+        return get_error_argument_result(f"{', '.join(empty_fields)} can't be empty.")
+
+    memory_ids = args.getlist("memory_id")
+    query = args.get("query")
+    similarity_threshold = float(args.get("similarity_threshold", 0.2))
+    keywords_similarity_weight = float(args.get("keywords_similarity_weight", 0.7))
+    top_n = int(args.get("top_n", 5))
+    agent_id = args.get("agent_id", "")
+    session_id = args.get("session_id", "")
+
+    filter_dict = {
+        "memory_id": memory_ids,
+        "agent_id": agent_id,
+        "session_id": session_id
+    }
+    params = {
+        "query": query,
+        "similarity_threshold": similarity_threshold,
+        "keywords_similarity_weight": keywords_similarity_weight,
+        "top_n": top_n
+    }
+    res = memory_message_service.query_message(filter_dict, params)
+    return get_json_result(message=True, data=res)
+
+
+@manager.route("", methods=["GET"]) # noqa: F821
+@login_required
+async def get_messages():
+    args = request.args
+    memory_ids = args.getlist("memory_id")
+    agent_id = args.get("agent_id", "")
+    session_id = args.get("session_id", "")
+    limit = int(args.get("limit", 10))
+    if not memory_ids:
+        return get_error_argument_result("memory_ids is required.")
+    memory_list = MemoryService.get_by_ids(memory_ids)
+    uids = [memory.tenant_id for memory in memory_list]
+    res = MessageService.get_recent_messages(
+        uids,
+        memory_ids,
+        agent_id,
+        session_id,
+        limit
+    )
+    return get_json_result(message=True, data=res)
+
+
+@manager.route("/<memory_id>:<message_id>/content", methods=["GET"]) # noqa: F821
+@login_required
+async def get_message_content(memory_id:str, message_id: int):
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return get_json_result(code=RetCode.NOT_FOUND, message=f"Memory '{memory_id}' not found.")
+
+    res = MessageService.get_by_message_id(memory_id, message_id, memory.tenant_id)
+    if res:
+        return get_json_result(message=True, data=res)
+    else:
+        return get_json_result(code=RetCode.NOT_FOUND, message=f"Message '{message_id}' in memory '{memory_id}' not found.")

api/apps/sdk/agents.py

@@ -14,20 +14,29 @@
 #  limitations under the License.
 #
 
+import asyncio
+import base64
+import hashlib
+import hmac
+import ipaddress
 import json
 import logging
 import time
 from typing import Any, cast
 
+import jwt
+
 from agent.canvas import Canvas
 from api.db import CanvasCategory
 from api.db.services.canvas_service import UserCanvasService
+from api.db.services.file_service import FileService
 from api.db.services.user_canvas_version import UserCanvasVersionService
 from common.constants import RetCode
 from common.misc_utils import get_uuid
 from api.utils.api_utils import get_data_error_result, get_error_data_result, get_json_result, get_request_json, token_required
 from api.utils.api_utils import get_result
 from quart import request, Response
+from rag.utils.redis_conn import REDIS_CONN
 
 
 @manager.route('/agents', methods=['GET'])  # noqa: F821
@@ -132,48 +141,785 @@ def delete_agent(tenant_id: str, agent_id: str):
     UserCanvasService.delete_by_id(agent_id)
     return get_json_result(data=True)
 
+@manager.route("/webhook/<agent_id>", methods=["POST", "GET", "PUT", "PATCH", "DELETE", "HEAD"])  # noqa: F821
+@manager.route("/webhook_test/<agent_id>",methods=["POST", "GET", "PUT", "PATCH", "DELETE", "HEAD"],)  # noqa: F821
+async def webhook(agent_id: str):
+    is_test = request.path.startswith("/api/v1/webhook_test")
+    start_ts = time.time()
 
-@manager.route('/webhook/<agent_id>', methods=['POST'])  # noqa: F821
-@token_required
-async def webhook(tenant_id: str, agent_id: str):
-    req = await get_request_json()
-    if not UserCanvasService.accessible(req["id"], tenant_id):
-        return get_json_result(
-            data=False, message='Only owner of canvas authorized for this operation.',
-            code=RetCode.OPERATING_ERROR)
-
-    e, cvs = UserCanvasService.get_by_id(req["id"])
-    if not e:
-        return get_data_error_result(message="canvas not found.")
-
-    if not isinstance(cvs.dsl, str):
-        cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
+    # 1. Fetch canvas by agent_id
+    exists, cvs = UserCanvasService.get_by_id(agent_id)
+    if not exists:
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message="Canvas not found."),RetCode.BAD_REQUEST
 
+    # 2. Check canvas category
     if cvs.canvas_category == CanvasCategory.DataFlow:
-        return get_data_error_result(message="Dataflow can not be triggered by webhook.")
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message="Dataflow can not be triggered by webhook."),RetCode.BAD_REQUEST
+
+    # 3. Load DSL from canvas
+    dsl = getattr(cvs, "dsl", None)
+    if not isinstance(dsl, dict):
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message="Invalid DSL format."),RetCode.BAD_REQUEST
+
+    # 4. Check webhook configuration in DSL
+    components = dsl.get("components", {})
+    for k, _ in components.items():
+        cpn_obj = components[k]["obj"]
+        if cpn_obj["component_name"].lower() == "begin" and cpn_obj["params"]["mode"] == "Webhook":
+            webhook_cfg = cpn_obj["params"]
+
+    if not webhook_cfg:
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message="Webhook not configured for this agent."),RetCode.BAD_REQUEST
+
+    # 5. Validate request method against webhook_cfg.methods
+    allowed_methods = webhook_cfg.get("methods", [])
+    request_method = request.method.upper()
+    if allowed_methods and request_method not in allowed_methods:
+        return get_data_error_result(
+            code=RetCode.BAD_REQUEST,message=f"HTTP method '{request_method}' not allowed for this webhook."
+        ),RetCode.BAD_REQUEST
+
+    # 6. Validate webhook security
+    async def validate_webhook_security(security_cfg: dict):
+        """Validate webhook security rules based on security configuration."""
+
+        if not security_cfg:
+            return  # No security config → allowed by default
+
+        # 1. Validate max body size
+        await _validate_max_body_size(security_cfg)
+
+        # 2. Validate IP whitelist
+        _validate_ip_whitelist(security_cfg)
+
+        # # 3. Validate rate limiting
+        _validate_rate_limit(security_cfg)
+
+        # 4. Validate authentication
+        auth_type = security_cfg.get("auth_type", "none")
+
+        if auth_type == "none":
+            return
+
+        if auth_type == "token":
+            _validate_token_auth(security_cfg)
+
+        elif auth_type == "basic":
+            _validate_basic_auth(security_cfg)
+
+        elif auth_type == "jwt":
+            _validate_jwt_auth(security_cfg)
+
+        else:
+            raise Exception(f"Unsupported auth_type: {auth_type}")
+
+    async def _validate_max_body_size(security_cfg):
+        """Check request size does not exceed max_body_size."""
+        max_size = security_cfg.get("max_body_size")
+        if not max_size:
+            return
+
+        # Convert "10MB" → bytes
+        units = {"kb": 1024, "mb": 1024**2}
+        size_str = max_size.lower()
+
+        for suffix, factor in units.items():
+            if size_str.endswith(suffix):
+                limit = int(size_str.replace(suffix, "")) * factor
+                break
+        else:
+            raise Exception("Invalid max_body_size format")
+        MAX_LIMIT = 10 * 1024 * 1024  # 10MB
+        if limit > MAX_LIMIT:
+            raise Exception("max_body_size exceeds maximum allowed size (10MB)")
+
+        content_length = request.content_length or 0
+        if content_length > limit:
+            raise Exception(f"Request body too large: {content_length} > {limit}")
+
+    def _validate_ip_whitelist(security_cfg):
+        """Allow only IPs listed in ip_whitelist."""
+        whitelist = security_cfg.get("ip_whitelist", [])
+        if not whitelist:
+            return
+
+        client_ip = request.remote_addr
+
+
+        for rule in whitelist:
+            if "/" in rule:
+                # CIDR notation
+                if ipaddress.ip_address(client_ip) in ipaddress.ip_network(rule, strict=False):
+                    return
+            else:
+                # Single IP
+                if client_ip == rule:
+                    return
+
+        raise Exception(f"IP {client_ip} is not allowed by whitelist")
+
+    def _validate_rate_limit(security_cfg):
+        """Simple in-memory rate limiting."""
+        rl = security_cfg.get("rate_limit")
+        if not rl:
+            return
+
+        limit = int(rl.get("limit", 60))
+        if limit <= 0:
+            raise Exception("rate_limit.limit must be > 0")
+        per = rl.get("per", "minute")
+
+        window = {
+            "second": 1,
+            "minute": 60,
+            "hour": 3600,
+            "day": 86400,
+        }.get(per)
+
+        if not window:
+            raise Exception(f"Invalid rate_limit.per: {per}")
+
+        capacity = limit
+        rate = limit / window
+        cost = 1
+
+        key = f"rl:tb:{agent_id}"
+        now = time.time()
+
+        try:
+            res = REDIS_CONN.lua_token_bucket(
+                keys=[key],
+                args=[capacity, rate, now, cost],
+                client=REDIS_CONN.REDIS,
+            )
+
+            allowed = int(res[0])
+            if allowed != 1:
+                raise Exception("Too many requests (rate limit exceeded)")
+
+        except Exception as e:
+            raise Exception(f"Rate limit error: {e}")
+
+    def _validate_token_auth(security_cfg):
+        """Validate header-based token authentication."""
+        token_cfg = security_cfg.get("token",{})
+        header = token_cfg.get("token_header")
+        token_value = token_cfg.get("token_value")
+
+        provided = request.headers.get(header)
+        if provided != token_value:
+            raise Exception("Invalid token authentication")
+
+    def _validate_basic_auth(security_cfg):
+        """Validate HTTP Basic Auth credentials."""
+        auth_cfg = security_cfg.get("basic_auth", {})
+        username = auth_cfg.get("username")
+        password = auth_cfg.get("password")
+
+        auth = request.authorization
+        if not auth or auth.username != username or auth.password != password:
+            raise Exception("Invalid Basic Auth credentials")
+
+    def _validate_jwt_auth(security_cfg):
+        """Validate JWT token in Authorization header."""
+        jwt_cfg = security_cfg.get("jwt", {})
+        secret = jwt_cfg.get("secret")
+        if not secret:
+            raise Exception("JWT secret not configured")
+
+        auth_header = request.headers.get("Authorization", "")
+        if not auth_header.startswith("Bearer "):
+            raise Exception("Missing Bearer token")
+
+        token = auth_header[len("Bearer "):].strip()
+        if not token:
+            raise Exception("Empty Bearer token")
+
+        alg = (jwt_cfg.get("algorithm") or "HS256").upper()
+
+        decode_kwargs = {
+            "key": secret,
+            "algorithms": [alg],
+        }
+        options = {}
+        if jwt_cfg.get("audience"):
+            decode_kwargs["audience"] = jwt_cfg["audience"]
+            options["verify_aud"] = True
+        else:
+            options["verify_aud"] = False
+
+        if jwt_cfg.get("issuer"):
+            decode_kwargs["issuer"] = jwt_cfg["issuer"]
+            options["verify_iss"] = True
+        else:
+            options["verify_iss"] = False
+        try:
+            decoded = jwt.decode(
+                token,
+                options=options,
+                **decode_kwargs,
+            )
+        except Exception as e:
+            raise Exception(f"Invalid JWT: {str(e)}")
+
+        raw_required_claims = jwt_cfg.get("required_claims", [])
+        if isinstance(raw_required_claims, str):
+            required_claims = [raw_required_claims]
+        elif isinstance(raw_required_claims, (list, tuple, set)):
+            required_claims = list(raw_required_claims)
+        else:
+            required_claims = []
+
+        required_claims = [
+            c for c in required_claims
+            if isinstance(c, str) and c.strip()
+        ]
+
+        RESERVED_CLAIMS = {"exp", "sub", "aud", "iss", "nbf", "iat"}
+        for claim in required_claims:
+            if claim in RESERVED_CLAIMS:
+                raise Exception(f"Reserved JWT claim cannot be required: {claim}")
+
+        for claim in required_claims:
+            if claim not in decoded:
+                raise Exception(f"Missing JWT claim: {claim}")
+
+        return decoded
 
     try:
-        canvas = Canvas(cvs.dsl, tenant_id, agent_id)
+        security_config=webhook_cfg.get("security", {})
+        await validate_webhook_security(security_config)
     except Exception as e:
-        return get_json_result(
-            data=False, message=str(e),
-            code=RetCode.EXCEPTION_ERROR)
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message=str(e)),RetCode.BAD_REQUEST
+    if not isinstance(cvs.dsl, str):
+        dsl = json.dumps(cvs.dsl, ensure_ascii=False)
+    try:
+        canvas = Canvas(dsl, cvs.user_id, agent_id, canvas_id=agent_id)
+    except Exception as e:
+        resp=get_data_error_result(code=RetCode.BAD_REQUEST,message=str(e))
+        resp.status_code = RetCode.BAD_REQUEST
+        return resp
+
+    # 7. Parse request body
+    async def parse_webhook_request(content_type):
+        """Parse request based on content-type and return structured data."""
+
+        # 1. Query
+        query_data = {k: v for k, v in request.args.items()}
+
+        # 2. Headers
+        header_data = {k: v for k, v in request.headers.items()}
+
+        # 3. Body
+        ctype = request.headers.get("Content-Type", "").split(";")[0].strip()
+        if ctype and ctype != content_type:
+            raise ValueError(
+                f"Invalid Content-Type: expect '{content_type}', got '{ctype}'"
+            )
+
+        body_data: dict = {}
 
-    async def sse():
-        nonlocal canvas
         try:
-            async for ans in canvas.run(query=req.get("query", ""), files=req.get("files", []), user_id=req.get("user_id", tenant_id), webhook_payload=req):
-                yield "data:" + json.dumps(ans, ensure_ascii=False) + "\n\n"
+            if ctype == "application/json":
+                body_data = await request.get_json() or {}
 
-            cvs.dsl = json.loads(str(canvas))
-            UserCanvasService.update_by_id(req["id"], cvs.to_dict())
-        except Exception as e:
-            logging.exception(e)
-            yield "data:" + json.dumps({"code": 500, "message": str(e), "data": False}, ensure_ascii=False) + "\n\n"
+            elif ctype == "multipart/form-data":
+                nonlocal canvas
+                form = await request.form
+                files = await request.files
 
-    resp = Response(sse(), mimetype="text/event-stream")
-    resp.headers.add_header("Cache-control", "no-cache")
-    resp.headers.add_header("Connection", "keep-alive")
-    resp.headers.add_header("X-Accel-Buffering", "no")
-    resp.headers.add_header("Content-Type", "text/event-stream; charset=utf-8")
-    return resp
+                body_data = {}
+
+                for key, value in form.items():
+                    body_data[key] = value
+
+                if len(files) > 10:
+                    raise Exception("Too many uploaded files")
+                for key, file in files.items():
+                    desc = FileService.upload_info(
+                        cvs.user_id,           # user
+                        file,              # FileStorage
+                        None                   # url (None for webhook)
+                    )
+                    file_parsed= await canvas.get_files_async([desc])
+                    body_data[key] = file_parsed
+
+            elif ctype == "application/x-www-form-urlencoded":
+                form = await request.form
+                body_data = dict(form)
+
+            else:
+                # text/plain / octet-stream / empty / unknown
+                raw = await request.get_data()
+                if raw:
+                    try:
+                        body_data = json.loads(raw.decode("utf-8"))
+                    except Exception:
+                        body_data = {}
+                else:
+                    body_data = {}
+
+        except Exception:
+            body_data = {}
+
+        return {
+            "query": query_data,
+            "headers": header_data,
+            "body": body_data,
+            "content_type": ctype,
+        }
+
+    def extract_by_schema(data, schema, name="section"):
+        """
+        Extract only fields defined in schema.
+        Required fields must exist.
+        Optional fields default to type-based default values.
+        Type validation included.
+        """
+        props = schema.get("properties", {})
+        required = schema.get("required", [])
+
+        extracted = {}
+
+        for field, field_schema in props.items():
+            field_type = field_schema.get("type")
+
+            # 1. Required field missing
+            if field in required and field not in data:
+                raise Exception(f"{name} missing required field: {field}")
+
+            # 2. Optional → default value
+            if field not in data:
+                extracted[field] = default_for_type(field_type)
+                continue
+
+            raw_value = data[field]
+
+            # 3. Auto convert value
+            try:
+                value = auto_cast_value(raw_value, field_type)
+            except Exception as e:
+                raise Exception(f"{name}.{field} auto-cast failed: {str(e)}")
+
+            # 4. Type validation
+            if not validate_type(value, field_type):
+                raise Exception(
+                    f"{name}.{field} type mismatch: expected {field_type}, got {type(value).__name__}"
+                )
+
+            extracted[field] = value
+
+        return extracted
+
+
+    def default_for_type(t):
+        """Return default value for the given schema type."""
+        if t == "file":
+            return []
+        if t == "object":
+            return {}
+        if t == "boolean":
+            return False
+        if t == "number":
+            return 0
+        if t == "string":
+            return ""
+        if t and t.startswith("array"):
+            return []
+        if t == "null":
+            return None
+        return None
+
+    def auto_cast_value(value, expected_type):
+        """Convert string values into schema type when possible."""
+
+        # Non-string values already good
+        if not isinstance(value, str):
+            return value
+
+        v = value.strip()
+
+        # Boolean
+        if expected_type == "boolean":
+            if v.lower() in ["true", "1"]:
+                return True
+            if v.lower() in ["false", "0"]:
+                return False
+            raise Exception(f"Cannot convert '{value}' to boolean")
+
+        # Number
+        if expected_type == "number":
+            # integer
+            if v.isdigit() or (v.startswith("-") and v[1:].isdigit()):
+                return int(v)
+
+            # float
+            try:
+                return float(v)
+            except Exception:
+                raise Exception(f"Cannot convert '{value}' to number")
+
+        # Object
+        if expected_type == "object":
+            try:
+                parsed = json.loads(v)
+                if isinstance(parsed, dict):
+                    return parsed
+                else:
+                    raise Exception("JSON is not an object")
+            except Exception:
+                raise Exception(f"Cannot convert '{value}' to object")
+
+        # Array <T>
+        if expected_type.startswith("array"):
+            try:
+                parsed = json.loads(v)
+                if isinstance(parsed, list):
+                    return parsed
+                else:
+                    raise Exception("JSON is not an array")
+            except Exception:
+                raise Exception(f"Cannot convert '{value}' to array")
+
+        # String (accept original)
+        if expected_type == "string":
+            return value
+
+        # File
+        if expected_type == "file":
+            return value
+        # Default: do nothing
+        return value
+
+
+    def validate_type(value, t):
+        """Validate value type against schema type t."""
+        if t == "file":
+            return isinstance(value, list)
+
+        if t == "string":
+            return isinstance(value, str)
+
+        if t == "number":
+            return isinstance(value, (int, float))
+
+        if t == "boolean":
+            return isinstance(value, bool)
+
+        if t == "object":
+            return isinstance(value, dict)
+
+        # array<string> / array<number> / array<object>
+        if t.startswith("array"):
+            if not isinstance(value, list):
+                return False
+
+            if "<" in t and ">" in t:
+                inner = t[t.find("<") + 1 : t.find(">")]
+
+                # Check each element type
+                for item in value:
+                    if not validate_type(item, inner):
+                        return False
+
+            return True
+
+        return True
+    parsed = await parse_webhook_request(webhook_cfg.get("content_types"))
+    SCHEMA = webhook_cfg.get("schema", {"query": {}, "headers": {}, "body": {}})
+
+    # Extract strictly by schema
+    try:
+        query_clean  = extract_by_schema(parsed["query"],   SCHEMA.get("query", {}),  name="query")
+        header_clean = extract_by_schema(parsed["headers"], SCHEMA.get("headers", {}), name="headers")
+        body_clean   = extract_by_schema(parsed["body"],    SCHEMA.get("body", {}),    name="body")
+    except Exception as e:
+        return get_data_error_result(code=RetCode.BAD_REQUEST,message=str(e)),RetCode.BAD_REQUEST
+
+    clean_request = {
+        "query": query_clean,
+        "headers": header_clean,
+        "body": body_clean,
+        "input": parsed
+    }
+
+    execution_mode = webhook_cfg.get("execution_mode", "Immediately")
+    response_cfg = webhook_cfg.get("response", {})
+
+    def append_webhook_trace(agent_id: str, start_ts: float,event: dict, ttl=600):
+        key = f"webhook-trace-{agent_id}-logs"
+
+        raw = REDIS_CONN.get(key)
+        obj = json.loads(raw) if raw else {"webhooks": {}}
+
+        ws = obj["webhooks"].setdefault(
+            str(start_ts),
+            {"start_ts": start_ts, "events": []}
+        )
+
+        ws["events"].append({
+            "ts": time.time(),
+            **event
+        })
+
+        REDIS_CONN.set_obj(key, obj, ttl)
+
+    if execution_mode == "Immediately":
+        status = response_cfg.get("status", 200)
+        try:
+            status = int(status)
+        except (TypeError, ValueError):
+            return get_data_error_result(code=RetCode.BAD_REQUEST,message=str(f"Invalid response status code: {status}")),RetCode.BAD_REQUEST
+
+        if not (200 <= status <= 399):
+            return get_data_error_result(code=RetCode.BAD_REQUEST,message=str(f"Invalid response status code: {status}, must be between 200 and 399")),RetCode.BAD_REQUEST
+
+        body_tpl = response_cfg.get("body_template", "")
+
+        def parse_body(body: str):
+            if not body:
+                return None, "application/json"
+
+            try:
+                parsed = json.loads(body)
+                return parsed, "application/json"
+            except (json.JSONDecodeError, TypeError):
+                return body, "text/plain"
+
+
+        body, content_type = parse_body(body_tpl)
+        resp = Response(
+            json.dumps(body, ensure_ascii=False) if content_type == "application/json" else body,
+            status=status,
+            content_type=content_type,
+        )
+
+        async def background_run():
+            try:
+                async for ans in canvas.run(
+                    query="",
+                    user_id=cvs.user_id,
+                    webhook_payload=clean_request
+                ):
+                    if is_test:
+                        append_webhook_trace(agent_id, start_ts, ans)
+
+                if is_test:
+                    append_webhook_trace(
+                        agent_id,
+                        start_ts,
+                        {
+                            "event": "finished",
+                            "elapsed_time": time.time() - start_ts,
+                            "success": True,
+                        }
+                    )
+
+                cvs.dsl = json.loads(str(canvas))
+                UserCanvasService.update_by_id(cvs.user_id, cvs.to_dict())
+
+            except Exception as e:
+                logging.exception("Webhook background run failed")
+                if is_test:
+                    try:
+                        append_webhook_trace(
+                            agent_id,
+                            start_ts,
+                            {
+                                "event": "error",
+                                "message": str(e),
+                                "error_type": type(e).__name__,
+                            }
+                        )
+                        append_webhook_trace(
+                            agent_id,
+                            start_ts,
+                            {
+                                "event": "finished",
+                                "elapsed_time": time.time() - start_ts,
+                                "success": False,
+                            }
+                        )
+                    except Exception:
+                        logging.exception("Failed to append webhook trace")
+
+        asyncio.create_task(background_run())
+        return resp
+    else:
+        async def sse():
+            nonlocal canvas
+            contents: list[str] = []
+            status = 200
+            try:
+                async for ans in canvas.run(
+                    query="",
+                    user_id=cvs.user_id,
+                    webhook_payload=clean_request,
+                ):
+                    if ans["event"] == "message":
+                        content = ans["data"]["content"]
+                        if ans["data"].get("start_to_think", False):
+                            content = "<think>"
+                        elif ans["data"].get("end_to_think", False):
+                            content = "</think>"
+                        if content:
+                            contents.append(content)
+                    if ans["event"] == "message_end":
+                        status = int(ans["data"].get("status", status))
+                    if is_test:
+                        append_webhook_trace(
+                            agent_id,
+                            start_ts,
+                            ans
+                        )
+                if is_test:
+                    append_webhook_trace(
+                        agent_id,
+                        start_ts,
+                        {
+                            "event": "finished",
+                            "elapsed_time": time.time() - start_ts,
+                            "success": True,
+                        }
+                    )
+                final_content = "".join(contents)
+                return {
+                    "message": final_content,
+                    "success": True,
+                    "code":  status,
+                }
+
+            except Exception as e:
+                if is_test:
+                    append_webhook_trace(
+                        agent_id,
+                        start_ts,
+                        {
+                            "event": "error",
+                            "message": str(e),
+                            "error_type": type(e).__name__,
+                        }
+                    )
+                    append_webhook_trace(
+                        agent_id,
+                        start_ts,
+                        {
+                            "event": "finished",
+                            "elapsed_time": time.time() - start_ts,
+                            "success": False,
+                        }
+                    )
+                return {"code": 400, "message": str(e),"success":False}
+
+        result = await sse()
+        return Response(
+            json.dumps(result),
+            status=result["code"],
+            mimetype="application/json",
+        )
+
+
+@manager.route("/webhook_trace/<agent_id>", methods=["GET"])  # noqa: F821
+async def webhook_trace(agent_id: str):
+    def encode_webhook_id(start_ts: str) -> str:
+        WEBHOOK_ID_SECRET = "webhook_id_secret"
+        sig = hmac.new(
+            WEBHOOK_ID_SECRET.encode("utf-8"),
+            start_ts.encode("utf-8"),
+            hashlib.sha256,
+        ).digest()
+        return base64.urlsafe_b64encode(sig).decode("utf-8").rstrip("=")
+
+    def decode_webhook_id(enc_id: str, webhooks: dict) -> str | None:
+        for ts in webhooks.keys():
+            if encode_webhook_id(ts) == enc_id:
+                return ts
+        return None
+    since_ts = request.args.get("since_ts", type=float)
+    webhook_id = request.args.get("webhook_id")
+
+    key = f"webhook-trace-{agent_id}-logs"
+    raw = REDIS_CONN.get(key)
+
+    if since_ts is None:
+        now = time.time()
+        return get_json_result(
+            data={
+                "webhook_id": None,
+                "events": [],
+                "next_since_ts": now,
+                "finished": False,
+            }
+        )
+
+    if not raw:
+        return get_json_result(
+            data={
+                "webhook_id": None,
+                "events": [],
+                "next_since_ts": since_ts,
+                "finished": False,
+            }
+        )
+
+    obj = json.loads(raw)
+    webhooks = obj.get("webhooks", {})
+
+    if webhook_id is None:
+        candidates = [
+            float(k) for k in webhooks.keys() if float(k) > since_ts
+        ]
+
+        if not candidates:
+            return get_json_result(
+                data={
+                    "webhook_id": None,
+                    "events": [],
+                    "next_since_ts": since_ts,
+                    "finished": False,
+                }
+            )
+
+        start_ts = min(candidates)
+        real_id = str(start_ts)
+        webhook_id = encode_webhook_id(real_id)
+
+        return get_json_result(
+            data={
+                "webhook_id": webhook_id,
+                "events": [],
+                "next_since_ts": start_ts,
+                "finished": False,
+            }
+        )
+
+    real_id = decode_webhook_id(webhook_id, webhooks)
+
+    if not real_id:
+        return get_json_result(
+            data={
+                "webhook_id": webhook_id,
+                "events": [],
+                "next_since_ts": since_ts,
+                "finished": True,
+            }
+        )
+
+    ws = webhooks.get(str(real_id))
+    events = ws.get("events", [])
+    new_events = [e for e in events if e.get("ts", 0) > since_ts]
+
+    next_ts = since_ts
+    for e in new_events:
+        next_ts = max(next_ts, e["ts"])
+
+    finished = any(e.get("event") == "finished" for e in new_events)
+
+    return get_json_result(
+        data={
+            "webhook_id": webhook_id,
+            "events": new_events,
+            "next_since_ts": next_ts,
+            "finished": finished,
+        }
+    )

api/apps/sdk/chat.py

@@ -252,7 +252,6 @@ async def delete_chats(tenant_id):
             continue
         temp_dict = {"status": StatusEnum.INVALID.value}
         success_count += DialogService.update_by_id(id, temp_dict)
-        print(success_count, "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$", flush=True)
 
     if errors:
         if success_count > 0:
@@ -288,7 +287,7 @@ def list_chat(tenant_id):
     chats = DialogService.get_list(tenant_id, page_number, items_per_page, orderby, desc, id, name)
     if not chats:
         return get_result(data=[])
-    list_assts = []
+    list_assistants = []
     key_mapping = {
         "parameters": "variables",
         "prologue": "opener",
@@ -322,5 +321,5 @@ def list_chat(tenant_id):
         del res["kb_ids"]
         res["datasets"] = kb_list
         res["avatar"] = res.pop("icon")
-        list_assts.append(res)
-    return get_result(data=list_assts)
+        list_assistants.append(res)
+    return get_result(data=list_assistants)

api/apps/sdk/dataset.py

@@ -495,7 +495,7 @@ def knowledge_graph(tenant_id, dataset_id):
     }
 
     obj = {"graph": {}, "mind_map": {}}
-    if not settings.docStoreConn.indexExist(search.index_name(kb.tenant_id), dataset_id):
+    if not settings.docStoreConn.index_exist(search.index_name(kb.tenant_id), dataset_id):
         return get_result(data=obj)
     sres = settings.retriever.search(req, search.index_name(kb.tenant_id), [dataset_id])
     if not len(sres.ids):

api/apps/sdk/dify_retrieval.py

@@ -150,7 +150,7 @@ async def retrieval(tenant_id):
         )
 
         if use_kg:
-            ck = settings.kg_retriever.retrieval(question,
+            ck = await settings.kg_retriever.retrieval(question,
                                                  [tenant_id],
                                                  [kb_id],
                                                  embd_mdl,

api/apps/sdk/doc.py

@@ -1080,7 +1080,7 @@ def list_chunks(tenant_id, dataset_id, document_id):
         res["chunks"].append(final_chunk)
         _ = Chunk(**final_chunk)
 
-    elif settings.docStoreConn.indexExist(search.index_name(tenant_id), dataset_id):
+    elif settings.docStoreConn.index_exist(search.index_name(tenant_id), dataset_id):
         sres = settings.retriever.search(query, search.index_name(tenant_id), [dataset_id], emb_mdl=None, highlight=True)
         res["total"] = sres.total
         for id in sres.ids:
@@ -1286,6 +1286,9 @@ async def rm_chunk(tenant_id, dataset_id, document_id):
     if "chunk_ids" in req:
         unique_chunk_ids, duplicate_messages = check_duplicate_ids(req["chunk_ids"], "chunk")
         condition["id"] = unique_chunk_ids
+    else:
+        unique_chunk_ids = []
+        duplicate_messages = []
     chunk_number = settings.docStoreConn.delete(condition, search.index_name(tenant_id), dataset_id)
     if chunk_number != 0:
         DocumentService.decrement_chunk_num(document_id, dataset_id, 1, chunk_number, 0)
@@ -1576,7 +1579,7 @@ async def retrieval_test(tenant_id):
             if cks:
                 ranks["chunks"] = cks
         if use_kg:
-            ck = settings.kg_retriever.retrieval(question, [k.tenant_id for k in kbs], kb_ids, embd_mdl, LLMBundle(kb.tenant_id, LLMType.CHAT))
+            ck = await settings.kg_retriever.retrieval(question, [k.tenant_id for k in kbs], kb_ids, embd_mdl, LLMBundle(kb.tenant_id, LLMType.CHAT))
             if ck["content_with_weight"]:
                 ranks["chunks"].insert(0, ck)
 

api/apps/sdk/files.py

@@ -205,7 +205,8 @@ async def create(tenant_id):
         if not FileService.is_parent_folder_exist(pf_id):
             return get_json_result(data=False, message="Parent Folder Doesn't Exist!", code=RetCode.BAD_REQUEST)
         if FileService.query(name=req["name"], parent_id=pf_id):
-            return get_json_result(data=False, message="Duplicated folder name in the same folder.", code=409)
+            return get_json_result(data=False, message="Duplicated folder name in the same folder.",
+                                   code=RetCode.CONFLICT)
 
         if input_file_type == FileType.FOLDER.value:
             file_type = FileType.FOLDER.value
@@ -565,11 +566,13 @@ async def rename(tenant_id):
 
         if file.type != FileType.FOLDER.value and pathlib.Path(req["name"].lower()).suffix != pathlib.Path(
                 file.name.lower()).suffix:
-            return get_json_result(data=False, message="The extension of file can't be changed", code=RetCode.BAD_REQUEST)
+            return get_json_result(data=False, message="The extension of file can't be changed",
+                                   code=RetCode.BAD_REQUEST)
 
         for existing_file in FileService.query(name=req["name"], pf_id=file.parent_id):
             if existing_file.name == req["name"]:
-                return get_json_result(data=False, message="Duplicated file name in the same folder.", code=409)
+                return get_json_result(data=False, message="Duplicated file name in the same folder.",
+                                       code=RetCode.CONFLICT)
 
         if not FileService.update_by_id(req["file_id"], {"name": req["name"]}):
             return get_json_result(message="Database error (File rename)!", code=RetCode.SERVER_ERROR)
@@ -631,9 +634,10 @@ async def get(tenant_id, file_id):
     except Exception as e:
         return server_error_response(e)
 
+
 @manager.route("/file/download/<attachment_id>", methods=["GET"])  # noqa: F821
 @token_required
-async def download_attachment(tenant_id,attachment_id):
+async def download_attachment(tenant_id, attachment_id):
     try:
         ext = request.args.get("ext", "markdown")
         data = await asyncio.to_thread(settings.STORAGE_IMPL.get, tenant_id, attachment_id)
@@ -645,6 +649,7 @@ async def download_attachment(tenant_id,attachment_id):
     except Exception as e:
         return server_error_response(e)
 
+
 @manager.route('/file/mv', methods=['POST'])  # noqa: F821
 @token_required
 async def move(tenant_id):

api/apps/sdk/session.py

@@ -14,6 +14,7 @@
 #  limitations under the License.
 #
 import json
+import copy
 import re
 import time
 
@@ -32,7 +33,7 @@ from api.db.services.dialog_service import DialogService, async_ask, async_chat,
 from api.db.services.document_service import DocumentService
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import LLMBundle
-from common.metadata_utils import apply_meta_data_filter
+from common.metadata_utils import apply_meta_data_filter, convert_conditions, meta_filter
 from api.db.services.search_service import SearchService
 from api.db.services.user_service import UserTenantService
 from common.misc_utils import get_uuid
@@ -59,7 +60,7 @@ async def create(tenant_id, chat_id):
         "name": req.get("name", "New session"),
         "message": [{"role": "assistant", "content": dia[0].prompt_config.get("prologue")}],
         "user_id": req.get("user_id", ""),
-        "reference": [{}],
+        "reference": [],
     }
     if not conv.get("name"):
         return get_error_data_result(message="`name` can not be empty.")
@@ -87,7 +88,7 @@ async def create_agent_session(tenant_id, agent_id):
         cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
 
     session_id = get_uuid()
-    canvas = Canvas(cvs.dsl, tenant_id, agent_id)
+    canvas = Canvas(cvs.dsl, tenant_id, agent_id, canvas_id=cvs.id)
     canvas.reset()
 
     cvs.dsl = json.loads(str(canvas))
@@ -128,11 +129,33 @@ async def chat_completion(tenant_id, chat_id):
         req = {"question": ""}
     if not req.get("session_id"):
         req["question"] = ""
-    if not DialogService.query(tenant_id=tenant_id, id=chat_id, status=StatusEnum.VALID.value):
+    dia = DialogService.query(tenant_id=tenant_id, id=chat_id, status=StatusEnum.VALID.value)
+    if not dia:
         return get_error_data_result(f"You don't own the chat {chat_id}")
+    dia = dia[0]
     if req.get("session_id"):
         if not ConversationService.query(id=req["session_id"], dialog_id=chat_id):
             return get_error_data_result(f"You don't own the session {req['session_id']}")
+
+    metadata_condition = req.get("metadata_condition") or {}
+    if metadata_condition and not isinstance(metadata_condition, dict):
+        return get_error_data_result(message="metadata_condition must be an object.")
+
+    if metadata_condition and req.get("question"):
+        metas = DocumentService.get_meta_by_kbs(dia.kb_ids or [])
+        filtered_doc_ids = meta_filter(
+            metas,
+            convert_conditions(metadata_condition),
+            metadata_condition.get("logic", "and"),
+        )
+        if metadata_condition.get("conditions") and not filtered_doc_ids:
+            filtered_doc_ids = ["-999"]
+
+        if filtered_doc_ids:
+            req["doc_ids"] = ",".join(filtered_doc_ids)
+        else:
+            req.pop("doc_ids", None)
+
     if req.get("stream", True):
         resp = Response(rag_completion(tenant_id, chat_id, **req), mimetype="text/event-stream")
         resp.headers.add_header("Cache-control", "no-cache")
@@ -195,7 +218,19 @@ async def chat_completion_openai_like(tenant_id, chat_id):
             {"role": "user", "content": "Can you tell me how to install neovim"},
         ],
         stream=stream,
-        extra_body={"reference": reference}
+        extra_body={
+            "reference": reference,
+            "metadata_condition": {
+                "logic": "and",
+                "conditions": [
+                    {
+                        "name": "author",
+                        "comparison_operator": "is",
+                        "value": "bob"
+                    }
+                ]
+            }
+        }
     )
 
     if stream:
@@ -211,7 +246,11 @@ async def chat_completion_openai_like(tenant_id, chat_id):
     """
     req = await get_request_json()
 
-    need_reference = bool(req.get("reference", False))
+    extra_body = req.get("extra_body") or {}
+    if extra_body and not isinstance(extra_body, dict):
+        return get_error_data_result("extra_body must be an object.")
+
+    need_reference = bool(extra_body.get("reference", False))
 
     messages = req.get("messages", [])
     # To prevent empty [] input
@@ -229,6 +268,22 @@ async def chat_completion_openai_like(tenant_id, chat_id):
         return get_error_data_result(f"You don't own the chat {chat_id}")
     dia = dia[0]
 
+    metadata_condition = extra_body.get("metadata_condition") or {}
+    if metadata_condition and not isinstance(metadata_condition, dict):
+        return get_error_data_result(message="metadata_condition must be an object.")
+
+    doc_ids_str = None
+    if metadata_condition:
+        metas = DocumentService.get_meta_by_kbs(dia.kb_ids or [])
+        filtered_doc_ids = meta_filter(
+            metas,
+            convert_conditions(metadata_condition),
+            metadata_condition.get("logic", "and"),
+        )
+        if metadata_condition.get("conditions") and not filtered_doc_ids:
+            filtered_doc_ids = ["-999"]
+        doc_ids_str = ",".join(filtered_doc_ids) if filtered_doc_ids else None
+
     # Filter system and non-sense assistant messages
     msg = []
     for m in messages:
@@ -276,14 +331,17 @@ async def chat_completion_openai_like(tenant_id, chat_id):
             }
 
             try:
-                async for ans in async_chat(dia, msg, True, toolcall_session=toolcall_session, tools=tools, quote=need_reference):
+                chat_kwargs = {"toolcall_session": toolcall_session, "tools": tools, "quote": need_reference}
+                if doc_ids_str:
+                    chat_kwargs["doc_ids"] = doc_ids_str
+                async for ans in async_chat(dia, msg, True, **chat_kwargs):
                     last_ans = ans
                     answer = ans["answer"]
 
                     reasoning_match = re.search(r"<think>(.*?)</think>", answer, flags=re.DOTALL)
                     if reasoning_match:
                         reasoning_part = reasoning_match.group(1)
-                        content_part = answer[reasoning_match.end():]
+                        content_part = answer[reasoning_match.end() :]
                     else:
                         reasoning_part = ""
                         content_part = answer
@@ -328,8 +386,7 @@ async def chat_completion_openai_like(tenant_id, chat_id):
             response["choices"][0]["delta"]["content"] = None
             response["choices"][0]["delta"]["reasoning_content"] = None
             response["choices"][0]["finish_reason"] = "stop"
-            response["usage"] = {"prompt_tokens": len(prompt), "completion_tokens": token_used,
-                                 "total_tokens": len(prompt) + token_used}
+            response["usage"] = {"prompt_tokens": len(prompt), "completion_tokens": token_used, "total_tokens": len(prompt) + token_used}
             if need_reference:
                 response["choices"][0]["delta"]["reference"] = chunks_format(last_ans.get("reference", []))
                 response["choices"][0]["delta"]["final_content"] = last_ans.get("answer", "")
@@ -344,7 +401,10 @@ async def chat_completion_openai_like(tenant_id, chat_id):
         return resp
     else:
         answer = None
-        async for ans in async_chat(dia, msg, False, toolcall_session=toolcall_session, tools=tools, quote=need_reference):
+        chat_kwargs = {"toolcall_session": toolcall_session, "tools": tools, "quote": need_reference}
+        if doc_ids_str:
+            chat_kwargs["doc_ids"] = doc_ids_str
+        async for ans in async_chat(dia, msg, False, **chat_kwargs):
             # focus answer content only
             answer = ans
             break
@@ -388,7 +448,7 @@ async def chat_completion_openai_like(tenant_id, chat_id):
 @token_required
 async def agents_completion_openai_compatibility(tenant_id, agent_id):
     req = await get_request_json()
-    tiktokenenc = tiktoken.get_encoding("cl100k_base")
+    tiktoken_encode = tiktoken.get_encoding("cl100k_base")
     messages = req.get("messages", [])
     if not messages:
         return get_error_data_result("You must provide at least one message.")
@@ -396,7 +456,7 @@ async def agents_completion_openai_compatibility(tenant_id, agent_id):
         return get_error_data_result(f"You don't own the agent {agent_id}")
 
     filtered_messages = [m for m in messages if m["role"] in ["user", "assistant"]]
-    prompt_tokens = sum(len(tiktokenenc.encode(m["content"])) for m in filtered_messages)
+    prompt_tokens = sum(len(tiktoken_encode.encode(m["content"])) for m in filtered_messages)
     if not filtered_messages:
         return jsonify(
             get_data_openai(
@@ -404,7 +464,7 @@ async def agents_completion_openai_compatibility(tenant_id, agent_id):
                 content="No valid messages found (user or assistant).",
                 finish_reason="stop",
                 model=req.get("model", ""),
-                completion_tokens=len(tiktokenenc.encode("No valid messages found (user or assistant).")),
+                completion_tokens=len(tiktoken_encode.encode("No valid messages found (user or assistant).")),
                 prompt_tokens=prompt_tokens,
             )
         )
@@ -441,15 +501,19 @@ async def agents_completion_openai_compatibility(tenant_id, agent_id):
             ):
             return jsonify(response)
 
+        return None
+
 
 @manager.route("/agents/<agent_id>/completions", methods=["POST"])  # noqa: F821
 @token_required
 async def agent_completions(tenant_id, agent_id):
     req = await get_request_json()
+    return_trace = bool(req.get("return_trace", False))
 
     if req.get("stream", True):
 
         async def generate():
+            trace_items = []
             async for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
                 if isinstance(answer, str):
                     try:
@@ -457,7 +521,21 @@ async def agent_completions(tenant_id, agent_id):
                     except Exception:
                         continue
 
-                if ans.get("event") not in ["message", "message_end"]:
+                event = ans.get("event")
+                if event == "node_finished":
+                    if return_trace:
+                        data = ans.get("data", {})
+                        trace_items.append(
+                            {
+                                "component_id": data.get("component_id"),
+                                "trace": [copy.deepcopy(data)],
+                            }
+                        )
+                        ans.setdefault("data", {})["trace"] = trace_items
+                        answer = "data:" + json.dumps(ans, ensure_ascii=False) + "\n\n"
+                    yield answer
+
+                if event not in ["message", "message_end"]:
                     continue
 
                 yield answer
@@ -474,6 +552,7 @@ async def agent_completions(tenant_id, agent_id):
     full_content = ""
     reference = {}
     final_ans = ""
+    trace_items = []
     async for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
         try:
             ans = json.loads(answer[5:])
@@ -484,11 +563,22 @@ async def agent_completions(tenant_id, agent_id):
             if ans.get("data", {}).get("reference", None):
                 reference.update(ans["data"]["reference"])
 
+            if return_trace and ans.get("event") == "node_finished":
+                data = ans.get("data", {})
+                trace_items.append(
+                    {
+                        "component_id": data.get("component_id"),
+                        "trace": [copy.deepcopy(data)],
+                    }
+                )
+
             final_ans = ans
         except Exception as e:
             return get_result(data=f"**ERROR**: {str(e)}")
     final_ans["data"]["content"] = full_content
     final_ans["data"]["reference"] = reference
+    if return_trace and final_ans:
+        final_ans["data"]["trace"] = trace_items
     return get_result(data=final_ans)
 
 
@@ -832,6 +922,7 @@ async def chatbot_completions(dialog_id):
     async for answer in iframe_completion(dialog_id, **req):
         return get_result(data=answer)
 
+    return None
 
 @manager.route("/chatbots/<dialog_id>/info", methods=["GET"])  # noqa: F821
 async def chatbots_inputs(dialog_id):
@@ -879,6 +970,7 @@ async def agent_bot_completions(agent_id):
     async for answer in agent_completion(objs[0].tenant_id, agent_id, **req):
         return get_result(data=answer)
 
+    return None
 
 @manager.route("/agentbots/<agent_id>/inputs", methods=["GET"])  # noqa: F821
 async def begin_inputs(agent_id):
@@ -894,7 +986,7 @@ async def begin_inputs(agent_id):
     if not e:
         return get_error_data_result(f"Can't find agent by ID: {agent_id}")
 
-    canvas = Canvas(json.dumps(cvs.dsl), objs[0].tenant_id)
+    canvas = Canvas(json.dumps(cvs.dsl), objs[0].tenant_id, canvas_id=cvs.id)
     return get_result(
         data={"title": cvs.title, "avatar": cvs.avatar, "inputs": canvas.get_component_input_form("begin"),
               "prologue": canvas.get_prologue(), "mode": canvas.get_mode()})
@@ -1024,7 +1116,7 @@ async def retrieval_test_embedded():
             local_doc_ids, rerank_mdl=rerank_mdl, highlight=req.get("highlight"), rank_feature=labels
         )
         if use_kg:
-            ck = settings.kg_retriever.retrieval(_question, tenant_ids, kb_ids, embd_mdl,
+            ck = await settings.kg_retriever.retrieval(_question, tenant_ids, kb_ids, embd_mdl,
                                                  LLMBundle(kb.tenant_id, LLMType.CHAT))
             if ck["content_with_weight"]:
                 ranks["chunks"].insert(0, ck)

api/apps/system_app.py

@@ -177,7 +177,7 @@ def healthz():
     return jsonify(result), (200 if all_ok else 500)
 
 
-@manager.route("/ping", methods=["GET"]) # noqa: F821
+@manager.route("/ping", methods=["GET"])  # noqa: F821
 def ping():
     return "pong", 200
 
@@ -213,7 +213,7 @@ def new_token():
         if not tenants:
             return get_data_error_result(message="Tenant not found!")
 
-        tenant_id = [tenant for tenant in tenants if tenant.role == 'owner'][0].tenant_id
+        tenant_id = [tenant for tenant in tenants if tenant.role == "owner"][0].tenant_id
         obj = {
             "tenant_id": tenant_id,
             "token": generate_confirmation_token(),
@@ -268,13 +268,12 @@ def token_list():
         if not tenants:
             return get_data_error_result(message="Tenant not found!")
 
-        tenant_id = [tenant for tenant in tenants if tenant.role == 'owner'][0].tenant_id
+        tenant_id = [tenant for tenant in tenants if tenant.role == "owner"][0].tenant_id
         objs = APITokenService.query(tenant_id=tenant_id)
         objs = [o.to_dict() for o in objs]
         for o in objs:
             if not o["beta"]:
-                o["beta"] = generate_confirmation_token().replace(
-                    "ragflow-", "")[:32]
+                o["beta"] = generate_confirmation_token().replace("ragflow-", "")[:32]
                 APITokenService.filter_update([APIToken.tenant_id == tenant_id, APIToken.token == o["token"]], o)
         return get_json_result(data=objs)
     except Exception as e:
@@ -307,13 +306,19 @@ def rm(token):
               type: boolean
               description: Deletion status.
     """
-    APITokenService.filter_delete(
-        [APIToken.tenant_id == current_user.id, APIToken.token == token]
-    )
-    return get_json_result(data=True)
+    try:
+        tenants = UserTenantService.query(user_id=current_user.id)
+        if not tenants:
+            return get_data_error_result(message="Tenant not found!")
+
+        tenant_id = tenants[0].tenant_id
+        APITokenService.filter_delete([APIToken.tenant_id == tenant_id, APIToken.token == token])
+        return get_json_result(data=True)
+    except Exception as e:
+        return server_error_response(e)
 
 
-@manager.route('/config', methods=['GET'])  # noqa: F821
+@manager.route("/config", methods=["GET"])  # noqa: F821
 def get_config():
     """
     Get system configuration.
@@ -330,6 +335,4 @@ def get_config():
                         type: integer 0 means disabled, 1 means enabled
                         description: Whether user registration is enabled
     """
-    return get_json_result(data={
-        "registerEnabled": settings.REGISTER_ENABLED
-    })
+    return get_json_result(data={"registerEnabled": settings.REGISTER_ENABLED})

api/apps/user_app.py

@@ -660,7 +660,7 @@ def user_register(user_id, user):
     tenant_llm = get_init_tenant_llm(user_id)
 
     if not UserService.save(**user):
-        return
+        return None
     TenantService.insert(**tenant)
     UserTenantService.insert(**usr_tenant)
     TenantLLMService.insert_many(tenant_llm)

api/db/db_models.py

@@ -1189,7 +1189,7 @@ class Memory(DataBaseModel):
     permissions = CharField(max_length=16, null=False, index=True, help_text="me|team", default="me")
     description = TextField(null=True, help_text="description")
     memory_size = IntegerField(default=5242880, null=False, index=False)
-    forgetting_policy = CharField(max_length=32, null=False, default="fifo", index=False, help_text="lru|fifo")
+    forgetting_policy = CharField(max_length=32, null=False, default="FIFO", index=False, help_text="LRU|FIFO")
     temperature = FloatField(default=0.5, index=False)
     system_prompt = TextField(null=True, help_text="system prompt", index=False)
     user_prompt = TextField(null=True, help_text="user prompt", index=False)

api/db/init_data.py

@@ -30,6 +30,7 @@ from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.tenant_llm_service import LLMFactoriesService, TenantLLMService
 from api.db.services.llm_service import LLMService, LLMBundle, get_init_tenant_llm
 from api.db.services.user_service import TenantService, UserTenantService
+from api.db.joint_services.memory_message_service import init_message_id_sequence, init_memory_size_cache
 from common.constants import LLMType
 from common.file_utils import get_project_base_directory
 from common import settings
@@ -169,6 +170,8 @@ def init_web_data():
     #    init_superuser()
 
     add_graph_templates()
+    init_message_id_sequence()
+    init_memory_size_cache()
     logging.info("init web data success:{}".format(time.time() - start_time))
 
 

api/db/joint_services/memory_message_service.py

@@ -0,0 +1,389 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import logging
+from typing import List
+
+from api.db.services.task_service import TaskService
+from common import settings
+from common.time_utils import current_timestamp, timestamp_to_date, format_iso_8601_to_ymd_hms
+from common.constants import MemoryType, LLMType
+from common.doc_store.doc_store_base import FusionExpr
+from common.misc_utils import get_uuid
+from api.db.db_utils import bulk_insert_into_db
+from api.db.db_models import Task
+from api.db.services.memory_service import MemoryService
+from api.db.services.tenant_llm_service import TenantLLMService
+from api.db.services.llm_service import LLMBundle
+from api.utils.memory_utils import get_memory_type_human
+from memory.services.messages import MessageService
+from memory.services.query import MsgTextQuery, get_vector
+from memory.utils.prompt_util import PromptAssembler
+from memory.utils.msg_util import get_json_result_from_llm_response
+from rag.utils.redis_conn import REDIS_CONN
+
+
+async def save_to_memory(memory_id: str, message_dict: dict):
+    """
+    :param memory_id:
+    :param message_dict: {
+        "user_id": str,
+        "agent_id": str,
+        "session_id": str,
+        "user_input": str,
+        "agent_response": str
+    }
+    """
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return False, f"Memory '{memory_id}' not found."
+
+    tenant_id = memory.tenant_id
+    extracted_content = await extract_by_llm(
+        tenant_id,
+        memory.llm_id,
+        {"temperature": memory.temperature},
+        get_memory_type_human(memory.memory_type),
+        message_dict.get("user_input", ""),
+        message_dict.get("agent_response", "")
+    ) if memory.memory_type != MemoryType.RAW.value else []  # if only RAW, no need to extract
+    raw_message_id = REDIS_CONN.generate_auto_increment_id(namespace="memory")
+    message_list = [{
+        "message_id": raw_message_id,
+        "message_type": MemoryType.RAW.name.lower(),
+        "source_id": 0,
+        "memory_id": memory_id,
+        "user_id": "",
+        "agent_id": message_dict["agent_id"],
+        "session_id": message_dict["session_id"],
+        "content": f"User Input: {message_dict.get('user_input')}\nAgent Response: {message_dict.get('agent_response')}",
+        "valid_at": timestamp_to_date(current_timestamp()),
+        "invalid_at": None,
+        "forget_at": None,
+        "status": True
+    }, *[{
+        "message_id": REDIS_CONN.generate_auto_increment_id(namespace="memory"),
+        "message_type": content["message_type"],
+        "source_id": raw_message_id,
+        "memory_id": memory_id,
+        "user_id": "",
+        "agent_id": message_dict["agent_id"],
+        "session_id": message_dict["session_id"],
+        "content": content["content"],
+        "valid_at": content["valid_at"],
+        "invalid_at": content["invalid_at"] if content["invalid_at"] else None,
+        "forget_at": None,
+        "status": True
+    } for content in extracted_content]]
+    return await embed_and_save(memory, message_list)
+
+
+async def save_extracted_to_memory_only(memory_id: str, message_dict, source_message_id: int):
+    memory = MemoryService.get_by_memory_id(memory_id)
+    if not memory:
+        return False, f"Memory '{memory_id}' not found."
+
+    if memory.memory_type == MemoryType.RAW.value:
+        return True, f"Memory '{memory_id}' don't need to extract."
+
+    tenant_id = memory.tenant_id
+    extracted_content = await extract_by_llm(
+        tenant_id,
+        memory.llm_id,
+        {"temperature": memory.temperature},
+        get_memory_type_human(memory.memory_type),
+        message_dict.get("user_input", ""),
+        message_dict.get("agent_response", "")
+    )
+    message_list = [{
+        "message_id": REDIS_CONN.generate_auto_increment_id(namespace="memory"),
+        "message_type": content["message_type"],
+        "source_id": source_message_id,
+        "memory_id": memory_id,
+        "user_id": "",
+        "agent_id": message_dict["agent_id"],
+        "session_id": message_dict["session_id"],
+        "content": content["content"],
+        "valid_at": content["valid_at"],
+        "invalid_at": content["invalid_at"] if content["invalid_at"] else None,
+        "forget_at": None,
+        "status": True
+    } for content in extracted_content]
+    if not message_list:
+        return True, "No memory extracted from raw message."
+
+    return await embed_and_save(memory, message_list)
+
+
+async def extract_by_llm(tenant_id: str, llm_id: str, extract_conf: dict, memory_type: List[str], user_input: str,
+                         agent_response: str, system_prompt: str = "", user_prompt: str="") -> List[dict]:
+    llm_type = TenantLLMService.llm_id2llm_type(llm_id)
+    if not llm_type:
+        raise RuntimeError(f"Unknown type of LLM '{llm_id}'")
+    if not system_prompt:
+        system_prompt = PromptAssembler.assemble_system_prompt({"memory_type": memory_type})
+    conversation_content = f"User Input: {user_input}\nAgent Response: {agent_response}"
+    conversation_time = timestamp_to_date(current_timestamp())
+    user_prompts = []
+    if user_prompt:
+        user_prompts.append({"role": "user", "content": user_prompt})
+        user_prompts.append({"role": "user", "content": f"Conversation: {conversation_content}\nConversation Time: {conversation_time}\nCurrent Time: {conversation_time}"})
+    else:
+        user_prompts.append({"role": "user", "content": PromptAssembler.assemble_user_prompt(conversation_content, conversation_time, conversation_time)})
+    llm = LLMBundle(tenant_id, llm_type, llm_id)
+    res = await llm.async_chat(system_prompt, user_prompts, extract_conf)
+    res_json = get_json_result_from_llm_response(res)
+    return [{
+        "content": extracted_content["content"],
+        "valid_at": format_iso_8601_to_ymd_hms(extracted_content["valid_at"]),
+        "invalid_at": format_iso_8601_to_ymd_hms(extracted_content["invalid_at"]) if extracted_content.get("invalid_at") else "",
+        "message_type": message_type
+    } for message_type, extracted_content_list in res_json.items() for extracted_content in extracted_content_list]
+
+
+async def embed_and_save(memory, message_list: list[dict]):
+    embedding_model = LLMBundle(memory.tenant_id, llm_type=LLMType.EMBEDDING, llm_name=memory.embd_id)
+    vector_list, _ = embedding_model.encode([msg["content"] for msg in message_list])
+    for idx, msg in enumerate(message_list):
+        msg["content_embed"] = vector_list[idx]
+    vector_dimension = len(vector_list[0])
+    if not MessageService.has_index(memory.tenant_id, memory.id):
+        created = MessageService.create_index(memory.tenant_id, memory.id, vector_size=vector_dimension)
+        if not created:
+            return False, "Failed to create message index."
+
+    new_msg_size = sum([MessageService.calculate_message_size(m) for m in message_list])
+    current_memory_size = get_memory_size_cache(memory.tenant_id, memory.id)
+    if new_msg_size + current_memory_size > memory.memory_size:
+        size_to_delete = current_memory_size + new_msg_size - memory.memory_size
+        if memory.forgetting_policy == "FIFO":
+            message_ids_to_delete, delete_size = MessageService.pick_messages_to_delete_by_fifo(memory.id, memory.tenant_id,
+                                                                                                size_to_delete)
+            MessageService.delete_message({"message_id": message_ids_to_delete}, memory.tenant_id, memory.id)
+            decrease_memory_size_cache(memory.id, delete_size)
+        else:
+            return False, "Failed to insert message into memory. Memory size reached limit and cannot decide which to delete."
+    fail_cases = MessageService.insert_message(message_list, memory.tenant_id, memory.id)
+    if fail_cases:
+        return False, "Failed to insert message into memory. Details: " + "; ".join(fail_cases)
+
+    increase_memory_size_cache(memory.id, new_msg_size)
+    return True, "Message saved successfully."
+
+
+def query_message(filter_dict: dict, params: dict):
+    """
+    :param filter_dict: {
+        "memory_id": List[str],
+        "agent_id": optional
+        "session_id": optional
+    }
+    :param params: {
+        "query": question str,
+        "similarity_threshold": float,
+        "keywords_similarity_weight": float,
+        "top_n": int
+    }
+    """
+    memory_ids = filter_dict["memory_id"]
+    memory_list = MemoryService.get_by_ids(memory_ids)
+    if not memory_list:
+        return []
+
+    condition_dict = {k: v for k, v in filter_dict.items() if v}
+    uids = [memory.tenant_id for memory in memory_list]
+
+    question = params["query"]
+    question = question.strip()
+    memory = memory_list[0]
+    embd_model = LLMBundle(memory.tenant_id, llm_type=LLMType.EMBEDDING, llm_name=memory.embd_id)
+    match_dense = get_vector(question, embd_model, similarity=params["similarity_threshold"])
+    match_text, _ = MsgTextQuery().question(question, min_match=params["similarity_threshold"])
+    keywords_similarity_weight = params.get("keywords_similarity_weight", 0.7)
+    fusion_expr = FusionExpr("weighted_sum", params["top_n"], {"weights": ",".join([str(1 - keywords_similarity_weight), str(keywords_similarity_weight)])})
+
+    return MessageService.search_message(memory_ids, condition_dict, uids, [match_text, match_dense, fusion_expr], params["top_n"])
+
+
+def init_message_id_sequence():
+    message_id_redis_key = "id_generator:memory"
+    if REDIS_CONN.exist(message_id_redis_key):
+        current_max_id = REDIS_CONN.get(message_id_redis_key)
+        logging.info(f"No need to init message_id sequence, current max id is {current_max_id}.")
+    else:
+        max_id = 1
+        exist_memory_list = MemoryService.get_all_memory()
+        if not exist_memory_list:
+            REDIS_CONN.set(message_id_redis_key, max_id)
+        else:
+            max_id = MessageService.get_max_message_id(
+                uid_list=[m.tenant_id for m in exist_memory_list],
+                memory_ids=[m.id for m in exist_memory_list]
+            )
+            REDIS_CONN.set(message_id_redis_key, max_id)
+        logging.info(f"Init message_id sequence done, current max id is {max_id}.")
+
+
+def get_memory_size_cache(memory_id: str, uid: str):
+    redis_key = f"memory_{memory_id}"
+    if REDIS_CONN.exist(redis_key):
+        return int(REDIS_CONN.get(redis_key))
+    else:
+        memory_size_map = MessageService.calculate_memory_size(
+            [memory_id],
+            [uid]
+        )
+        memory_size = memory_size_map.get(memory_id, 0)
+        set_memory_size_cache(memory_id, memory_size)
+        return memory_size
+
+
+def set_memory_size_cache(memory_id: str, size: int):
+    redis_key = f"memory_{memory_id}"
+    return REDIS_CONN.set(redis_key, size)
+
+
+def increase_memory_size_cache(memory_id: str, size: int):
+    redis_key = f"memory_{memory_id}"
+    return REDIS_CONN.incrby(redis_key, size)
+
+
+def decrease_memory_size_cache(memory_id: str, size: int):
+    redis_key = f"memory_{memory_id}"
+    return REDIS_CONN.decrby(redis_key, size)
+
+
+def init_memory_size_cache():
+    memory_list = MemoryService.get_all_memory()
+    if not memory_list:
+        logging.info("No memory found, no need to init memory size.")
+    else:
+        for m in memory_list:
+            get_memory_size_cache(m.id, m.tenant_id)
+        logging.info("Memory size cache init done.")
+
+
+def judge_system_prompt_is_default(system_prompt: str, memory_type: int|list[str]):
+    memory_type_list = memory_type if isinstance(memory_type, list) else get_memory_type_human(memory_type)
+    return system_prompt == PromptAssembler.assemble_system_prompt({"memory_type": memory_type_list})
+
+
+async def queue_save_to_memory_task(memory_ids: list[str], message_dict: dict):
+    """
+    :param memory_ids:
+    :param message_dict: {
+        "user_id": str,
+        "agent_id": str,
+        "session_id": str,
+        "user_input": str,
+        "agent_response": str
+    }
+    """
+    def new_task(_memory_id: str, _source_id: int):
+        return {
+            "id": get_uuid(),
+            "doc_id": _memory_id,
+            "task_type": "memory",
+            "progress": 0.0,
+            "digest": str(_source_id)
+        }
+
+    not_found_memory = []
+    failed_memory = []
+    for memory_id in memory_ids:
+        memory = MemoryService.get_by_memory_id(memory_id)
+        if not memory:
+            not_found_memory.append(memory_id)
+            continue
+
+        raw_message_id = REDIS_CONN.generate_auto_increment_id(namespace="memory")
+        raw_message = {
+            "message_id": raw_message_id,
+            "message_type": MemoryType.RAW.name.lower(),
+            "source_id": 0,
+            "memory_id": memory_id,
+            "user_id": "",
+            "agent_id": message_dict["agent_id"],
+            "session_id": message_dict["session_id"],
+            "content": f"User Input: {message_dict.get('user_input')}\nAgent Response: {message_dict.get('agent_response')}",
+            "valid_at": timestamp_to_date(current_timestamp()),
+            "invalid_at": None,
+            "forget_at": None,
+            "status": True
+        }
+        res, msg = await embed_and_save(memory, [raw_message])
+        if not res:
+            failed_memory.append({"memory_id": memory_id, "fail_msg": msg})
+            continue
+
+        task = new_task(memory_id, raw_message_id)
+        bulk_insert_into_db(Task, [task], replace_on_conflict=True)
+        task_message = {
+            "id": task["id"],
+            "task_id": task["id"],
+            "task_type": task["task_type"],
+            "memory_id": memory_id,
+            "source_id": raw_message_id,
+            "message_dict": message_dict
+        }
+        if not REDIS_CONN.queue_product(settings.get_svr_queue_name(priority=0), message=task_message):
+            failed_memory.append({"memory_id": memory_id, "fail_msg": "Can't access Redis."})
+
+    error_msg = ""
+    if not_found_memory:
+        error_msg = f"Memory {not_found_memory} not found."
+    if failed_memory:
+        error_msg += "".join([f"Memory {fm['memory_id']} failed. Detail: {fm['fail_msg']}" for fm in failed_memory])
+
+    if error_msg:
+        return False, error_msg
+
+    return True, "All add to task."
+
+
+async def handle_save_to_memory_task(task_param: dict):
+    """
+    :param task_param: {
+        "id": task_id
+        "memory_id": id
+        "source_id": id
+        "message_dict": {
+            "user_id": str,
+            "agent_id": str,
+            "session_id": str,
+            "user_input": str,
+            "agent_response": str
+        }
+    }
+    """
+    _, task = TaskService.get_by_id(task_param["id"])
+    if not task:
+        return False, f"Task {task_param['id']} is not found."
+    if task.progress == -1:
+        return False, f"Task {task_param['id']} is already failed."
+    now_time = current_timestamp()
+    TaskService.update_by_id(task_param["id"], {"begin_at": timestamp_to_date(now_time)})
+
+    memory_id = task_param["memory_id"]
+    source_id = task_param["source_id"]
+    message_dict = task_param["message_dict"]
+    success, msg = await save_extracted_to_memory_only(memory_id, message_dict, source_id)
+    if success:
+        TaskService.update_progress(task.id, {"progress": 1.0,  "progress_msg": msg})
+        return True, msg
+
+    logging.error(msg)
+    TaskService.update_progress(task.id, {"progress": -1, "progress_msg": None})
+    return False, msg

api/db/joint_services/user_account_service.py

@@ -34,6 +34,8 @@ from api.db.services.task_service import TaskService
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.user_canvas_version import UserCanvasVersionService
 from api.db.services.user_service import TenantService, UserService, UserTenantService
+from api.db.services.memory_service import MemoryService
+from memory.services.messages import MessageService
 from rag.nlp import search
 from common.constants import ActiveEnum
 from common import settings
@@ -200,7 +202,16 @@ def delete_user_data(user_id: str) -> dict:
             done_msg += f"- Deleted {llm_delete_res} tenant-LLM records.\n"
             langfuse_delete_res = TenantLangfuseService.delete_ty_tenant_id(tenant_id)
             done_msg += f"- Deleted {langfuse_delete_res} langfuse records.\n"
-            # step1.3 delete own tenant
+            # step1.3 delete memory and messages
+            user_memory = MemoryService.get_by_tenant_id(tenant_id)
+            if user_memory:
+                for memory in user_memory:
+                    if MessageService.has_index(tenant_id, memory.id):
+                        MessageService.delete_index(tenant_id, memory.id)
+                done_msg += " Deleted memory index."
+                memory_delete_res = MemoryService.delete_by_ids([m.id for m in user_memory])
+                done_msg += f"Deleted {memory_delete_res} memory datasets."
+            # step1.4 delete own tenant
             tenant_delete_res = TenantService.delete_by_id(tenant_id)
             done_msg += f"- Deleted {tenant_delete_res} tenant.\n"
         # step2 delete user-tenant relation

api/db/services/canvas_service.py

@@ -123,6 +123,19 @@ class UserCanvasService(CommonService):
             logging.exception(e)
             return False, None
 
+    @classmethod
+    @DB.connection_context()
+    def get_basic_info_by_canvas_ids(cls, canvas_id):
+        fields = [
+            cls.model.id,
+            cls.model.avatar,
+            cls.model.user_id,
+            cls.model.title,
+            cls.model.permission,
+            cls.model.canvas_category
+        ]
+        return cls.model.select(*fields).where(cls.model.id.in_(canvas_id)).dicts()
+
     @classmethod
     @DB.connection_context()
     def get_by_tenant_ids(cls, joined_tenant_ids, user_id,
@@ -198,7 +211,7 @@ async def completion(tenant_id, agent_id, session_id=None, **kwargs):
         if not isinstance(cvs.dsl, str):
             cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
         session_id=get_uuid()
-        canvas = Canvas(cvs.dsl, tenant_id, agent_id)
+        canvas = Canvas(cvs.dsl, tenant_id, agent_id, canvas_id=cvs.id)
         canvas.reset()
         conv = {
             "id": session_id,

api/db/services/common_service.py

@@ -169,10 +169,12 @@ class CommonService:
         """
         if "id" not in kwargs:
             kwargs["id"] = get_uuid()
-        kwargs["create_time"] = current_timestamp()
-        kwargs["create_date"] = datetime_format(datetime.now())
-        kwargs["update_time"] = current_timestamp()
-        kwargs["update_date"] = datetime_format(datetime.now())
+        timestamp = current_timestamp()
+        cur_datetime = datetime_format(datetime.now())
+        kwargs["create_time"] = timestamp
+        kwargs["create_date"] = cur_datetime
+        kwargs["update_time"] = timestamp
+        kwargs["update_date"] = cur_datetime
         sample_obj = cls.model(**kwargs).save(force_insert=True)
         return sample_obj
 
@@ -207,10 +209,14 @@ class CommonService:
             data_list (list): List of dictionaries containing record data to update.
                              Each dictionary must include an 'id' field.
         """
+
+        timestamp = current_timestamp()
+        cur_datetime = datetime_format(datetime.now())
+        for data in data_list:
+            data["update_time"] = timestamp
+            data["update_date"] = cur_datetime
         with DB.atomic():
             for data in data_list:
-                data["update_time"] = current_timestamp()
-                data["update_date"] = datetime_format(datetime.now())
                 cls.model.update(data).where(cls.model.id == data["id"]).execute()
 
     @classmethod

api/db/services/conversation_service.py

@@ -116,6 +116,16 @@ async def async_completion(tenant_id, chat_id, question, name="New session", ses
                                     ensure_ascii=False) + "\n\n"
             yield "data:" + json.dumps({"code": 0, "message": "", "data": True}, ensure_ascii=False) + "\n\n"
             return
+        else:
+            answer = {
+                "answer": conv["message"][0]["content"],
+                "reference": {},
+                "audio_binary": None,
+                "id": None,
+                "session_id": session_id
+            }
+            yield answer
+            return
 
     conv = ConversationService.query(id=session_id, dialog_id=chat_id)
     if not conv:

api/db/services/dialog_service.py

@@ -406,7 +406,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
                     dialog.vector_similarity_weight,
                     doc_ids=attachments,
                     top=dialog.top_k,
-                    aggs=False,
+                    aggs=True,
                     rerank_mdl=rerank_mdl,
                     rank_feature=label_question(" ".join(questions), kbs),
                 )
@@ -421,7 +421,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
                 kbinfos["chunks"].extend(tav_res["chunks"])
                 kbinfos["doc_aggs"].extend(tav_res["doc_aggs"])
             if prompt_config.get("use_kg"):
-                ck = settings.kg_retriever.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl,
+                ck = await settings.kg_retriever.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl,
                                                        LLMBundle(dialog.tenant_id, LLMType.CHAT))
                 if ck["content_with_weight"]:
                     kbinfos["chunks"].insert(0, ck)
@@ -769,7 +769,7 @@ async def async_ask(question, kb_ids, tenant_id, chat_llm_name=None, search_conf
         vector_similarity_weight=search_config.get("vector_similarity_weight", 0.3),
         top=search_config.get("top_k", 1024),
         doc_ids=doc_ids,
-        aggs=False,
+        aggs=True,
         rerank_mdl=rerank_mdl,
         rank_feature=label_question(question, kbs)
     )

api/db/services/document_service.py

@@ -33,12 +33,13 @@ from api.db.db_models import DB, Document, Knowledgebase, Task, Tenant, UserTena
 from api.db.db_utils import bulk_insert_into_db
 from api.db.services.common_service import CommonService
 from api.db.services.knowledgebase_service import KnowledgebaseService
+from common.metadata_utils import dedupe_list
 from common.misc_utils import get_uuid
 from common.time_utils import current_timestamp, get_format_time
 from common.constants import LLMType, ParserType, StatusEnum, TaskStatus, SVR_CONSUMER_GROUP_NAME
 from rag.nlp import rag_tokenizer, search
 from rag.utils.redis_conn import REDIS_CONN
-from rag.utils.doc_store_conn import OrderByExpr
+from common.doc_store.doc_store_base import OrderByExpr
 from common import settings
 
 
@@ -124,26 +125,26 @@ class DocumentService(CommonService):
 
     @classmethod
     @DB.connection_context()
-    def get_by_kb_id(cls, kb_id, page_number, items_per_page,
-                     orderby, desc, keywords, run_status, types, suffix, doc_ids=None):
+    def get_by_kb_id(cls, kb_id, page_number, items_per_page, orderby, desc, keywords, run_status, types, suffix, doc_ids=None, return_empty_metadata=False):
         fields = cls.get_cls_model_fields()
         if keywords:
-            docs = cls.model.select(*[*fields, UserCanvas.title.alias("pipeline_name"), User.nickname])\
-                .join(File2Document, on=(File2Document.document_id == cls.model.id))\
-                .join(File, on=(File.id == File2Document.file_id))\
-                .join(UserCanvas, on=(cls.model.pipeline_id == UserCanvas.id), join_type=JOIN.LEFT_OUTER)\
-                .join(User, on=(cls.model.created_by == User.id), join_type=JOIN.LEFT_OUTER)\
-                .where(
-                    (cls.model.kb_id == kb_id),
-                    (fn.LOWER(cls.model.name).contains(keywords.lower()))
-                )
+            docs = (
+                cls.model.select(*[*fields, UserCanvas.title.alias("pipeline_name"), User.nickname])
+                .join(File2Document, on=(File2Document.document_id == cls.model.id))
+                .join(File, on=(File.id == File2Document.file_id))
+                .join(UserCanvas, on=(cls.model.pipeline_id == UserCanvas.id), join_type=JOIN.LEFT_OUTER)
+                .join(User, on=(cls.model.created_by == User.id), join_type=JOIN.LEFT_OUTER)
+                .where((cls.model.kb_id == kb_id), (fn.LOWER(cls.model.name).contains(keywords.lower())))
+            )
         else:
-            docs = cls.model.select(*[*fields, UserCanvas.title.alias("pipeline_name"), User.nickname])\
-                .join(File2Document, on=(File2Document.document_id == cls.model.id))\
-                .join(UserCanvas, on=(cls.model.pipeline_id == UserCanvas.id), join_type=JOIN.LEFT_OUTER)\
-                .join(File, on=(File.id == File2Document.file_id))\
-                .join(User, on=(cls.model.created_by == User.id), join_type=JOIN.LEFT_OUTER)\
+            docs = (
+                cls.model.select(*[*fields, UserCanvas.title.alias("pipeline_name"), User.nickname])
+                .join(File2Document, on=(File2Document.document_id == cls.model.id))
+                .join(UserCanvas, on=(cls.model.pipeline_id == UserCanvas.id), join_type=JOIN.LEFT_OUTER)
+                .join(File, on=(File.id == File2Document.file_id))
+                .join(User, on=(cls.model.created_by == User.id), join_type=JOIN.LEFT_OUTER)
                 .where(cls.model.kb_id == kb_id)
+            )
 
         if doc_ids:
             docs = docs.where(cls.model.id.in_(doc_ids))
@@ -153,6 +154,8 @@ class DocumentService(CommonService):
             docs = docs.where(cls.model.type.in_(types))
         if suffix:
             docs = docs.where(cls.model.suffix.in_(suffix))
+        if return_empty_metadata:
+            docs = docs.where(fn.COALESCE(fn.JSON_LENGTH(cls.model.meta_fields), 0) == 0)
 
         count = docs.count()
         if desc:
@@ -160,7 +163,6 @@ class DocumentService(CommonService):
         else:
             docs = docs.order_by(cls.model.getter_by(orderby).asc())
 
-
         if page_number and items_per_page:
             docs = docs.paginate(page_number, items_per_page)
 
@@ -180,6 +182,16 @@ class DocumentService(CommonService):
              "1": 2,
              "2": 2
             }
+            "metadata": {
+                "key1": {
+                 "key1_value1": 1,
+                 "key1_value2": 2,
+                },
+                "key2": {
+                 "key2_value1": 2,
+                 "key2_value2": 1,
+                },
+            }
         }, total
         where "1" => RUNNING, "2" => CANCEL
         """
@@ -200,19 +212,42 @@ class DocumentService(CommonService):
         if suffix:
             query = query.where(cls.model.suffix.in_(suffix))
 
-        rows = query.select(cls.model.run, cls.model.suffix)
+        rows = query.select(cls.model.run, cls.model.suffix, cls.model.meta_fields)
         total = rows.count()
 
         suffix_counter = {}
         run_status_counter = {}
+        metadata_counter = {}
+        empty_metadata_count = 0
 
         for row in rows:
             suffix_counter[row.suffix] = suffix_counter.get(row.suffix, 0) + 1
             run_status_counter[str(row.run)] = run_status_counter.get(str(row.run), 0) + 1
+            meta_fields = row.meta_fields or {}
+            if not meta_fields:
+                empty_metadata_count += 1
+                continue
+            has_valid_meta = False
+            for key, value in meta_fields.items():
+                values = value if isinstance(value, list) else [value]
+                for vv in values:
+                    if vv is None:
+                        continue
+                    if isinstance(vv, str) and not vv.strip():
+                        continue
+                    sv = str(vv)
+                    if key not in metadata_counter:
+                        metadata_counter[key] = {}
+                    metadata_counter[key][sv] = metadata_counter[key].get(sv, 0) + 1
+                    has_valid_meta = True
+            if not has_valid_meta:
+                empty_metadata_count += 1
 
+        metadata_counter["empty_metadata"] = {"true": empty_metadata_count}
         return {
             "suffix": suffix_counter,
-            "run_status": run_status_counter
+            "run_status": run_status_counter,
+            "metadata": metadata_counter,
         }, total
 
     @classmethod
@@ -307,21 +342,7 @@ class DocumentService(CommonService):
         cls.clear_chunk_num(doc.id)
         try:
             TaskService.filter_delete([Task.doc_id == doc.id])
-            page = 0
-            page_size = 1000
-            all_chunk_ids = []
-            while True:
-                chunks = settings.docStoreConn.search(["img_id"], [], {"doc_id": doc.id}, [], OrderByExpr(),
-                                                      page * page_size, page_size, search.index_name(tenant_id),
-                                                      [doc.kb_id])
-                chunk_ids = settings.docStoreConn.get_chunk_ids(chunks)
-                if not chunk_ids:
-                    break
-                all_chunk_ids.extend(chunk_ids)
-                page += 1
-            for cid in all_chunk_ids:
-                if settings.STORAGE_IMPL.obj_exist(doc.kb_id, cid):
-                    settings.STORAGE_IMPL.rm(doc.kb_id, cid)
+            cls.delete_chunk_images(doc, tenant_id)
             if doc.thumbnail and not doc.thumbnail.startswith(IMG_BASE64_PREFIX):
                 if settings.STORAGE_IMPL.obj_exist(doc.kb_id, doc.thumbnail):
                     settings.STORAGE_IMPL.rm(doc.kb_id, doc.thumbnail)
@@ -343,6 +364,23 @@ class DocumentService(CommonService):
             pass
         return cls.delete_by_id(doc.id)
 
+    @classmethod
+    @DB.connection_context()
+    def delete_chunk_images(cls, doc, tenant_id):
+        page = 0
+        page_size = 1000
+        while True:
+            chunks = settings.docStoreConn.search(["img_id"], [], {"doc_id": doc.id}, [], OrderByExpr(),
+                                                  page * page_size, page_size, search.index_name(tenant_id),
+                                                  [doc.kb_id])
+            chunk_ids = settings.docStoreConn.get_doc_ids(chunks)
+            if not chunk_ids:
+                break
+            for cid in chunk_ids:
+                if settings.STORAGE_IMPL.obj_exist(doc.kb_id, cid):
+                    settings.STORAGE_IMPL.rm(doc.kb_id, cid)
+            page += 1
+
     @classmethod
     @DB.connection_context()
     def get_newly_uploaded(cls):
@@ -665,10 +703,14 @@ class DocumentService(CommonService):
             for k,v in r.meta_fields.items():
                 if k not in meta:
                     meta[k] = {}
-                v = str(v)
-                if v not in meta[k]:
-                    meta[k][v] = []
-                meta[k][v].append(doc_id)
+                if not isinstance(v, list):
+                    v = [v]
+                for vv in v:
+                    if vv not in meta[k]:
+                        if isinstance(vv, list) or isinstance(vv, dict):
+                            continue
+                        meta[k][vv] = []
+                    meta[k][vv].append(doc_id)
         return meta
 
     @classmethod
@@ -766,7 +808,10 @@ class DocumentService(CommonService):
                 match_provided = "match" in upd
                 if isinstance(meta[key], list):
                     if not match_provided:
-                        meta[key] = new_value
+                        if isinstance(new_value, list):
+                            meta[key] = dedupe_list(new_value)
+                        else:
+                            meta[key] = new_value
                         changed = True
                     else:
                         match_value = upd.get("match")
@@ -779,7 +824,7 @@ class DocumentService(CommonService):
                             else:
                                 new_list.append(item)
                         if replaced:
-                            meta[key] = new_list
+                            meta[key] = dedupe_list(new_list)
                             changed = True
                 else:
                     if not match_provided:
@@ -1199,8 +1244,8 @@ def doc_upload_and_parse(conversation_id, file_objs, user_id):
             d["q_%d_vec" % len(v)] = v
         for b in range(0, len(cks), es_bulk_size):
             if try_create_idx:
-                if not settings.docStoreConn.indexExist(idxnm, kb_id):
-                    settings.docStoreConn.createIdx(idxnm, kb_id, len(vectors[0]))
+                if not settings.docStoreConn.index_exist(idxnm, kb_id):
+                    settings.docStoreConn.create_idx(idxnm, kb_id, len(vectors[0]))
                 try_create_idx = False
             settings.docStoreConn.insert(cks[b:b + es_bulk_size], idxnm, kb_id)
 

api/db/services/evaluation_service.py

@@ -65,6 +65,7 @@ class EvaluationService(CommonService):
             (success, dataset_id or error_message)
         """
         try:
+            timestamp= current_timestamp()
             dataset_id = get_uuid()
             dataset = {
                 "id": dataset_id,
@@ -73,8 +74,8 @@ class EvaluationService(CommonService):
                 "description": description,
                 "kb_ids": kb_ids,
                 "created_by": user_id,
-                "create_time": current_timestamp(),
-                "update_time": current_timestamp(),
+                "create_time": timestamp,
+                "update_time": timestamp,
                 "status": StatusEnum.VALID.value
             }
 

api/db/services/file_service.py

@@ -100,7 +100,7 @@ class FileService(CommonService):
         # Returns:
         #     List of dictionaries containing dataset IDs and names
         kbs = (
-            cls.model.select(*[Knowledgebase.id, Knowledgebase.name])
+            cls.model.select(*[Knowledgebase.id, Knowledgebase.name, File2Document.document_id])
             .join(File2Document, on=(File2Document.file_id == file_id))
             .join(Document, on=(File2Document.document_id == Document.id))
             .join(Knowledgebase, on=(Knowledgebase.id == Document.kb_id))
@@ -110,7 +110,7 @@ class FileService(CommonService):
             return []
         kbs_info_list = []
         for kb in list(kbs.dicts()):
-            kbs_info_list.append({"kb_id": kb["id"], "kb_name": kb["name"]})
+            kbs_info_list.append({"kb_id": kb["id"], "kb_name": kb["name"], "document_id": kb["document_id"]})
         return kbs_info_list
 
     @classmethod

api/db/services/knowledgebase_service.py

@@ -425,6 +425,7 @@ class KnowledgebaseService(CommonService):
 
         # Update parser_config (always override with validated default/merged config)
         payload["parser_config"] = get_parser_config(parser_id, kwargs.get("parser_config"))
+        payload["parser_config"]["llm_id"] = _t.llm_id
 
         return True, payload
 

api/db/services/langfuse_service.py

@@ -64,10 +64,13 @@ class TenantLangfuseService(CommonService):
 
     @classmethod
     def save(cls, **kwargs):
-        kwargs["create_time"] = current_timestamp()
-        kwargs["create_date"] = datetime_format(datetime.now())
-        kwargs["update_time"] = current_timestamp()
-        kwargs["update_date"] = datetime_format(datetime.now())
+        current_ts = current_timestamp()
+        current_date = datetime_format(datetime.now())
+
+        kwargs["create_time"] = current_ts
+        kwargs["create_date"] = current_date
+        kwargs["update_time"] = current_ts
+        kwargs["update_date"] = current_date
         obj = cls.model.create(**kwargs)
         return obj
 

api/db/services/memory_service.py

@@ -15,7 +15,6 @@
 #
 from typing import List
 
-from api.apps import current_user
 from api.db.db_models import DB, Memory, User
 from api.db.services import duplicate_name
 from api.db.services.common_service import CommonService
@@ -23,6 +22,7 @@ from api.utils.memory_utils import calculate_memory_type
 from api.constants import MEMORY_NAME_LIMIT
 from common.misc_utils import get_uuid
 from common.time_utils import get_format_time, current_timestamp
+from memory.utils.prompt_util import PromptAssembler
 
 
 class MemoryService(CommonService):
@@ -34,6 +34,17 @@ class MemoryService(CommonService):
     def get_by_memory_id(cls, memory_id: str):
         return cls.model.select().where(cls.model.id == memory_id).first()
 
+    @classmethod
+    @DB.connection_context()
+    def get_by_tenant_id(cls, tenant_id: str):
+        return cls.model.select().where(cls.model.tenant_id == tenant_id)
+
+    @classmethod
+    @DB.connection_context()
+    def get_all_memory(cls):
+        memory_list = cls.model.select()
+        return list(memory_list)
+
     @classmethod
     @DB.connection_context()
     def get_with_owner_name_by_id(cls, memory_id: str):
@@ -53,7 +64,9 @@ class MemoryService(CommonService):
             cls.model.forgetting_policy,
             cls.model.temperature,
             cls.model.system_prompt,
-            cls.model.user_prompt
+            cls.model.user_prompt,
+            cls.model.create_date,
+            cls.model.create_time
         ]
         memory = cls.model.select(*fields).join(User, on=(cls.model.tenant_id == User.id)).where(
             cls.model.id == memory_id
@@ -72,7 +85,9 @@ class MemoryService(CommonService):
             cls.model.memory_type,
             cls.model.storage_type,
             cls.model.permissions,
-            cls.model.description
+            cls.model.description,
+            cls.model.create_time,
+            cls.model.create_date
         ]
         memories = cls.model.select(*fields).join(User, on=(cls.model.tenant_id == User.id))
         if filter_dict.get("tenant_id"):
@@ -102,6 +117,8 @@ class MemoryService(CommonService):
         if len(memory_name) > MEMORY_NAME_LIMIT:
             return False, f"Memory name {memory_name} exceeds limit of {MEMORY_NAME_LIMIT}."
 
+        timestamp = current_timestamp()
+        format_time = get_format_time()
         # build create dict
         memory_info = {
             "id": get_uuid(),
@@ -110,10 +127,11 @@ class MemoryService(CommonService):
             "tenant_id": tenant_id,
             "embd_id": embd_id,
             "llm_id": llm_id,
-            "create_time": current_timestamp(),
-            "create_date": get_format_time(),
-            "update_time": current_timestamp(),
-            "update_date": get_format_time(),
+            "system_prompt": PromptAssembler.assemble_system_prompt({"memory_type": memory_type}),
+            "create_time": timestamp,
+            "create_date": format_time,
+            "update_time": timestamp,
+            "update_date": format_time,
         }
         obj = cls.model(**memory_info).save(force_insert=True)
 
@@ -126,16 +144,18 @@ class MemoryService(CommonService):
 
     @classmethod
     @DB.connection_context()
-    def update_memory(cls, memory_id: str, update_dict: dict):
+    def update_memory(cls, tenant_id: str, memory_id: str, update_dict: dict):
         if not update_dict:
             return 0
         if "temperature" in update_dict and isinstance(update_dict["temperature"], str):
             update_dict["temperature"] = float(update_dict["temperature"])
+        if "memory_type" in update_dict and isinstance(update_dict["memory_type"], list):
+            update_dict["memory_type"] = calculate_memory_type(update_dict["memory_type"])
         if "name" in update_dict:
             update_dict["name"] = duplicate_name(
                 cls.query,
                 name=update_dict["name"],
-                tenant_id=current_user.id
+                tenant_id=tenant_id
             )
         update_dict.update({
             "update_time": current_timestamp(),

api/db/services/pipeline_operation_log_service.py

@@ -169,11 +169,12 @@ class PipelineOperationLogService(CommonService):
             operation_status=operation_status,
             avatar=avatar,
         )
-        log["create_time"] = current_timestamp()
-        log["create_date"] = datetime_format(datetime.now())
-        log["update_time"] = current_timestamp()
-        log["update_date"] = datetime_format(datetime.now())
-
+        timestamp = current_timestamp()
+        datetime_now = datetime_format(datetime.now())
+        log["create_time"] = timestamp
+        log["create_date"] = datetime_now
+        log["update_time"] = timestamp
+        log["update_date"] = datetime_now
         with DB.atomic():
             obj = cls.save(**log)
 

api/db/services/search_service.py

@@ -28,10 +28,13 @@ class SearchService(CommonService):
 
     @classmethod
     def save(cls, **kwargs):
-        kwargs["create_time"] = current_timestamp()
-        kwargs["create_date"] = datetime_format(datetime.now())
-        kwargs["update_time"] = current_timestamp()
-        kwargs["update_date"] = datetime_format(datetime.now())
+        current_ts = current_timestamp()
+        current_date = datetime_format(datetime.now())
+
+        kwargs["create_time"] = current_ts
+        kwargs["create_date"] = current_date
+        kwargs["update_time"] = current_ts
+        kwargs["update_date"] = current_date
         obj = cls.model.create(**kwargs)
         return obj
 

api/db/services/tenant_llm_service.py

@@ -97,7 +97,7 @@ class TenantLLMService(CommonService):
         if llm_type == LLMType.EMBEDDING.value:
             mdlnm = tenant.embd_id if not llm_name else llm_name
         elif llm_type == LLMType.SPEECH2TEXT.value:
-            mdlnm = tenant.asr_id
+            mdlnm = tenant.asr_id if not llm_name else llm_name
         elif llm_type == LLMType.IMAGE2TEXT.value:
             mdlnm = tenant.img2txt_id if not llm_name else llm_name
         elif llm_type == LLMType.CHAT.value:

api/db/services/user_service.py

@@ -116,10 +116,13 @@ class UserService(CommonService):
             kwargs["password"] = generate_password_hash(
                 str(kwargs["password"]))
 
-        kwargs["create_time"] = current_timestamp()
-        kwargs["create_date"] = datetime_format(datetime.now())
-        kwargs["update_time"] = current_timestamp()
-        kwargs["update_date"] = datetime_format(datetime.now())
+        current_ts = current_timestamp()
+        current_date = datetime_format(datetime.now())
+
+        kwargs["create_time"] = current_ts
+        kwargs["create_date"] = current_date
+        kwargs["update_time"] = current_ts
+        kwargs["update_date"] = current_date
         obj = cls.model(**kwargs).save(force_insert=True)
         return obj
 
@@ -161,7 +164,7 @@ class UserService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_all_users(cls):
-        users = cls.model.select()
+        users = cls.model.select().order_by(cls.model.email)
         return list(users)
 
 

api/utils/api_utils.py

@@ -163,6 +163,7 @@ def validate_request(*args, **kwargs):
             if error_arguments:
                 error_string += "required argument values: {}".format(",".join(["{}={}".format(a[0], a[1]) for a in error_arguments]))
             return error_string
+        return None
 
     def wrapper(func):
         @wraps(func)
@@ -409,7 +410,7 @@ def get_parser_config(chunk_method, parser_config):
     if default_config is None:
         return deep_merge(base_defaults, parser_config)
 
-    # Ensure raptor and graphrag fields have default values if not provided
+    # Ensure raptor and graph_rag fields have default values if not provided
     merged_config = deep_merge(base_defaults, default_config)
     merged_config = deep_merge(merged_config, parser_config)
 

api/utils/file_utils.py

@@ -42,7 +42,7 @@ def filename_type(filename):
     if re.match(r".*\.pdf$", filename):
         return FileType.PDF.value
 
-    if re.match(r".*\.(msg|eml|doc|docx|ppt|pptx|yml|xml|htm|json|jsonl|ldjson|csv|txt|ini|xls|xlsx|wps|rtf|hlp|pages|numbers|key|md|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|html|sql)$", filename):
+    if re.match(r".*\.(msg|eml|doc|docx|ppt|pptx|yml|xml|htm|json|jsonl|ldjson|csv|txt|ini|xls|xlsx|wps|rtf|hlp|pages|numbers|key|md|mdx|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|html|sql)$", filename):
         return FileType.DOC.value
 
     if re.match(r".*\.(wav|flac|ape|alac|wavpack|wv|mp3|aac|ogg|vorbis|opus)$", filename):

api/utils/web_utils.py

@@ -69,6 +69,7 @@ CONTENT_TYPE_MAP = {
     # Web
     "md": "text/markdown",
     "markdown": "text/markdown",
+    "mdx": "text/markdown",
     "htm": "text/html",
     "html": "text/html",
     "json": "application/json",

common/constants.py

@@ -54,6 +54,7 @@ class RetCode(IntEnum, CustomEnum):
     SERVER_ERROR = 500
     FORBIDDEN = 403
     NOT_FOUND = 404
+    CONFLICT = 409
 
 
 class StatusEnum(Enum):
@@ -124,6 +125,16 @@ class FileSource(StrEnum):
     MOODLE = "moodle"
     DROPBOX = "dropbox"
     BOX = "box"
+    R2 = "r2"
+    OCI_STORAGE = "oci_storage"
+    GOOGLE_CLOUD_STORAGE = "google_cloud_storage"
+    AIRTABLE = "airtable"
+    ASANA = "asana"
+    GITHUB = "github"
+    GITLAB = "gitlab"
+    IMAP = "imap"
+    BITBUCKET = "bitbucket"
+    ZENDESK = "zendesk"
 
 class PipelineTaskType(StrEnum):
     PARSE = "Parse"
@@ -131,6 +142,7 @@ class PipelineTaskType(StrEnum):
     RAPTOR = "RAPTOR"
     GRAPH_RAG = "GraphRAG"
     MINDMAP = "Mindmap"
+    MEMORY = "Memory"
 
 
 VALID_PIPELINE_TASK_TYPES = {PipelineTaskType.PARSE, PipelineTaskType.DOWNLOAD, PipelineTaskType.RAPTOR,
@@ -165,7 +177,7 @@ class MemoryStorageType(StrEnum):
 
 
 class ForgettingPolicy(StrEnum):
-    FIFO = "fifo"
+    FIFO = "FIFO"
 
 
 # environment

common/data_source/__init__.py

@@ -34,8 +34,11 @@ from .google_drive.connector import GoogleDriveConnector
 from .jira.connector import JiraConnector
 from .sharepoint_connector import SharePointConnector
 from .teams_connector import TeamsConnector
-from .webdav_connector import WebDAVConnector
 from .moodle_connector import MoodleConnector
+from .airtable_connector import AirtableConnector
+from .asana_connector import AsanaConnector
+from .imap_connector import ImapConnector
+from .zendesk_connector import ZendeskConnector
 from .config import BlobType, DocumentSource
 from .models import Document, TextSection, ImageSection, BasicExpertInfo
 from .exceptions import (
@@ -58,7 +61,6 @@ __all__ = [
     "JiraConnector",
     "SharePointConnector",
     "TeamsConnector",
-    "WebDAVConnector",
     "MoodleConnector",
     "BlobType",
     "DocumentSource",
@@ -70,5 +72,9 @@ __all__ = [
     "ConnectorValidationError",
     "CredentialExpiredError",
     "InsufficientPermissionsError",
-    "UnexpectedValidationError"
+    "UnexpectedValidationError",
+    "AirtableConnector",
+    "AsanaConnector",
+    "ImapConnector",
+    "ZendeskConnector",
 ]

common/data_source/airtable_connector.py

@@ -0,0 +1,169 @@
+from datetime import datetime, timezone
+import logging
+from typing import Any, Generator
+
+import requests
+
+from pyairtable import Api as AirtableApi
+
+from common.data_source.config import AIRTABLE_CONNECTOR_SIZE_THRESHOLD, INDEX_BATCH_SIZE, DocumentSource
+from common.data_source.exceptions import ConnectorMissingCredentialError
+from common.data_source.interfaces import LoadConnector, PollConnector
+from common.data_source.models import Document, GenerateDocumentsOutput, SecondsSinceUnixEpoch
+from common.data_source.utils import extract_size_bytes, get_file_ext
+
+class AirtableClientNotSetUpError(PermissionError):
+    def __init__(self) -> None:
+        super().__init__(
+            "Airtable client is not set up. Did you forget to call load_credentials()?"
+        )
+
+
+class AirtableConnector(LoadConnector, PollConnector):
+    """
+    Lightweight Airtable connector.
+
+    This connector ingests Airtable attachments as raw blobs without
+    parsing file content or generating text/image sections.
+    """
+
+    def __init__(
+        self,
+        base_id: str,
+        table_name_or_id: str,
+        batch_size: int = INDEX_BATCH_SIZE,
+    ) -> None:
+        self.base_id = base_id
+        self.table_name_or_id = table_name_or_id
+        self.batch_size = batch_size
+        self._airtable_client: AirtableApi | None = None
+        self.size_threshold = AIRTABLE_CONNECTOR_SIZE_THRESHOLD
+
+    # -------------------------
+    # Credentials
+    # -------------------------
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        self._airtable_client = AirtableApi(credentials["airtable_access_token"])
+        return None
+
+    @property
+    def airtable_client(self) -> AirtableApi:
+        if not self._airtable_client:
+            raise AirtableClientNotSetUpError()
+        return self._airtable_client
+
+    # -------------------------
+    # Core logic
+    # -------------------------
+    def load_from_state(self) -> GenerateDocumentsOutput:
+        """
+        Fetch all Airtable records and ingest attachments as raw blobs.
+
+        Each attachment is converted into a single Document(blob=...).
+        """
+        if not self._airtable_client:
+            raise ConnectorMissingCredentialError("Airtable credentials not loaded")
+
+        table = self.airtable_client.table(self.base_id, self.table_name_or_id)
+        records = table.all()
+
+        logging.info(
+            f"Starting Airtable blob ingestion for table {self.table_name_or_id}, "
+            f"{len(records)} records found."
+        )
+
+        batch: list[Document] = []
+
+        for record in records:
+            print(record)
+            record_id = record.get("id")
+            fields = record.get("fields", {})
+            created_time = record.get("createdTime")
+
+            for field_value in fields.values():
+                # We only care about attachment fields (lists of dicts with url/filename)
+                if not isinstance(field_value, list):
+                    continue
+
+                for attachment in field_value:
+                    url = attachment.get("url")
+                    filename = attachment.get("filename")
+                    attachment_id = attachment.get("id")
+
+                    if not url or not filename or not attachment_id:
+                        continue
+
+                    try:
+                        resp = requests.get(url, timeout=30)
+                        resp.raise_for_status()
+                        content = resp.content
+                    except Exception:
+                        logging.exception(
+                            f"Failed to download attachment {filename} "
+                            f"(record={record_id})"
+                        )
+                        continue
+                    size_bytes = extract_size_bytes(attachment)
+                    if (
+                        self.size_threshold is not None
+                        and isinstance(size_bytes, int)
+                        and size_bytes > self.size_threshold
+                    ):
+                        logging.warning(
+                            f"{filename} exceeds size threshold of {self.size_threshold}. Skipping."
+                        )
+                        continue
+                    batch.append(
+                        Document(
+                            id=f"airtable:{record_id}:{attachment_id}",
+                            blob=content,
+                            source=DocumentSource.AIRTABLE,
+                            semantic_identifier=filename,
+                            extension=get_file_ext(filename),
+                            size_bytes=size_bytes if size_bytes else 0,
+                            doc_updated_at=datetime.strptime(created_time, "%Y-%m-%dT%H:%M:%S.%fZ").replace(tzinfo=timezone.utc)
+                        )
+                    )
+
+                    if len(batch) >= self.batch_size:
+                        yield batch
+                        batch = []
+
+        if batch:
+            yield batch
+
+    def poll_source(self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch) -> Generator[list[Document], None, None]:
+        """Poll source to get documents"""
+        start_dt = datetime.fromtimestamp(start, tz=timezone.utc)
+        end_dt = datetime.fromtimestamp(end, tz=timezone.utc)
+
+        for batch in self.load_from_state():
+            filtered: list[Document] = []
+
+            for doc in batch:
+                if not doc.doc_updated_at:
+                    continue
+
+                doc_dt = doc.doc_updated_at.astimezone(timezone.utc)
+
+                if start_dt <= doc_dt < end_dt:
+                    filtered.append(doc)
+
+            if filtered:
+                yield filtered
+
+if __name__ == "__main__":
+    import os
+
+    logging.basicConfig(level=logging.DEBUG)
+    connector = AirtableConnector("xxx","xxx")
+    connector.load_credentials({"airtable_access_token": os.environ.get("AIRTABLE_ACCESS_TOKEN")})
+    connector.validate_connector_settings()
+    document_batches = connector.load_from_state()
+    try:
+        first_batch = next(document_batches)
+        print(f"Loaded {len(first_batch)} documents in first batch.")
+        for doc in first_batch:
+            print(f"- {doc.semantic_identifier} ({doc.size_bytes} bytes)")
+    except StopIteration:
+        print("No documents available in Dropbox.")

common/data_source/asana_connector.py

@@ -0,0 +1,454 @@
+from collections.abc import Iterator
+import time
+from datetime import datetime
+import logging
+from typing import Any, Dict
+import asana
+import requests
+from common.data_source.config import CONTINUE_ON_CONNECTOR_FAILURE, INDEX_BATCH_SIZE, DocumentSource
+from common.data_source.interfaces import LoadConnector, PollConnector
+from common.data_source.models import Document, GenerateDocumentsOutput, SecondsSinceUnixEpoch
+from common.data_source.utils import extract_size_bytes, get_file_ext
+
+
+
+# https://github.com/Asana/python-asana/tree/master?tab=readme-ov-file#documentation-for-api-endpoints
+class AsanaTask:
+    def __init__(
+        self,
+        id: str,
+        title: str,
+        text: str,
+        link: str,
+        last_modified: datetime,
+        project_gid: str,
+        project_name: str,
+    ) -> None:
+        self.id = id
+        self.title = title
+        self.text = text
+        self.link = link
+        self.last_modified = last_modified
+        self.project_gid = project_gid
+        self.project_name = project_name
+
+    def __str__(self) -> str:
+        return f"ID: {self.id}\nTitle: {self.title}\nLast modified: {self.last_modified}\nText: {self.text}"
+
+
+class AsanaAPI:
+    def __init__(
+        self, api_token: str, workspace_gid: str, team_gid: str | None
+    ) -> None:
+        self._user = None
+        self.workspace_gid = workspace_gid
+        self.team_gid = team_gid
+
+        self.configuration = asana.Configuration()
+        self.api_client = asana.ApiClient(self.configuration)
+        self.tasks_api = asana.TasksApi(self.api_client)
+        self.attachments_api = asana.AttachmentsApi(self.api_client)
+        self.stories_api = asana.StoriesApi(self.api_client)
+        self.users_api = asana.UsersApi(self.api_client)
+        self.project_api = asana.ProjectsApi(self.api_client)
+        self.project_memberships_api = asana.ProjectMembershipsApi(self.api_client)
+        self.workspaces_api = asana.WorkspacesApi(self.api_client)
+
+        self.api_error_count = 0
+        self.configuration.access_token = api_token
+        self.task_count = 0
+
+    def get_tasks(
+        self, project_gids: list[str] | None, start_date: str
+    ) -> Iterator[AsanaTask]:
+        """Get all tasks from the projects with the given gids that were modified since the given date.
+        If project_gids is None, get all tasks from all projects in the workspace."""
+        logging.info("Starting to fetch Asana projects")
+        projects = self.project_api.get_projects(
+            opts={
+                "workspace": self.workspace_gid,
+                "opt_fields": "gid,name,archived,modified_at",
+            }
+        )
+        start_seconds = int(time.mktime(datetime.now().timetuple()))
+        projects_list = []
+        project_count = 0
+        for project_info in projects:
+            project_gid = project_info["gid"]
+            if project_gids is None or project_gid in project_gids:
+                projects_list.append(project_gid)
+            else:
+                logging.debug(
+                    f"Skipping project: {project_gid} - not in accepted project_gids"
+                )
+            project_count += 1
+            if project_count % 100 == 0:
+                logging.info(f"Processed {project_count} projects")
+        logging.info(f"Found {len(projects_list)} projects to process")
+        for project_gid in projects_list:
+            for task in self._get_tasks_for_project(
+                project_gid, start_date, start_seconds
+            ):
+                yield task
+        logging.info(f"Completed fetching {self.task_count} tasks from Asana")
+        if self.api_error_count > 0:
+            logging.warning(
+                f"Encountered {self.api_error_count} API errors during task fetching"
+            )
+
+    def _get_tasks_for_project(
+        self, project_gid: str, start_date: str, start_seconds: int
+    ) -> Iterator[AsanaTask]:
+        project = self.project_api.get_project(project_gid, opts={})
+        project_name = project.get("name", project_gid)
+        team = project.get("team") or {}
+        team_gid = team.get("gid")
+
+        if project.get("archived"):
+            logging.info(f"Skipping archived project: {project_name} ({project_gid})")
+            return
+        if not team_gid:
+            logging.info(
+                f"Skipping project without a team: {project_name} ({project_gid})"
+            )
+            return
+        if project.get("privacy_setting") == "private":
+            if self.team_gid and team_gid != self.team_gid:
+                logging.info(
+                    f"Skipping private project not in configured team: {project_name} ({project_gid})"
+                )
+                return
+            logging.info(
+                f"Processing private project in configured team: {project_name} ({project_gid})"
+            )
+
+        simple_start_date = start_date.split(".")[0].split("+")[0]
+        logging.info(
+            f"Fetching tasks modified since {simple_start_date} for project: {project_name} ({project_gid})"
+        )
+
+        opts = {
+            "opt_fields": "name,memberships,memberships.project,completed_at,completed_by,created_at,"
+            "created_by,custom_fields,dependencies,due_at,due_on,external,html_notes,liked,likes,"
+            "modified_at,notes,num_hearts,parent,projects,resource_subtype,resource_type,start_on,"
+            "workspace,permalink_url",
+            "modified_since": start_date,
+        }
+        tasks_from_api = self.tasks_api.get_tasks_for_project(project_gid, opts)
+        for data in tasks_from_api:
+            self.task_count += 1
+            if self.task_count % 10 == 0:
+                end_seconds = time.mktime(datetime.now().timetuple())
+                runtime_seconds = end_seconds - start_seconds
+                if runtime_seconds > 0:
+                    logging.info(
+                        f"Processed {self.task_count} tasks in {runtime_seconds:.0f} seconds "
+                        f"({self.task_count / runtime_seconds:.2f} tasks/second)"
+                    )
+
+            logging.debug(f"Processing Asana task: {data['name']}")
+
+            text = self._construct_task_text(data)
+
+            try:
+                text += self._fetch_and_add_comments(data["gid"])
+
+                last_modified_date = self.format_date(data["modified_at"])
+                text += f"Last modified: {last_modified_date}\n"
+
+                task = AsanaTask(
+                    id=data["gid"],
+                    title=data["name"],
+                    text=text,
+                    link=data["permalink_url"],
+                    last_modified=datetime.fromisoformat(data["modified_at"]),
+                    project_gid=project_gid,
+                    project_name=project_name,
+                )
+                yield task
+            except Exception:
+                logging.error(
+                    f"Error processing task {data['gid']} in project {project_gid}",
+                    exc_info=True,
+                )
+                self.api_error_count += 1
+
+    def _construct_task_text(self, data: Dict) -> str:
+        text = f"{data['name']}\n\n"
+
+        if data["notes"]:
+            text += f"{data['notes']}\n\n"
+
+        if data["created_by"] and data["created_by"]["gid"]:
+            creator = self.get_user(data["created_by"]["gid"])["name"]
+            created_date = self.format_date(data["created_at"])
+            text += f"Created by: {creator} on {created_date}\n"
+
+        if data["due_on"]:
+            due_date = self.format_date(data["due_on"])
+            text += f"Due date: {due_date}\n"
+
+        if data["completed_at"]:
+            completed_date = self.format_date(data["completed_at"])
+            text += f"Completed on: {completed_date}\n"
+
+        text += "\n"
+        return text
+
+    def _fetch_and_add_comments(self, task_gid: str) -> str:
+        text = ""
+        stories_opts: Dict[str, str] = {}
+        story_start = time.time()
+        stories = self.stories_api.get_stories_for_task(task_gid, stories_opts)
+
+        story_count = 0
+        comment_count = 0
+
+        for story in stories:
+            story_count += 1
+            if story["resource_subtype"] == "comment_added":
+                comment = self.stories_api.get_story(
+                    story["gid"], opts={"opt_fields": "text,created_by,created_at"}
+                )
+                commenter = self.get_user(comment["created_by"]["gid"])["name"]
+                text += f"Comment by {commenter}: {comment['text']}\n\n"
+                comment_count += 1
+
+        story_duration = time.time() - story_start
+        logging.debug(
+            f"Processed {story_count} stories (including {comment_count} comments) in {story_duration:.2f} seconds"
+        )
+
+        return text
+
+    def get_attachments(self, task_gid: str) -> list[dict]:
+        """
+        Fetch full attachment info (including download_url) for a task.
+        """
+        attachments: list[dict] = []
+
+        try:
+            # Step 1: list attachment compact records
+            for att in self.attachments_api.get_attachments_for_object(
+                parent=task_gid,
+                opts={}
+            ):
+                gid = att.get("gid")
+                if not gid:
+                    continue
+
+                try:
+                    # Step 2: expand to full attachment
+                    full = self.attachments_api.get_attachment(
+                        attachment_gid=gid,
+                        opts={
+                            "opt_fields": "name,download_url,size,created_at"
+                        }
+                    )
+
+                    if full.get("download_url"):
+                        attachments.append(full)
+
+                except Exception:
+                    logging.exception(
+                        f"Failed to fetch attachment detail {gid} for task {task_gid}"
+                    )
+                    self.api_error_count += 1
+
+        except Exception:
+            logging.exception(f"Failed to list attachments for task {task_gid}")
+            self.api_error_count += 1
+
+        return attachments
+
+    def get_accessible_emails(
+        self,
+        workspace_id: str,
+        project_ids: list[str] | None,
+        team_id: str | None,
+    ):
+
+        ws_users = self.users_api.get_users(
+            opts={
+                "workspace": workspace_id,
+                "opt_fields": "gid,name,email"
+            }
+        )
+
+        workspace_users = {
+            u["gid"]: u.get("email")
+            for u in ws_users
+            if u.get("email")
+        }
+
+        if not project_ids:
+            return set(workspace_users.values())
+
+
+        project_emails = set()
+
+        for pid in project_ids:
+            project = self.project_api.get_project(
+                pid,
+                opts={"opt_fields": "team,privacy_setting"}
+            )
+
+            if project["privacy_setting"] == "private":
+                if team_id and project.get("team", {}).get("gid") != team_id:
+                    continue
+
+            memberships = self.project_memberships_api.get_project_membership(
+                pid,
+                opts={"opt_fields": "user.gid,user.email"}
+            )
+
+            for m in memberships:
+                email = m["user"].get("email")
+                if email:
+                    project_emails.add(email)
+
+        return project_emails
+
+    def get_user(self, user_gid: str) -> Dict:
+        if self._user is not None:
+            return self._user
+        self._user = self.users_api.get_user(user_gid, {"opt_fields": "name,email"})
+
+        if not self._user:
+            logging.warning(f"Unable to fetch user information for user_gid: {user_gid}")
+            return {"name": "Unknown"}
+        return self._user
+
+    def format_date(self, date_str: str) -> str:
+        date = datetime.fromisoformat(date_str)
+        return time.strftime("%Y-%m-%d", date.timetuple())
+
+    def get_time(self) -> str:
+        return time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
+
+
+class AsanaConnector(LoadConnector, PollConnector):
+    def __init__(
+        self,
+        asana_workspace_id: str,
+        asana_project_ids: str | None = None,
+        asana_team_id: str | None = None,
+        batch_size: int = INDEX_BATCH_SIZE,
+        continue_on_failure: bool = CONTINUE_ON_CONNECTOR_FAILURE,
+    ) -> None:
+        self.workspace_id = asana_workspace_id
+        self.project_ids_to_index: list[str] | None = (
+            asana_project_ids.split(",") if asana_project_ids else None
+        )
+        self.asana_team_id = asana_team_id if asana_team_id else None
+        self.batch_size = batch_size
+        self.continue_on_failure = continue_on_failure
+        self.size_threshold = None
+        logging.info(
+            f"AsanaConnector initialized with workspace_id: {asana_workspace_id}"
+        )
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        self.api_token = credentials["asana_api_token_secret"]
+        self.asana_client = AsanaAPI(
+            api_token=self.api_token,
+            workspace_gid=self.workspace_id,
+            team_gid=self.asana_team_id,
+        )
+        self.workspace_users_email = self.asana_client.get_accessible_emails(self.workspace_id, self.project_ids_to_index, self.asana_team_id)
+        logging.info("Asana credentials loaded and API client initialized")
+        return None
+
+    def poll_source(
+        self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch | None
+    ) -> GenerateDocumentsOutput:
+        start_time = datetime.fromtimestamp(start).isoformat()
+        logging.info(f"Starting Asana poll from {start_time}")
+        docs_batch: list[Document] = []
+        tasks = self.asana_client.get_tasks(self.project_ids_to_index, start_time)
+        for task in tasks:
+            docs = self._task_to_documents(task)
+            docs_batch.extend(docs)
+
+            if len(docs_batch) >= self.batch_size:
+                logging.info(f"Yielding batch of {len(docs_batch)} documents")
+                yield docs_batch
+                docs_batch = []
+
+        if docs_batch:
+            logging.info(f"Yielding final batch of {len(docs_batch)} documents")
+            yield docs_batch
+
+        logging.info("Asana poll completed")
+
+    def load_from_state(self) -> GenerateDocumentsOutput:
+        logging.info("Starting full index of all Asana tasks")
+        return self.poll_source(start=0, end=None)
+
+    def _task_to_documents(self, task: AsanaTask) -> list[Document]:
+        docs: list[Document] = []
+
+        attachments = self.asana_client.get_attachments(task.id)
+
+        for att in attachments:
+            try:
+                resp = requests.get(att["download_url"], timeout=30)
+                resp.raise_for_status()
+                file_blob = resp.content
+                filename = att.get("name", "attachment")
+                size_bytes = extract_size_bytes(att)
+                if (
+                    self.size_threshold is not None
+                    and isinstance(size_bytes, int)
+                    and size_bytes > self.size_threshold
+                ):
+                    logging.warning(
+                        f"{filename} exceeds size threshold of {self.size_threshold}. Skipping."
+                    )
+                    continue
+                docs.append(
+                    Document(
+                        id=f"asana:{task.id}:{att['gid']}",
+                        blob=file_blob,
+                        extension=get_file_ext(filename) or "",
+                        size_bytes=size_bytes,
+                        doc_updated_at=task.last_modified,
+                        source=DocumentSource.ASANA,
+                        semantic_identifier=filename,
+                        primary_owners=list(self.workspace_users_email),
+                    )
+                )
+            except Exception:
+                logging.exception(
+                    f"Failed to download attachment {att.get('gid')} for task {task.id}"
+                )
+
+        return docs
+
+
+
+if __name__ == "__main__":
+    import time
+    import os
+
+    logging.info("Starting Asana connector test")
+    connector = AsanaConnector(
+        os.environ["WORKSPACE_ID"],
+        os.environ["PROJECT_IDS"],
+        os.environ["TEAM_ID"],
+    )
+    connector.load_credentials(
+        {
+            "asana_api_token_secret": os.environ["API_TOKEN"],
+        }
+    )
+    logging.info("Loading all documents from Asana")
+    all_docs = connector.load_from_state()
+    current = time.time()
+    one_day_ago = current - 24 * 60 * 60  # 1 day
+    logging.info("Polling for documents updated in the last 24 hours")
+    latest_docs = connector.poll_source(one_day_ago, current)
+    for docs in all_docs:
+        for doc in docs:
+            print(doc.id)
+    logging.info("Asana connector test completed")

common/data_source/bitbucket/connector.py

@@ -0,0 +1,388 @@
+from __future__ import annotations
+
+import copy
+from collections.abc import Callable
+from collections.abc import Iterator
+from datetime import datetime
+from datetime import timezone
+from typing import Any
+from typing import TYPE_CHECKING
+
+from typing_extensions import override
+
+from common.data_source.config import INDEX_BATCH_SIZE
+from common.data_source.config import DocumentSource
+from common.data_source.config import REQUEST_TIMEOUT_SECONDS
+from common.data_source.exceptions import ( 
+    ConnectorMissingCredentialError,
+    CredentialExpiredError,
+    InsufficientPermissionsError,
+    UnexpectedValidationError,
+)
+from common.data_source.interfaces import CheckpointedConnector
+from common.data_source.interfaces import CheckpointOutput
+from common.data_source.interfaces import IndexingHeartbeatInterface
+from common.data_source.interfaces import SecondsSinceUnixEpoch
+from common.data_source.interfaces import SlimConnectorWithPermSync
+from common.data_source.models import ConnectorCheckpoint
+from common.data_source.models import ConnectorFailure
+from common.data_source.models import DocumentFailure
+from common.data_source.models import SlimDocument
+from common.data_source.bitbucket.utils import (
+    build_auth_client,
+    list_repositories,
+    map_pr_to_document,
+    paginate,
+    PR_LIST_RESPONSE_FIELDS,
+    SLIM_PR_LIST_RESPONSE_FIELDS,
+)
+
+if TYPE_CHECKING:
+    import httpx
+
+
+class BitbucketConnectorCheckpoint(ConnectorCheckpoint):
+    """Checkpoint state for resumable Bitbucket PR indexing.
+
+    Fields:
+        repos_queue: Materialized list of repository slugs to process.
+        current_repo_index: Index of the repository currently being processed.
+        next_url: Bitbucket "next" URL for continuing pagination within the current repo.
+    """
+
+    repos_queue: list[str] = []
+    current_repo_index: int = 0
+    next_url: str | None = None
+
+
+class BitbucketConnector(
+    CheckpointedConnector[BitbucketConnectorCheckpoint],
+    SlimConnectorWithPermSync,
+):
+    """Connector for indexing Bitbucket Cloud pull requests.
+
+    Args:
+        workspace: Bitbucket workspace ID.
+        repositories: Comma-separated list of repository slugs to index.
+        projects: Comma-separated list of project keys to index all repositories within.
+        batch_size: Max number of documents to yield per batch.
+    """
+
+    def __init__(
+        self,
+        workspace: str,
+        repositories: str | None = None,
+        projects: str | None = None,
+        batch_size: int = INDEX_BATCH_SIZE,
+    ) -> None:
+        self.workspace = workspace
+        self._repositories = (
+            [s.strip() for s in repositories.split(",") if s.strip()]
+            if repositories
+            else None
+        )
+        self._projects: list[str] | None = (
+            [s.strip() for s in projects.split(",") if s.strip()] if projects else None
+        )
+        self.batch_size = batch_size
+        self.email: str | None = None
+        self.api_token: str | None = None
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        """Load API token-based credentials.
+
+        Expects a dict with keys: `bitbucket_email`, `bitbucket_api_token`.
+        """
+        self.email = credentials.get("bitbucket_email")
+        self.api_token = credentials.get("bitbucket_api_token")
+        if not self.email or not self.api_token:
+            raise ConnectorMissingCredentialError("Bitbucket")
+        return None
+
+    def _client(self) -> httpx.Client:
+        """Build an authenticated HTTP client or raise if credentials missing."""
+        if not self.email or not self.api_token:
+            raise ConnectorMissingCredentialError("Bitbucket")
+        return build_auth_client(self.email, self.api_token)
+
+    def _iter_pull_requests_for_repo(
+        self,
+        client: httpx.Client,
+        repo_slug: str,
+        params: dict[str, Any] | None = None,
+        start_url: str | None = None,
+        on_page: Callable[[str | None], None] | None = None,
+    ) -> Iterator[dict[str, Any]]:
+        base = f"https://api.bitbucket.org/2.0/repositories/{self.workspace}/{repo_slug}/pullrequests"
+        yield from paginate(
+            client,
+            base,
+            params,
+            start_url=start_url,
+            on_page=on_page,
+        )
+
+    def _build_params(
+        self,
+        fields: str = PR_LIST_RESPONSE_FIELDS,
+        start: SecondsSinceUnixEpoch | None = None,
+        end: SecondsSinceUnixEpoch | None = None,
+    ) -> dict[str, Any]:
+        """Build Bitbucket fetch params.
+
+        Always include OPEN, MERGED, and DECLINED PRs. If both ``start`` and
+        ``end`` are provided, apply a single updated_on time window.
+        """
+
+        def _iso(ts: SecondsSinceUnixEpoch) -> str:
+            return datetime.fromtimestamp(ts, tz=timezone.utc).isoformat()
+
+        def _tc_epoch(
+            lower_epoch: SecondsSinceUnixEpoch | None,
+            upper_epoch: SecondsSinceUnixEpoch | None,
+        ) -> str | None:
+            if lower_epoch is not None and upper_epoch is not None:
+                lower_iso = _iso(lower_epoch)
+                upper_iso = _iso(upper_epoch)
+                return f'(updated_on > "{lower_iso}" AND updated_on <= "{upper_iso}")'
+            return None
+
+        params: dict[str, Any] = {"fields": fields, "pagelen": 50}
+        time_clause = _tc_epoch(start, end)
+        q = '(state = "OPEN" OR state = "MERGED" OR state = "DECLINED")'
+        if time_clause:
+            q = f"{q} AND {time_clause}"
+        params["q"] = q
+        return params
+
+    def _iter_target_repositories(self, client: httpx.Client) -> Iterator[str]:
+        """Yield repository slugs based on configuration.
+
+        Priority:
+        - repositories list
+        - projects list (list repos by project key)
+        - workspace (all repos)
+        """
+        if self._repositories:
+            for slug in self._repositories:
+                yield slug
+            return
+        if self._projects:
+            for project_key in self._projects:
+                for repo in list_repositories(client, self.workspace, project_key):
+                    slug_val = repo.get("slug")
+                    if isinstance(slug_val, str) and slug_val:
+                        yield slug_val
+            return
+        for repo in list_repositories(client, self.workspace, None):
+            slug_val = repo.get("slug")
+            if isinstance(slug_val, str) and slug_val:
+                yield slug_val
+
+    @override
+    def load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: BitbucketConnectorCheckpoint,
+    ) -> CheckpointOutput[BitbucketConnectorCheckpoint]:
+        """Resumable PR ingestion across repos and pages within a time window.
+
+        Yields Documents (or ConnectorFailure for per-PR mapping failures) and returns
+        an updated checkpoint that records repo position and next page URL.
+        """
+        new_checkpoint = copy.deepcopy(checkpoint)
+
+        with self._client() as client:
+            # Materialize target repositories once
+            if not new_checkpoint.repos_queue:
+                # Preserve explicit order; otherwise ensure deterministic ordering
+                repos_list = list(self._iter_target_repositories(client))
+                new_checkpoint.repos_queue = sorted(set(repos_list))
+                new_checkpoint.current_repo_index = 0
+                new_checkpoint.next_url = None
+
+            repos = new_checkpoint.repos_queue
+            if not repos or new_checkpoint.current_repo_index >= len(repos):
+                new_checkpoint.has_more = False
+                return new_checkpoint
+
+            repo_slug = repos[new_checkpoint.current_repo_index]
+
+            first_page_params = self._build_params(
+                fields=PR_LIST_RESPONSE_FIELDS,
+                start=start,
+                end=end,
+            )
+
+            def _on_page(next_url: str | None) -> None:
+                new_checkpoint.next_url = next_url
+
+            for pr in self._iter_pull_requests_for_repo(
+                client,
+                repo_slug,
+                params=first_page_params,
+                start_url=new_checkpoint.next_url,
+                on_page=_on_page,
+            ):
+                try:
+                    document = map_pr_to_document(pr, self.workspace, repo_slug)
+                    yield document
+                except Exception as e:
+                    pr_id = pr.get("id")
+                    pr_link = (
+                        f"https://bitbucket.org/{self.workspace}/{repo_slug}/pull-requests/{pr_id}"
+                        if pr_id is not None
+                        else None
+                    )
+                    yield ConnectorFailure(
+                        failed_document=DocumentFailure(
+                            document_id=(
+                                f"{DocumentSource.BITBUCKET.value}:{self.workspace}:{repo_slug}:pr:{pr_id}"
+                                if pr_id is not None
+                                else f"{DocumentSource.BITBUCKET.value}:{self.workspace}:{repo_slug}:pr:unknown"
+                            ),
+                            document_link=pr_link,
+                        ),
+                        failure_message=f"Failed to process Bitbucket PR: {e}",
+                        exception=e,
+                    )
+
+            # Advance to next repository (if any) and set has_more accordingly
+            new_checkpoint.current_repo_index += 1
+            new_checkpoint.next_url = None
+            new_checkpoint.has_more = new_checkpoint.current_repo_index < len(repos)
+
+        return new_checkpoint
+
+    @override
+    def build_dummy_checkpoint(self) -> BitbucketConnectorCheckpoint:
+        """Create an initial checkpoint with work remaining."""
+        return BitbucketConnectorCheckpoint(has_more=True)
+
+    @override
+    def validate_checkpoint_json(
+        self, checkpoint_json: str
+    ) -> BitbucketConnectorCheckpoint:
+        """Validate and deserialize a checkpoint instance from JSON."""
+        return BitbucketConnectorCheckpoint.model_validate_json(checkpoint_json)
+
+    def retrieve_all_slim_docs_perm_sync(
+        self,
+        start: SecondsSinceUnixEpoch | None = None,
+        end: SecondsSinceUnixEpoch | None = None,
+        callback: IndexingHeartbeatInterface | None = None,
+    ) -> Iterator[list[SlimDocument]]:
+        """Return only document IDs for all existing pull requests."""
+        batch: list[SlimDocument] = []
+        params = self._build_params(
+            fields=SLIM_PR_LIST_RESPONSE_FIELDS,
+            start=start,
+            end=end,
+        )
+        with self._client() as client:
+            for slug in self._iter_target_repositories(client):
+                for pr in self._iter_pull_requests_for_repo(
+                    client, slug, params=params
+                ):
+                    pr_id = pr["id"]
+                    doc_id = f"{DocumentSource.BITBUCKET.value}:{self.workspace}:{slug}:pr:{pr_id}"
+                    batch.append(SlimDocument(id=doc_id))
+                    if len(batch) >= self.batch_size:
+                        yield batch
+                        batch = []
+                        if callback:
+                            if callback.should_stop():
+                                # Note: this is not actually used for permission sync yet, just pruning
+                                raise RuntimeError(
+                                    "bitbucket_pr_sync: Stop signal detected"
+                                )
+                            callback.progress("bitbucket_pr_sync", len(batch))
+        if batch:
+            yield batch
+
+    def validate_connector_settings(self) -> None:
+        """Validate Bitbucket credentials and workspace access by probing a lightweight endpoint.
+
+        Raises:
+            CredentialExpiredError: on HTTP 401
+            InsufficientPermissionsError: on HTTP 403
+            UnexpectedValidationError: on any other failure
+        """
+        try:
+            with self._client() as client:
+                url = f"https://api.bitbucket.org/2.0/repositories/{self.workspace}"
+                resp = client.get(
+                    url,
+                    params={"pagelen": 1, "fields": "pagelen"},
+                    timeout=REQUEST_TIMEOUT_SECONDS,
+                )
+                if resp.status_code == 401:
+                    raise CredentialExpiredError(
+                        "Invalid or expired Bitbucket credentials (HTTP 401)."
+                    )
+                if resp.status_code == 403:
+                    raise InsufficientPermissionsError(
+                        "Insufficient permissions to access Bitbucket workspace (HTTP 403)."
+                    )
+                if resp.status_code < 200 or resp.status_code >= 300:
+                    raise UnexpectedValidationError(
+                        f"Unexpected Bitbucket error (status={resp.status_code})."
+                    )
+        except Exception as e:
+            # Network or other unexpected errors
+            if isinstance(
+                e,
+                (
+                    CredentialExpiredError,
+                    InsufficientPermissionsError,
+                    UnexpectedValidationError,
+                    ConnectorMissingCredentialError,
+                ),
+            ):
+                raise
+            raise UnexpectedValidationError(
+                f"Unexpected error while validating Bitbucket settings: {e}"
+            )
+
+if __name__ == "__main__":
+    bitbucket = BitbucketConnector(
+        workspace="<YOUR_WORKSPACE>"
+    )
+
+    bitbucket.load_credentials({
+        "bitbucket_email": "<YOUR_EMAIL>",
+        "bitbucket_api_token": "<YOUR_API_TOKEN>",
+    })
+
+    bitbucket.validate_connector_settings()
+    print("Credentials validated successfully.")
+
+    start_time = datetime.fromtimestamp(0, tz=timezone.utc)
+    end_time = datetime.now(timezone.utc)
+
+    for doc_batch in bitbucket.retrieve_all_slim_docs_perm_sync(
+        start=start_time.timestamp(),
+        end=end_time.timestamp(),
+    ):
+        for doc in doc_batch:
+            print(doc)
+
+
+    bitbucket_checkpoint = bitbucket.build_dummy_checkpoint()
+    
+    while bitbucket_checkpoint.has_more:
+        gen = bitbucket.load_from_checkpoint(
+            start=start_time.timestamp(),
+            end=end_time.timestamp(),
+            checkpoint=bitbucket_checkpoint,
+        )
+
+        while True:
+            try:
+                doc = next(gen)  
+                print(doc)
+            except StopIteration as e:
+                bitbucket_checkpoint = e.value  
+                break
+        

common/data_source/bitbucket/utils.py

@@ -0,0 +1,288 @@
+from __future__ import annotations
+
+import time
+from collections.abc import Callable
+from collections.abc import Iterator
+from datetime import datetime
+from datetime import timezone
+from typing import Any
+
+import httpx
+
+from common.data_source.config import REQUEST_TIMEOUT_SECONDS, DocumentSource
+from common.data_source.cross_connector_utils.rate_limit_wrapper import (
+    rate_limit_builder,
+)
+from common.data_source.utils import sanitize_filename
+from common.data_source.models import BasicExpertInfo, Document
+from common.data_source.cross_connector_utils.retry_wrapper import retry_builder
+
+# Fields requested from Bitbucket PR list endpoint to ensure rich PR data
+PR_LIST_RESPONSE_FIELDS: str = ",".join(
+    [
+        "next",
+        "page",
+        "pagelen",
+        "values.author",
+        "values.close_source_branch",
+        "values.closed_by",
+        "values.comment_count",
+        "values.created_on",
+        "values.description",
+        "values.destination",
+        "values.draft",
+        "values.id",
+        "values.links",
+        "values.merge_commit",
+        "values.participants",
+        "values.reason",
+        "values.rendered",
+        "values.reviewers",
+        "values.source",
+        "values.state",
+        "values.summary",
+        "values.task_count",
+        "values.title",
+        "values.type",
+        "values.updated_on",
+    ]
+)
+
+# Minimal fields for slim retrieval (IDs only)
+SLIM_PR_LIST_RESPONSE_FIELDS: str = ",".join(
+    [
+        "next",
+        "page",
+        "pagelen",
+        "values.id",
+    ]
+)
+
+
+# Minimal fields for repository list calls
+REPO_LIST_RESPONSE_FIELDS: str = ",".join(
+    [
+        "next",
+        "page",
+        "pagelen",
+        "values.slug",
+        "values.full_name",
+        "values.project.key",
+    ]
+)
+
+
+class BitbucketRetriableError(Exception):
+    """Raised for retriable Bitbucket conditions (429, 5xx)."""
+
+
+class BitbucketNonRetriableError(Exception):
+    """Raised for non-retriable Bitbucket client errors (4xx except 429)."""
+
+
+@retry_builder(
+    tries=6,
+    delay=1,
+    backoff=2,
+    max_delay=30,
+    exceptions=(BitbucketRetriableError, httpx.RequestError),
+)
+@rate_limit_builder(max_calls=60, period=60)
+def bitbucket_get(
+    client: httpx.Client, url: str, params: dict[str, Any] | None = None
+) -> httpx.Response:
+    """Perform a GET against Bitbucket with retry and rate limiting.
+
+    Retries on 429 and 5xx responses, and on transport errors. Honors
+    `Retry-After` header for 429 when present by sleeping before retrying.
+    """
+    try:
+        response = client.get(url, params=params, timeout=REQUEST_TIMEOUT_SECONDS)
+    except httpx.RequestError:
+        # Allow retry_builder to handle retries of transport errors
+        raise
+
+    try:
+        response.raise_for_status()
+    except httpx.HTTPStatusError as e:
+        status = e.response.status_code if e.response is not None else None
+        if status == 429:
+            retry_after = e.response.headers.get("Retry-After") if e.response else None
+            if retry_after is not None:
+                try:
+                    time.sleep(int(retry_after))
+                except (TypeError, ValueError):
+                    pass
+            raise BitbucketRetriableError("Bitbucket rate limit exceeded (429)") from e
+        if status is not None and 500 <= status < 600:
+            raise BitbucketRetriableError(f"Bitbucket server error: {status}") from e
+        if status is not None and 400 <= status < 500:
+            raise BitbucketNonRetriableError(f"Bitbucket client error: {status}") from e
+        # Unknown status, propagate
+        raise
+
+    return response
+
+
+def build_auth_client(email: str, api_token: str) -> httpx.Client:
+    """Create an authenticated httpx client for Bitbucket Cloud API."""
+    return httpx.Client(auth=(email, api_token), http2=True)
+
+
+def paginate(
+    client: httpx.Client,
+    url: str,
+    params: dict[str, Any] | None = None,
+    start_url: str | None = None,
+    on_page: Callable[[str | None], None] | None = None,
+) -> Iterator[dict[str, Any]]:
+    """Iterate over paginated Bitbucket API responses yielding individual values.
+
+    Args:
+        client: Authenticated HTTP client.
+        url: Base collection URL (first page when start_url is None).
+        params: Query params for the first page.
+        start_url: If provided, start from this absolute URL (ignores params).
+        on_page: Optional callback invoked after each page with the next page URL.
+    """
+    next_url = start_url or url
+    # If resuming from a next URL, do not pass params again
+    query = params.copy() if params else None
+    query = None if start_url else query
+    while next_url:
+        resp = bitbucket_get(client, next_url, params=query)
+        data = resp.json()
+        values = data.get("values", [])
+        for item in values:
+            yield item
+        next_url = data.get("next")
+        if on_page is not None:
+            on_page(next_url)
+        # only include params on first call, next_url will contain all necessary params
+        query = None
+
+
+def list_repositories(
+    client: httpx.Client, workspace: str, project_key: str | None = None
+) -> Iterator[dict[str, Any]]:
+    """List repositories in a workspace, optionally filtered by project key."""
+    base_url = f"https://api.bitbucket.org/2.0/repositories/{workspace}"
+    params: dict[str, Any] = {
+        "fields": REPO_LIST_RESPONSE_FIELDS,
+        "pagelen": 100,
+        # Ensure deterministic ordering
+        "sort": "full_name",
+    }
+    if project_key:
+        params["q"] = f'project.key="{project_key}"'
+    yield from paginate(client, base_url, params)
+
+
+def map_pr_to_document(pr: dict[str, Any], workspace: str, repo_slug: str) -> Document:
+    """Map a Bitbucket pull request JSON to Onyx Document."""
+    pr_id = pr["id"]
+    title = pr.get("title") or f"PR {pr_id}"
+    description = pr.get("description") or ""
+    state = pr.get("state")
+    draft = pr.get("draft", False)
+    author = pr.get("author", {})
+    reviewers = pr.get("reviewers", [])
+    participants = pr.get("participants", [])
+
+    link = pr.get("links", {}).get("html", {}).get("href") or (
+        f"https://bitbucket.org/{workspace}/{repo_slug}/pull-requests/{pr_id}"
+    )
+
+    created_on = pr.get("created_on")
+    updated_on = pr.get("updated_on")
+    updated_dt = (
+        datetime.fromisoformat(updated_on.replace("Z", "+00:00")).astimezone(
+            timezone.utc
+        )
+        if isinstance(updated_on, str)
+        else None
+    )
+
+    source_branch = pr.get("source", {}).get("branch", {}).get("name", "")
+    destination_branch = pr.get("destination", {}).get("branch", {}).get("name", "")
+
+    approved_by = [
+        _get_user_name(p.get("user", {})) for p in participants if p.get("approved")
+    ]
+
+    primary_owner = None
+    if author:
+        primary_owner = BasicExpertInfo(
+            display_name=_get_user_name(author),
+        )
+
+    # secondary_owners = [ 
+    #     BasicExpertInfo(display_name=_get_user_name(r)) for r in reviewers
+    # ] or None 
+
+    reviewer_names = [_get_user_name(r) for r in reviewers]
+
+    # Create a concise summary of key PR info
+    created_date = created_on.split("T")[0] if created_on else "N/A"
+    updated_date = updated_on.split("T")[0] if updated_on else "N/A"
+    content_text = (
+        "Pull Request Information:\n"
+        f"- Pull Request ID: {pr_id}\n"
+        f"- Title: {title}\n"
+        f"- State: {state or 'N/A'} {'(Draft)' if draft else ''}\n"
+    )
+    if state == "DECLINED":
+        content_text += f"- Reason: {pr.get('reason', 'N/A')}\n"
+    content_text += (
+        f"- Author: {_get_user_name(author) if author else 'N/A'}\n"
+        f"- Reviewers: {', '.join(reviewer_names) if reviewer_names else 'N/A'}\n"
+        f"- Branch: {source_branch} -> {destination_branch}\n"
+        f"- Created: {created_date}\n"
+        f"- Updated: {updated_date}"
+    )
+    if description:
+        content_text += f"\n\nDescription:\n{description}"
+
+    metadata: dict[str, str | list[str]] = {
+        "object_type": "PullRequest",
+        "workspace": workspace,
+        "repository": repo_slug,
+        "pr_key": f"{workspace}/{repo_slug}#{pr_id}",
+        "id": str(pr_id),
+        "title": title,
+        "state": state or "",
+        "draft": str(bool(draft)),
+        "link": link,
+        "author": _get_user_name(author) if author else "",
+        "reviewers": reviewer_names,
+        "approved_by": approved_by,
+        "comment_count": str(pr.get("comment_count", "")),
+        "task_count": str(pr.get("task_count", "")),
+        "created_on": created_on or "",
+        "updated_on": updated_on or "",
+        "source_branch": source_branch,
+        "destination_branch": destination_branch,
+        "closed_by": (
+            _get_user_name(pr.get("closed_by", {})) if pr.get("closed_by") else ""
+        ),
+        "close_source_branch": str(bool(pr.get("close_source_branch", False))),
+    }
+
+    name = sanitize_filename(title, "md")
+
+    return Document(
+        id=f"{DocumentSource.BITBUCKET.value}:{workspace}:{repo_slug}:pr:{pr_id}",
+        blob=content_text.encode("utf-8"),
+        source=DocumentSource.BITBUCKET,
+        extension=".md",
+        semantic_identifier=f"#{pr_id}: {name}",
+        size_bytes=len(content_text.encode("utf-8")),
+        doc_updated_at=updated_dt,
+        primary_owners=[primary_owner] if primary_owner else None,
+        # secondary_owners=secondary_owners,
+        metadata=metadata,
+    )
+
+
+def _get_user_name(user: dict[str, Any]) -> str:
+    return user.get("display_name") or user.get("nickname") or "unknown"

common/data_source/blob_connector.py

@@ -56,7 +56,7 @@ class BlobStorageConnector(LoadConnector, PollConnector):
 
         # Validate credentials
         if self.bucket_type == BlobType.R2:
-            if not all(
+            if not all( 
                 credentials.get(key)
                 for key in ["r2_access_key_id", "r2_secret_access_key", "account_id"]
             ):
@@ -64,15 +64,23 @@ class BlobStorageConnector(LoadConnector, PollConnector):
 
         elif self.bucket_type == BlobType.S3:
             authentication_method = credentials.get("authentication_method", "access_key")
+
             if authentication_method == "access_key":
                 if not all(
                     credentials.get(key)
                     for key in ["aws_access_key_id", "aws_secret_access_key"]
                 ):
                     raise ConnectorMissingCredentialError("Amazon S3")
+
             elif authentication_method == "iam_role":
                 if not credentials.get("aws_role_arn"):
                     raise ConnectorMissingCredentialError("Amazon S3 IAM role ARN is required")
+                
+            elif authentication_method == "assume_role":
+                pass
+
+            else:
+                raise ConnectorMissingCredentialError("Unsupported S3 authentication method")
 
         elif self.bucket_type == BlobType.GOOGLE_CLOUD_STORAGE:
             if not all(
@@ -120,55 +128,72 @@ class BlobStorageConnector(LoadConnector, PollConnector):
         paginator = self.s3_client.get_paginator("list_objects_v2")
         pages = paginator.paginate(Bucket=self.bucket_name, Prefix=self.prefix)
 
-        batch: list[Document] = []
+        # Collect all objects first to count filename occurrences
+        all_objects = []
         for page in pages:
             if "Contents" not in page:
                 continue
-
             for obj in page["Contents"]:
                 if obj["Key"].endswith("/"):
                     continue
-
                 last_modified = obj["LastModified"].replace(tzinfo=timezone.utc)
+                if start < last_modified <= end:
+                    all_objects.append(obj)
+        
+        # Count filename occurrences to determine which need full paths
+        filename_counts: dict[str, int] = {}
+        for obj in all_objects:
+            file_name = os.path.basename(obj["Key"])
+            filename_counts[file_name] = filename_counts.get(file_name, 0) + 1
 
-                if not (start < last_modified <= end):
+        batch: list[Document] = []
+        for obj in all_objects:
+            last_modified = obj["LastModified"].replace(tzinfo=timezone.utc)
+            file_name = os.path.basename(obj["Key"])
+            key = obj["Key"]
+
+            size_bytes = extract_size_bytes(obj)
+            if (
+                self.size_threshold is not None
+                and isinstance(size_bytes, int)
+                and size_bytes > self.size_threshold
+            ):
+                logging.warning(
+                    f"{file_name} exceeds size threshold of {self.size_threshold}. Skipping."
+                )
+                continue
+            
+            try:
+                blob = download_object(self.s3_client, self.bucket_name, key, self.size_threshold)
+                if blob is None:
                     continue
 
-                file_name = os.path.basename(obj["Key"])
-                key = obj["Key"]
+                # Use full path only if filename appears multiple times
+                if filename_counts.get(file_name, 0) > 1:
+                    relative_path = key
+                    if self.prefix and key.startswith(self.prefix):
+                        relative_path = key[len(self.prefix):]
+                    semantic_id = relative_path.replace('/', ' / ') if relative_path else file_name
+                else:
+                    semantic_id = file_name
 
-                size_bytes = extract_size_bytes(obj)
-                if (
-                    self.size_threshold is not None
-                    and isinstance(size_bytes, int)
-                    and size_bytes > self.size_threshold
-                ):
-                    logging.warning(
-                        f"{file_name} exceeds size threshold of {self.size_threshold}. Skipping."
+                batch.append(
+                    Document(
+                        id=f"{self.bucket_type}:{self.bucket_name}:{key}",
+                        blob=blob,
+                        source=DocumentSource(self.bucket_type.value),
+                        semantic_identifier=semantic_id,
+                        extension=get_file_ext(file_name),
+                        doc_updated_at=last_modified,
+                        size_bytes=size_bytes if size_bytes else 0
                     )
-                    continue
-                try:
-                    blob = download_object(self.s3_client, self.bucket_name, key, self.size_threshold)
-                    if blob is None:
-                        continue
+                )
+                if len(batch) == self.batch_size:
+                    yield batch
+                    batch = []
 
-                    batch.append(
-                        Document(
-                            id=f"{self.bucket_type}:{self.bucket_name}:{key}",
-                            blob=blob,
-                            source=DocumentSource(self.bucket_type.value),
-                            semantic_identifier=file_name,
-                            extension=get_file_ext(file_name),
-                            doc_updated_at=last_modified,
-                            size_bytes=size_bytes if size_bytes else 0
-                        )
-                    )
-                    if len(batch) == self.batch_size:
-                        yield batch
-                        batch = []
-
-                except Exception:
-                    logging.exception(f"Error decoding object {key}")
+            except Exception:
+                logging.exception(f"Error decoding object {key}")
         
         if batch:
             yield batch
@@ -276,4 +301,4 @@ if __name__ == "__main__":
     except ConnectorMissingCredentialError as e:
         print(f"Error: {e}")
     except Exception as e:
-        print(f"An unexpected error occurred: {e}")
+        print(f"An unexpected error occurred: {e}")

common/data_source/config.py

@@ -13,6 +13,9 @@ def get_current_tz_offset() -> int:
     return round(time_diff.total_seconds() / 3600)
 
 
+# Default request timeout, mostly used by connectors
+REQUEST_TIMEOUT_SECONDS = int(os.environ.get("REQUEST_TIMEOUT_SECONDS") or 60)
+
 ONE_MINUTE = 60
 ONE_HOUR = 3600
 ONE_DAY = ONE_HOUR * 24
@@ -53,6 +56,14 @@ class DocumentSource(str, Enum):
     S3_COMPATIBLE = "s3_compatible"
     DROPBOX = "dropbox"
     BOX = "box"
+    AIRTABLE = "airtable"
+    ASANA = "asana"
+    GITHUB = "github"
+    GITLAB = "gitlab"
+    IMAP = "imap"
+    BITBUCKET = "bitbucket"
+    ZENDESK = "zendesk"
+
 
 class FileOrigin(str, Enum):
     """File origins"""
@@ -83,6 +94,7 @@ _PAGE_EXPANSION_FIELDS = [
     "space",
     "metadata.labels",
     "history.lastUpdated",
+    "ancestors",
 ]
 
 
@@ -229,6 +241,8 @@ _REPLACEMENT_EXPANSIONS = "body.view.value"
 
 BOX_WEB_OAUTH_REDIRECT_URI = os.environ.get("BOX_WEB_OAUTH_REDIRECT_URI", "http://localhost:9380/v1/connector/box/oauth/web/callback")
 
+GITHUB_CONNECTOR_BASE_URL = os.environ.get("GITHUB_CONNECTOR_BASE_URL") or None
+
 class HtmlBasedConnectorTransformLinksStrategy(str, Enum):
     # remove links entirely
     STRIP = "strip"
@@ -250,6 +264,22 @@ WEB_CONNECTOR_IGNORED_ELEMENTS = os.environ.get(
     "WEB_CONNECTOR_IGNORED_ELEMENTS", "nav,footer,meta,script,style,symbol,aside"
 ).split(",")
 
+AIRTABLE_CONNECTOR_SIZE_THRESHOLD = int(
+    os.environ.get("AIRTABLE_CONNECTOR_SIZE_THRESHOLD", 10 * 1024 * 1024)
+)
+
+ASANA_CONNECTOR_SIZE_THRESHOLD = int(
+    os.environ.get("ASANA_CONNECTOR_SIZE_THRESHOLD", 10 * 1024 * 1024)
+)
+
+IMAP_CONNECTOR_SIZE_THRESHOLD = int(
+    os.environ.get("IMAP_CONNECTOR_SIZE_THRESHOLD", 10 * 1024 * 1024)
+)
+
+ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS = os.environ.get(
+    "ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS", ""
+).split(",")
+
 _USER_NOT_FOUND = "Unknown Confluence User"
 
 _COMMENT_EXPANSION_FIELDS = ["body.storage.value"]

common/data_source/confluence_connector.py

@@ -186,7 +186,7 @@ class OnyxConfluence:
         # between the db and redis everywhere the credentials might be updated
         new_credential_str = json.dumps(new_credentials)
         self.redis_client.set(
-            self.credential_key, new_credential_str, nx=True, ex=self.CREDENTIAL_TTL
+            self.credential_key, new_credential_str, exp=self.CREDENTIAL_TTL
         )
         self._credentials_provider.set_credentials(new_credentials)
 
@@ -1311,6 +1311,9 @@ class ConfluenceConnector(
         self._low_timeout_confluence_client: OnyxConfluence | None = None
         self._fetched_titles: set[str] = set()
         self.allow_images = False
+        # Track document names to detect duplicates
+        self._document_name_counts: dict[str, int] = {}
+        self._document_name_paths: dict[str, list[str]] = {}
 
         # Remove trailing slash from wiki_base if present
         self.wiki_base = wiki_base.rstrip("/")
@@ -1513,6 +1516,40 @@ class ConfluenceConnector(
                 self.wiki_base, page["_links"]["webui"], self.is_cloud
             )
 
+            # Build hierarchical path for semantic identifier
+            space_name = page.get("space", {}).get("name", "")
+            
+            # Build path from ancestors
+            path_parts = []
+            if space_name:
+                path_parts.append(space_name)
+            
+            # Add ancestor pages to path if available
+            if "ancestors" in page and page["ancestors"]:
+                for ancestor in page["ancestors"]:
+                    ancestor_title = ancestor.get("title", "")
+                    if ancestor_title:
+                        path_parts.append(ancestor_title)
+            
+            # Add current page title
+            path_parts.append(page_title)
+            
+            # Track page names for duplicate detection
+            full_path = " / ".join(path_parts) if len(path_parts) > 1 else page_title
+            
+            # Count occurrences of this page title
+            if page_title not in self._document_name_counts:
+                self._document_name_counts[page_title] = 0
+                self._document_name_paths[page_title] = []
+            self._document_name_counts[page_title] += 1
+            self._document_name_paths[page_title].append(full_path)
+            
+            # Use simple name if no duplicates, otherwise use full path
+            if self._document_name_counts[page_title] == 1:
+                semantic_identifier = page_title
+            else:
+                semantic_identifier = full_path
+
             # Get the page content
             page_content = extract_text_from_confluence_html(
                 self.confluence_client, page, self._fetched_titles
@@ -1559,11 +1596,11 @@ class ConfluenceConnector(
             return Document(
                 id=page_url,
                 source=DocumentSource.CONFLUENCE,
-                semantic_identifier=page_title,
+                semantic_identifier=semantic_identifier,
                 extension=".html",  # Confluence pages are HTML
                 blob=page_content.encode("utf-8"),  # Encode page content as bytes
-                size_bytes=len(page_content.encode("utf-8")),  # Calculate size in bytes
                 doc_updated_at=datetime_from_string(page["version"]["when"]),
+                size_bytes=len(page_content.encode("utf-8")),  # Calculate size in bytes
                 primary_owners=primary_owners if primary_owners else None,
                 metadata=metadata if metadata else None,
             )
@@ -1601,7 +1638,6 @@ class ConfluenceConnector(
             expand=",".join(_ATTACHMENT_EXPANSION_FIELDS),
         ):
             media_type: str = attachment.get("metadata", {}).get("mediaType", "")
-
             # TODO(rkuo): this check is partially redundant with validate_attachment_filetype
             # and checks in convert_attachment_to_content/process_attachment
             # but doing the check here avoids an unnecessary download. Due for refactoring.
@@ -1669,6 +1705,34 @@ class ConfluenceConnector(
                     self.wiki_base, attachment["_links"]["webui"], self.is_cloud
                 )
 
+                # Build semantic identifier with space and page context
+                attachment_title = attachment.get("title", object_url)
+                space_name = page.get("space", {}).get("name", "")
+                page_title = page.get("title", "")
+                
+                # Create hierarchical name: Space / Page / Attachment
+                attachment_path_parts = []
+                if space_name:
+                    attachment_path_parts.append(space_name)
+                if page_title:
+                    attachment_path_parts.append(page_title)
+                attachment_path_parts.append(attachment_title)
+                
+                full_attachment_path = " / ".join(attachment_path_parts) if len(attachment_path_parts) > 1 else attachment_title
+                
+                # Track attachment names for duplicate detection
+                if attachment_title not in self._document_name_counts:
+                    self._document_name_counts[attachment_title] = 0
+                    self._document_name_paths[attachment_title] = []
+                self._document_name_counts[attachment_title] += 1
+                self._document_name_paths[attachment_title].append(full_attachment_path)
+                
+                # Use simple name if no duplicates, otherwise use full path
+                if self._document_name_counts[attachment_title] == 1:
+                    attachment_semantic_identifier = attachment_title
+                else:
+                    attachment_semantic_identifier = full_attachment_path
+
                 primary_owners: list[BasicExpertInfo] | None = None
                 if "version" in attachment and "by" in attachment["version"]:
                     author = attachment["version"]["by"]
@@ -1680,11 +1744,12 @@ class ConfluenceConnector(
 
                 extension = Path(attachment.get("title", "")).suffix or ".unknown"
 
+
                 attachment_doc = Document(
                     id=attachment_id,
                     # sections=sections,
                     source=DocumentSource.CONFLUENCE,
-                    semantic_identifier=attachment.get("title", object_url),
+                    semantic_identifier=attachment_semantic_identifier,
                     extension=extension,
                     blob=file_blob,
                     size_bytes=len(file_blob),
@@ -1741,7 +1806,7 @@ class ConfluenceConnector(
             start_ts, end, self.batch_size
         )
         logging.debug(f"page_query_url: {page_query_url}")
-
+        
         # store the next page start for confluence server, cursor for confluence cloud
         def store_next_page_url(next_page_url: str) -> None:
             checkpoint.next_page_url = next_page_url

common/data_source/connector_runner.py

@@ -0,0 +1,217 @@
+import sys
+import time
+import logging
+from collections.abc import Generator
+from datetime import datetime
+from typing import Generic
+from typing import TypeVar
+from common.data_source.interfaces import (
+    BaseConnector,
+    CheckpointedConnector,
+    CheckpointedConnectorWithPermSync,
+    CheckpointOutput,
+    LoadConnector,
+    PollConnector,
+)
+from common.data_source.models import ConnectorCheckpoint, ConnectorFailure, Document
+
+
+TimeRange = tuple[datetime, datetime]
+
+CT = TypeVar("CT", bound=ConnectorCheckpoint)
+
+
+def batched_doc_ids(
+    checkpoint_connector_generator: CheckpointOutput[CT],
+    batch_size: int,
+) -> Generator[set[str], None, None]:
+    batch: set[str] = set()
+    for document, failure, next_checkpoint in CheckpointOutputWrapper[CT]()(
+        checkpoint_connector_generator
+    ):
+        if document is not None:
+            batch.add(document.id)
+        elif (
+            failure and failure.failed_document and failure.failed_document.document_id
+        ):
+            batch.add(failure.failed_document.document_id)
+
+        if len(batch) >= batch_size:
+            yield batch
+            batch = set()
+    if len(batch) > 0:
+        yield batch
+
+
+class CheckpointOutputWrapper(Generic[CT]):
+    """
+    Wraps a CheckpointOutput generator to give things back in a more digestible format,
+    specifically for Document outputs.
+    The connector format is easier for the connector implementor (e.g. it enforces exactly
+    one new checkpoint is returned AND that the checkpoint is at the end), thus the different
+    formats.
+    """
+
+    def __init__(self) -> None:
+        self.next_checkpoint: CT | None = None
+
+    def __call__(
+        self,
+        checkpoint_connector_generator: CheckpointOutput[CT],
+    ) -> Generator[
+        tuple[Document | None, ConnectorFailure | None, CT | None],
+        None,
+        None,
+    ]:
+        # grabs the final return value and stores it in the `next_checkpoint` variable
+        def _inner_wrapper(
+            checkpoint_connector_generator: CheckpointOutput[CT],
+        ) -> CheckpointOutput[CT]:
+            self.next_checkpoint = yield from checkpoint_connector_generator
+            return self.next_checkpoint  # not used
+
+        for document_or_failure in _inner_wrapper(checkpoint_connector_generator):
+            if isinstance(document_or_failure, Document):
+                yield document_or_failure, None, None
+            elif isinstance(document_or_failure, ConnectorFailure):
+                yield None, document_or_failure, None
+            else:
+                raise ValueError(
+                    f"Invalid document_or_failure type: {type(document_or_failure)}"
+                )
+
+        if self.next_checkpoint is None:
+            raise RuntimeError(
+                "Checkpoint is None. This should never happen - the connector should always return a checkpoint."
+            )
+
+        yield None, None, self.next_checkpoint
+
+
+class ConnectorRunner(Generic[CT]):
+    """
+    Handles:
+        - Batching
+        - Additional exception logging
+        - Combining different connector types to a single interface
+    """
+
+    def __init__(
+        self,
+        connector: BaseConnector,
+        batch_size: int,
+        # cannot be True for non-checkpointed connectors
+        include_permissions: bool,
+        time_range: TimeRange | None = None,
+    ):
+        if not isinstance(connector, CheckpointedConnector) and include_permissions:
+            raise ValueError(
+                "include_permissions cannot be True for non-checkpointed connectors"
+            )
+
+        self.connector = connector
+        self.time_range = time_range
+        self.batch_size = batch_size
+        self.include_permissions = include_permissions
+
+        self.doc_batch: list[Document] = []
+
+    def run(self, checkpoint: CT) -> Generator[
+        tuple[list[Document] | None, ConnectorFailure | None, CT | None],
+        None,
+        None,
+    ]:
+        """Adds additional exception logging to the connector."""
+        try:
+            if isinstance(self.connector, CheckpointedConnector):
+                if self.time_range is None:
+                    raise ValueError("time_range is required for CheckpointedConnector")
+
+                start = time.monotonic()
+                if self.include_permissions:
+                    if not isinstance(
+                        self.connector, CheckpointedConnectorWithPermSync
+                    ):
+                        raise ValueError(
+                            "Connector does not support permission syncing"
+                        )
+                    load_from_checkpoint = (
+                        self.connector.load_from_checkpoint_with_perm_sync
+                    )
+                else:
+                    load_from_checkpoint = self.connector.load_from_checkpoint
+                checkpoint_connector_generator = load_from_checkpoint(
+                    start=self.time_range[0].timestamp(),
+                    end=self.time_range[1].timestamp(),
+                    checkpoint=checkpoint,
+                )
+                next_checkpoint: CT | None = None
+                # this is guaranteed to always run at least once with next_checkpoint being non-None
+                for document, failure, next_checkpoint in CheckpointOutputWrapper[CT]()(
+                    checkpoint_connector_generator
+                ):
+                    if document is not None and isinstance(document, Document):
+                        self.doc_batch.append(document)
+
+                    if failure is not None:
+                        yield None, failure, None
+
+                    if len(self.doc_batch) >= self.batch_size:
+                        yield self.doc_batch, None, None
+                        self.doc_batch = []
+
+                # yield remaining documents
+                if len(self.doc_batch) > 0:
+                    yield self.doc_batch, None, None
+                    self.doc_batch = []
+
+                yield None, None, next_checkpoint
+
+                logging.debug(
+                    f"Connector took {time.monotonic() - start} seconds to get to the next checkpoint."
+                )
+
+            else:
+                finished_checkpoint = self.connector.build_dummy_checkpoint()
+                finished_checkpoint.has_more = False
+
+                if isinstance(self.connector, PollConnector):
+                    if self.time_range is None:
+                        raise ValueError("time_range is required for PollConnector")
+
+                    for document_batch in self.connector.poll_source(
+                        start=self.time_range[0].timestamp(),
+                        end=self.time_range[1].timestamp(),
+                    ):
+                        yield document_batch, None, None
+
+                    yield None, None, finished_checkpoint
+                elif isinstance(self.connector, LoadConnector):
+                    for document_batch in self.connector.load_from_state():
+                        yield document_batch, None, None
+
+                    yield None, None, finished_checkpoint
+                else:
+                    raise ValueError(f"Invalid connector. type: {type(self.connector)}")
+        except Exception:
+            exc_type, _, exc_traceback = sys.exc_info()
+
+            # Traverse the traceback to find the last frame where the exception was raised
+            tb = exc_traceback
+            if tb is None:
+                logging.error("No traceback found for exception")
+                raise
+
+            while tb.tb_next:
+                tb = tb.tb_next  # Move to the next frame in the traceback
+
+            # Get the local variables from the frame where the exception occurred
+            local_vars = tb.tb_frame.f_locals
+            local_vars_str = "\n".join(
+                f"{key}: {value}" for key, value in local_vars.items()
+            )
+            logging.error(
+                f"Error in connector. type: {exc_type};\n"
+                f"local_vars below -> \n{local_vars_str[:1024]}"
+            )
+            raise

common/data_source/cross_connector_utils/rate_limit_wrapper.py

@@ -0,0 +1,126 @@
+import time
+import logging
+from collections.abc import Callable
+from functools import wraps
+from typing import Any
+from typing import cast
+from typing import TypeVar
+
+import requests
+
+F = TypeVar("F", bound=Callable[..., Any])
+
+
+class RateLimitTriedTooManyTimesError(Exception):
+    pass
+
+
+class _RateLimitDecorator:
+    """Builds a generic wrapper/decorator for calls to external APIs that
+    prevents making more than `max_calls` requests per `period`
+
+    Implementation inspired by the `ratelimit` library:
+    https://github.com/tomasbasham/ratelimit.
+
+    NOTE: is not thread safe.
+    """
+
+    def __init__(
+        self,
+        max_calls: int,
+        period: float,  # in seconds
+        sleep_time: float = 2,  # in seconds
+        sleep_backoff: float = 2,  # applies exponential backoff
+        max_num_sleep: int = 0,
+    ):
+        self.max_calls = max_calls
+        self.period = period
+        self.sleep_time = sleep_time
+        self.sleep_backoff = sleep_backoff
+        self.max_num_sleep = max_num_sleep
+
+        self.call_history: list[float] = []
+        self.curr_calls = 0
+
+    def __call__(self, func: F) -> F:
+        @wraps(func)
+        def wrapped_func(*args: list, **kwargs: dict[str, Any]) -> Any:
+            # cleanup calls which are no longer relevant
+            self._cleanup()
+
+            # check if we've exceeded the rate limit
+            sleep_cnt = 0
+            while len(self.call_history) == self.max_calls:
+                sleep_time = self.sleep_time * (self.sleep_backoff**sleep_cnt)
+                logging.warning(
+                    f"Rate limit exceeded for function {func.__name__}. "
+                    f"Waiting {sleep_time} seconds before retrying."
+                )
+                time.sleep(sleep_time)
+                sleep_cnt += 1
+                if self.max_num_sleep != 0 and sleep_cnt >= self.max_num_sleep:
+                    raise RateLimitTriedTooManyTimesError(
+                        f"Exceeded '{self.max_num_sleep}' retries for function '{func.__name__}'"
+                    )
+
+                self._cleanup()
+
+            # add the current call to the call history
+            self.call_history.append(time.monotonic())
+            return func(*args, **kwargs)
+
+        return cast(F, wrapped_func)
+
+    def _cleanup(self) -> None:
+        curr_time = time.monotonic()
+        time_to_expire_before = curr_time - self.period
+        self.call_history = [
+            call_time
+            for call_time in self.call_history
+            if call_time > time_to_expire_before
+        ]
+
+
+rate_limit_builder = _RateLimitDecorator
+
+
+"""If you want to allow the external service to tell you when you've hit the rate limit,
+use the following instead"""
+
+R = TypeVar("R", bound=Callable[..., requests.Response])
+
+
+def wrap_request_to_handle_ratelimiting(
+    request_fn: R, default_wait_time_sec: int = 30, max_waits: int = 30
+) -> R:
+    def wrapped_request(*args: list, **kwargs: dict[str, Any]) -> requests.Response:
+        for _ in range(max_waits):
+            response = request_fn(*args, **kwargs)
+            if response.status_code == 429:
+                try:
+                    wait_time = int(
+                        response.headers.get("Retry-After", default_wait_time_sec)
+                    )
+                except ValueError:
+                    wait_time = default_wait_time_sec
+
+                time.sleep(wait_time)
+                continue
+
+            return response
+
+        raise RateLimitTriedTooManyTimesError(f"Exceeded '{max_waits}' retries")
+
+    return cast(R, wrapped_request)
+
+
+_rate_limited_get = wrap_request_to_handle_ratelimiting(requests.get)
+_rate_limited_post = wrap_request_to_handle_ratelimiting(requests.post)
+
+
+class _RateLimitedRequest:
+    get = _rate_limited_get
+    post = _rate_limited_post
+
+
+rl_requests = _RateLimitedRequest

common/data_source/cross_connector_utils/retry_wrapper.py

@@ -0,0 +1,88 @@
+from collections.abc import Callable
+import logging
+from logging import Logger
+from typing import Any
+from typing import cast
+from typing import TypeVar
+import requests
+from retry import retry
+
+from common.data_source.config import REQUEST_TIMEOUT_SECONDS
+
+
+F = TypeVar("F", bound=Callable[..., Any])
+logger = logging.getLogger(__name__)
+
+def retry_builder(
+    tries: int = 20,
+    delay: float = 0.1,
+    max_delay: float | None = 60,
+    backoff: float = 2,
+    jitter: tuple[float, float] | float = 1,
+    exceptions: type[Exception] | tuple[type[Exception], ...] = (Exception,),
+) -> Callable[[F], F]:
+    """Builds a generic wrapper/decorator for calls to external APIs that
+    may fail due to rate limiting, flakes, or other reasons. Applies exponential
+    backoff with jitter to retry the call."""
+
+    def retry_with_default(func: F) -> F:
+        @retry(
+            tries=tries,
+            delay=delay,
+            max_delay=max_delay,
+            backoff=backoff,
+            jitter=jitter,
+            logger=cast(Logger, logger),
+            exceptions=exceptions,
+        )
+        def wrapped_func(*args: list, **kwargs: dict[str, Any]) -> Any:
+            return func(*args, **kwargs)
+
+        return cast(F, wrapped_func)
+
+    return retry_with_default
+
+
+def request_with_retries(
+    method: str,
+    url: str,
+    *,
+    data: dict[str, Any] | None = None,
+    headers: dict[str, Any] | None = None,
+    params: dict[str, Any] | None = None,
+    timeout: int = REQUEST_TIMEOUT_SECONDS,
+    stream: bool = False,
+    tries: int = 8,
+    delay: float = 1,
+    backoff: float = 2,
+) -> requests.Response:
+    @retry(tries=tries, delay=delay, backoff=backoff, logger=cast(Logger, logger))
+    def _make_request() -> requests.Response:
+        response = requests.request(
+            method=method,
+            url=url,
+            data=data,
+            headers=headers,
+            params=params,
+            timeout=timeout,
+            stream=stream,
+        )
+        try:
+            response.raise_for_status()
+        except requests.exceptions.HTTPError:
+            logging.exception(
+                "Request failed:\n%s",
+                {
+                    "method": method,
+                    "url": url,
+                    "data": data,
+                    "headers": headers,
+                    "params": params,
+                    "timeout": timeout,
+                    "stream": stream,
+                },
+            )
+            raise
+        return response
+
+    return _make_request()

common/data_source/dropbox_connector.py

@@ -87,15 +87,69 @@ class DropboxConnector(LoadConnector, PollConnector):
         if self.dropbox_client is None:
             raise ConnectorMissingCredentialError("Dropbox")
 
+        # Collect all files first to count filename occurrences
+        all_files = []
+        self._collect_files_recursive(path, start, end, all_files)
+        
+        # Count filename occurrences
+        filename_counts: dict[str, int] = {}
+        for entry, _ in all_files:
+            filename_counts[entry.name] = filename_counts.get(entry.name, 0) + 1
+        
+        # Process files in batches
+        batch: list[Document] = []
+        for entry, downloaded_file in all_files:
+            modified_time = entry.client_modified
+            if modified_time.tzinfo is None:
+                modified_time = modified_time.replace(tzinfo=timezone.utc)
+            else:
+                modified_time = modified_time.astimezone(timezone.utc)
+            
+            # Use full path only if filename appears multiple times
+            if filename_counts.get(entry.name, 0) > 1:
+                # Remove leading slash and replace slashes with ' / '
+                relative_path = entry.path_display.lstrip('/')
+                semantic_id = relative_path.replace('/', ' / ') if relative_path else entry.name
+            else:
+                semantic_id = entry.name
+            
+            batch.append(
+                Document(
+                    id=f"dropbox:{entry.id}",
+                    blob=downloaded_file,
+                    source=DocumentSource.DROPBOX,
+                    semantic_identifier=semantic_id,
+                    extension=get_file_ext(entry.name),
+                    doc_updated_at=modified_time,
+                    size_bytes=entry.size if getattr(entry, "size", None) is not None else len(downloaded_file),
+                )
+            )
+            
+            if len(batch) == self.batch_size:
+                yield batch
+                batch = []
+        
+        if batch:
+            yield batch
+
+    def _collect_files_recursive(
+        self,
+        path: str,
+        start: SecondsSinceUnixEpoch | None,
+        end: SecondsSinceUnixEpoch | None,
+        all_files: list,
+    ) -> None:
+        """Recursively collect all files matching time criteria."""
+        if self.dropbox_client is None:
+            raise ConnectorMissingCredentialError("Dropbox")
+
         result = self.dropbox_client.files_list_folder(
             path,
-            limit=self.batch_size,
             recursive=False,
             include_non_downloadable_files=False,
         )
 
         while True:
-            batch: list[Document] = []
             for entry in result.entries:
                 if isinstance(entry, FileMetadata):
                     modified_time = entry.client_modified
@@ -112,27 +166,13 @@ class DropboxConnector(LoadConnector, PollConnector):
 
                     try:
                         downloaded_file = self._download_file(entry.path_display)
+                        all_files.append((entry, downloaded_file))
                     except Exception:
                         logger.exception(f"[Dropbox]: Error downloading file {entry.path_display}")
                         continue
 
-                    batch.append(
-                        Document(
-                            id=f"dropbox:{entry.id}",
-                            blob=downloaded_file,
-                            source=DocumentSource.DROPBOX,
-                            semantic_identifier=entry.name,
-                            extension=get_file_ext(entry.name),
-                            doc_updated_at=modified_time,
-                            size_bytes=entry.size if getattr(entry, "size", None) is not None else len(downloaded_file),
-                        )
-                    )
-
                 elif isinstance(entry, FolderMetadata):
-                    yield from self._yield_files_recursive(entry.path_lower, start, end)
-
-            if batch:
-                yield batch
+                    self._collect_files_recursive(entry.path_lower, start, end, all_files)
 
             if not result.has_more:
                 break

common/data_source/file_types.py

@@ -18,6 +18,7 @@ class UploadMimeTypes:
         "text/plain",
         "text/markdown",
         "text/x-markdown",
+        "text/mdx",
         "text/x-config",
         "text/tab-separated-values",
         "application/json",

common/data_source/github/connector.py

@@ -0,0 +1,973 @@
+import copy
+import logging
+from collections.abc import Callable
+from collections.abc import Generator
+from datetime import datetime
+from datetime import timedelta
+from datetime import timezone
+from enum import Enum
+from typing import Any
+from typing import cast
+
+from github import Github, Auth
+from github import RateLimitExceededException
+from github import Repository
+from github.GithubException import GithubException
+from github.Issue import Issue
+from github.NamedUser import NamedUser
+from github.PaginatedList import PaginatedList
+from github.PullRequest import PullRequest
+from pydantic import BaseModel
+from typing_extensions import override
+from common.data_source.utils import sanitize_filename
+from common.data_source.config import DocumentSource, GITHUB_CONNECTOR_BASE_URL
+from common.data_source.exceptions import (
+    ConnectorMissingCredentialError,
+    ConnectorValidationError,
+    CredentialExpiredError,
+    InsufficientPermissionsError,
+    UnexpectedValidationError,
+)
+from common.data_source.interfaces import CheckpointedConnectorWithPermSyncGH, CheckpointOutput
+from common.data_source.models import (
+    ConnectorCheckpoint,
+    ConnectorFailure,
+    Document,
+    DocumentFailure,
+    ExternalAccess,
+    SecondsSinceUnixEpoch,
+)
+from common.data_source.connector_runner import ConnectorRunner
+from .models import SerializedRepository
+from .rate_limit_utils import sleep_after_rate_limit_exception
+from .utils import deserialize_repository
+from .utils import get_external_access_permission
+
+ITEMS_PER_PAGE = 100
+CURSOR_LOG_FREQUENCY = 50
+
+_MAX_NUM_RATE_LIMIT_RETRIES = 5
+
+ONE_DAY = timedelta(days=1)
+SLIM_BATCH_SIZE = 100
+# Cases
+# X (from start) standard run, no fallback to cursor-based pagination
+# X (from start) standard run errors, fallback to cursor-based pagination
+#  X error in the middle of a page
+#  X no errors: run to completion
+# X (from checkpoint) standard run, no fallback to cursor-based pagination
+# X (from checkpoint) continue from cursor-based pagination
+#  - retrying
+#  - no retrying
+
+# things to check:
+# checkpoint state on return
+# checkpoint progress (no infinite loop)
+
+
+class DocMetadata(BaseModel):
+    repo: str
+
+
+def get_nextUrl_key(pag_list: PaginatedList[PullRequest | Issue]) -> str:
+    if "_PaginatedList__nextUrl" in pag_list.__dict__:
+        return "_PaginatedList__nextUrl"
+    for key in pag_list.__dict__:
+        if "__nextUrl" in key:
+            return key
+    for key in pag_list.__dict__:
+        if "nextUrl" in key:
+            return key
+    return ""
+
+
+def get_nextUrl(
+    pag_list: PaginatedList[PullRequest | Issue], nextUrl_key: str
+) -> str | None:
+    return getattr(pag_list, nextUrl_key) if nextUrl_key else None
+
+
+def set_nextUrl(
+    pag_list: PaginatedList[PullRequest | Issue], nextUrl_key: str, nextUrl: str
+) -> None:
+    if nextUrl_key:
+        setattr(pag_list, nextUrl_key, nextUrl)
+    elif nextUrl:
+        raise ValueError("Next URL key not found: " + str(pag_list.__dict__))
+
+
+def _paginate_until_error(
+    git_objs: Callable[[], PaginatedList[PullRequest | Issue]],
+    cursor_url: str | None,
+    prev_num_objs: int,
+    cursor_url_callback: Callable[[str | None, int], None],
+    retrying: bool = False,
+) -> Generator[PullRequest | Issue, None, None]:
+    num_objs = prev_num_objs
+    pag_list = git_objs()
+    nextUrl_key = get_nextUrl_key(pag_list)
+    if cursor_url:
+        set_nextUrl(pag_list, nextUrl_key, cursor_url)
+    elif retrying:
+        # if we are retrying, we want to skip the objects retrieved
+        # over previous calls. Unfortunately, this WILL retrieve all
+        # pages before the one we are resuming from, so we really
+        # don't want this case to be hit often
+        logging.warning(
+            "Retrying from a previous cursor-based pagination call. "
+            "This will retrieve all pages before the one we are resuming from, "
+            "which may take a while and consume many API calls."
+        )
+        pag_list = cast(PaginatedList[PullRequest | Issue], pag_list[prev_num_objs:])
+        num_objs = 0
+
+    try:
+        # this for loop handles cursor-based pagination
+        for issue_or_pr in pag_list:
+            num_objs += 1
+            yield issue_or_pr
+            # used to store the current cursor url in the checkpoint. This value
+            # is updated during iteration over pag_list.
+            cursor_url_callback(get_nextUrl(pag_list, nextUrl_key), num_objs)
+
+            if num_objs % CURSOR_LOG_FREQUENCY == 0:
+                logging.info(
+                    f"Retrieved {num_objs} objects with current cursor url: {get_nextUrl(pag_list, nextUrl_key)}"
+                )
+
+    except Exception as e:
+        logging.exception(f"Error during cursor-based pagination: {e}")
+        if num_objs - prev_num_objs > 0:
+            raise
+
+        if get_nextUrl(pag_list, nextUrl_key) is not None and not retrying:
+            logging.info(
+                "Assuming that this error is due to cursor "
+                "expiration because no objects were retrieved. "
+                "Retrying from the first page."
+            )
+            yield from _paginate_until_error(
+                git_objs, None, prev_num_objs, cursor_url_callback, retrying=True
+            )
+            return
+
+        # for no cursor url or if we reach this point after a retry, raise the error
+        raise
+
+
+def _get_batch_rate_limited(
+    # We pass in a callable because we want git_objs to produce a fresh
+    # PaginatedList each time it's called to avoid using the same object for cursor-based pagination
+    # from a partial offset-based pagination call.
+    git_objs: Callable[[], PaginatedList],
+    page_num: int,
+    cursor_url: str | None,
+    prev_num_objs: int,
+    cursor_url_callback: Callable[[str | None, int], None],
+    github_client: Github,
+    attempt_num: int = 0,
+) -> Generator[PullRequest | Issue, None, None]:
+    if attempt_num > _MAX_NUM_RATE_LIMIT_RETRIES:
+        raise RuntimeError(
+            "Re-tried fetching batch too many times. Something is going wrong with fetching objects from Github"
+        )
+    try:
+        if cursor_url:
+            # when this is set, we are resuming from an earlier
+            # cursor-based pagination call.
+            yield from _paginate_until_error(
+                git_objs, cursor_url, prev_num_objs, cursor_url_callback
+            )
+            return
+        objs = list(git_objs().get_page(page_num))
+        # fetch all data here to disable lazy loading later
+        # this is needed to capture the rate limit exception here (if one occurs)
+        for obj in objs:
+            if hasattr(obj, "raw_data"):
+                getattr(obj, "raw_data")
+        yield from objs
+    except RateLimitExceededException:
+        sleep_after_rate_limit_exception(github_client)
+        yield from _get_batch_rate_limited(
+            git_objs,
+            page_num,
+            cursor_url,
+            prev_num_objs,
+            cursor_url_callback,
+            github_client,
+            attempt_num + 1,
+        )
+    except GithubException as e:
+        if not (
+            e.status == 422
+            and (
+                "cursor" in (e.message or "")
+                or "cursor" in (e.data or {}).get("message", "")
+            )
+        ):
+            raise
+        # Fallback to a cursor-based pagination strategy
+        # This can happen for "large datasets," but there's no documentation
+        # On the error on the web as far as we can tell.
+        # Error message:
+        # "Pagination with the page parameter is not supported for large datasets,
+        # please use cursor based pagination (after/before)"
+        yield from _paginate_until_error(
+            git_objs, cursor_url, prev_num_objs, cursor_url_callback
+        )
+
+
+def _get_userinfo(user: NamedUser) -> dict[str, str]:
+    def _safe_get(attr_name: str) -> str | None:
+        try:
+            return cast(str | None, getattr(user, attr_name))
+        except GithubException:
+            logging.debug(f"Error getting {attr_name} for user")
+            return None
+
+    return {
+        k: v
+        for k, v in {
+            "login": _safe_get("login"),
+            "name": _safe_get("name"),
+            "email": _safe_get("email"),
+        }.items()
+        if v is not None
+    }
+
+
+def _convert_pr_to_document(
+    pull_request: PullRequest, repo_external_access: ExternalAccess | None
+) -> Document:
+    repo_name = pull_request.base.repo.full_name if pull_request.base else ""
+    doc_metadata = DocMetadata(repo=repo_name)
+    file_content_byte = pull_request.body.encode('utf-8') if pull_request.body else b""
+    name = sanitize_filename(pull_request.title, "md")
+
+    return Document(
+        id=pull_request.html_url,
+        blob= file_content_byte,
+        source=DocumentSource.GITHUB,
+        external_access=repo_external_access,
+        semantic_identifier=f"{pull_request.number}:{name}",
+        # updated_at is UTC time but is timezone unaware, explicitly add UTC
+        # as there is logic in indexing to prevent wrong timestamped docs
+        # due to local time discrepancies with UTC
+        doc_updated_at=(
+            pull_request.updated_at.replace(tzinfo=timezone.utc)
+            if pull_request.updated_at
+            else None
+        ),
+        extension=".md",
+        # this metadata is used in perm sync
+        size_bytes=len(file_content_byte) if file_content_byte else 0,
+        primary_owners=[],
+        doc_metadata=doc_metadata.model_dump(),
+        metadata={
+            k: [str(vi) for vi in v] if isinstance(v, list) else str(v)
+            for k, v in {
+                "object_type": "PullRequest",
+                "id": pull_request.number,
+                "merged": pull_request.merged,
+                "state": pull_request.state,
+                "user": _get_userinfo(pull_request.user) if pull_request.user else None,
+                "assignees": [
+                    _get_userinfo(assignee) for assignee in pull_request.assignees
+                ],
+                "repo": (
+                    pull_request.base.repo.full_name if pull_request.base else None
+                ),
+                "num_commits": str(pull_request.commits),
+                "num_files_changed": str(pull_request.changed_files),
+                "labels": [label.name for label in pull_request.labels],
+                "created_at": (
+                    pull_request.created_at.replace(tzinfo=timezone.utc)
+                    if pull_request.created_at
+                    else None
+                ),
+                "updated_at": (
+                    pull_request.updated_at.replace(tzinfo=timezone.utc)
+                    if pull_request.updated_at
+                    else None
+                ),
+                "closed_at": (
+                    pull_request.closed_at.replace(tzinfo=timezone.utc)
+                    if pull_request.closed_at
+                    else None
+                ),
+                "merged_at": (
+                    pull_request.merged_at.replace(tzinfo=timezone.utc)
+                    if pull_request.merged_at
+                    else None
+                ),
+                "merged_by": (
+                    _get_userinfo(pull_request.merged_by)
+                    if pull_request.merged_by
+                    else None
+                ),
+            }.items()
+            if v is not None
+        },
+    )
+
+
+def _fetch_issue_comments(issue: Issue) -> str:
+    comments = issue.get_comments()
+    return "\nComment: ".join(comment.body for comment in comments)
+
+
+def _convert_issue_to_document(
+    issue: Issue, repo_external_access: ExternalAccess | None
+) -> Document:
+    repo_name = issue.repository.full_name if issue.repository else ""
+    doc_metadata = DocMetadata(repo=repo_name)
+    file_content_byte = issue.body.encode('utf-8') if issue.body else b""
+    name = sanitize_filename(issue.title, "md")
+
+    return Document(
+        id=issue.html_url,
+        blob=file_content_byte,
+        source=DocumentSource.GITHUB,
+        extension=".md",
+        external_access=repo_external_access,
+        semantic_identifier=f"{issue.number}:{name}",
+        # updated_at is UTC time but is timezone unaware
+        doc_updated_at=issue.updated_at.replace(tzinfo=timezone.utc),
+        # this metadata is used in perm sync
+        doc_metadata=doc_metadata.model_dump(),
+        size_bytes=len(file_content_byte) if file_content_byte else 0,
+        primary_owners=[_get_userinfo(issue.user) if issue.user else None],
+        metadata={
+            k: [str(vi) for vi in v] if isinstance(v, list) else str(v)
+            for k, v in {
+                "object_type": "Issue",
+                "id": issue.number,
+                "state": issue.state,
+                "user": _get_userinfo(issue.user) if issue.user else None,
+                "assignees": [_get_userinfo(assignee) for assignee in issue.assignees],
+                "repo": issue.repository.full_name if issue.repository else None,
+                "labels": [label.name for label in issue.labels],
+                "created_at": (
+                    issue.created_at.replace(tzinfo=timezone.utc)
+                    if issue.created_at
+                    else None
+                ),
+                "updated_at": (
+                    issue.updated_at.replace(tzinfo=timezone.utc)
+                    if issue.updated_at
+                    else None
+                ),
+                "closed_at": (
+                    issue.closed_at.replace(tzinfo=timezone.utc)
+                    if issue.closed_at
+                    else None
+                ),
+                "closed_by": (
+                    _get_userinfo(issue.closed_by) if issue.closed_by else None
+                ),
+            }.items()
+            if v is not None
+        },
+    )
+
+
+class GithubConnectorStage(Enum):
+    START = "start"
+    PRS = "prs"
+    ISSUES = "issues"
+
+
+class GithubConnectorCheckpoint(ConnectorCheckpoint):
+    stage: GithubConnectorStage
+    curr_page: int
+
+    cached_repo_ids: list[int] | None = None
+    cached_repo: SerializedRepository | None = None
+
+    # Used for the fallback cursor-based pagination strategy
+    num_retrieved: int
+    cursor_url: str | None = None
+
+    def reset(self) -> None:
+        """
+        Resets curr_page, num_retrieved, and cursor_url to their initial values (0, 0, None)
+        """
+        self.curr_page = 0
+        self.num_retrieved = 0
+        self.cursor_url = None
+
+
+def make_cursor_url_callback(
+    checkpoint: GithubConnectorCheckpoint,
+) -> Callable[[str | None, int], None]:
+    def cursor_url_callback(cursor_url: str | None, num_objs: int) -> None:
+        # we want to maintain the old cursor url so code after retrieval
+        # can determine that we are using the fallback cursor-based pagination strategy
+        if cursor_url:
+            checkpoint.cursor_url = cursor_url
+        checkpoint.num_retrieved = num_objs
+
+    return cursor_url_callback
+
+
+class GithubConnector(CheckpointedConnectorWithPermSyncGH[GithubConnectorCheckpoint]):
+    def __init__(
+        self,
+        repo_owner: str,
+        repositories: str | None = None,
+        state_filter: str = "all",
+        include_prs: bool = True,
+        include_issues: bool = False,
+    ) -> None:
+        self.repo_owner = repo_owner
+        self.repositories = repositories
+        self.state_filter = state_filter
+        self.include_prs = include_prs
+        self.include_issues = include_issues
+        self.github_client: Github | None = None
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        # defaults to 30 items per page, can be set to as high as 100
+        token = credentials["github_access_token"]
+        auth = Auth.Token(token)
+
+        if GITHUB_CONNECTOR_BASE_URL:
+            self.github_client = Github(
+                auth=auth,
+                base_url=GITHUB_CONNECTOR_BASE_URL,
+                per_page=ITEMS_PER_PAGE,
+            )
+        else:
+            self.github_client = Github(
+                auth=auth,
+                per_page=ITEMS_PER_PAGE,
+            )
+
+        return None
+
+    def get_github_repo(
+        self, github_client: Github, attempt_num: int = 0
+    ) -> Repository.Repository:
+        if attempt_num > _MAX_NUM_RATE_LIMIT_RETRIES:
+            raise RuntimeError(
+                "Re-tried fetching repo too many times. Something is going wrong with fetching objects from Github"
+            )
+
+        try:
+            return github_client.get_repo(f"{self.repo_owner}/{self.repositories}")
+        except RateLimitExceededException:
+            sleep_after_rate_limit_exception(github_client)
+            return self.get_github_repo(github_client, attempt_num + 1)
+
+    def get_github_repos(
+        self, github_client: Github, attempt_num: int = 0
+    ) -> list[Repository.Repository]:
+        """Get specific repositories based on comma-separated repo_name string."""
+        if attempt_num > _MAX_NUM_RATE_LIMIT_RETRIES:
+            raise RuntimeError(
+                "Re-tried fetching repos too many times. Something is going wrong with fetching objects from Github"
+            )
+
+        try:
+            repos = []
+            # Split repo_name by comma and strip whitespace
+            repo_names = [
+                name.strip() for name in (cast(str, self.repositories)).split(",")
+            ]
+
+            for repo_name in repo_names:
+                if repo_name:  # Skip empty strings
+                    try:
+                        repo = github_client.get_repo(f"{self.repo_owner}/{repo_name}")
+                        repos.append(repo)
+                    except GithubException as e:
+                        logging.warning(
+                            f"Could not fetch repo {self.repo_owner}/{repo_name}: {e}"
+                        )
+
+            return repos
+        except RateLimitExceededException:
+            sleep_after_rate_limit_exception(github_client)
+            return self.get_github_repos(github_client, attempt_num + 1)
+
+    def get_all_repos(
+        self, github_client: Github, attempt_num: int = 0
+    ) -> list[Repository.Repository]:
+        if attempt_num > _MAX_NUM_RATE_LIMIT_RETRIES:
+            raise RuntimeError(
+                "Re-tried fetching repos too many times. Something is going wrong with fetching objects from Github"
+            )
+
+        try:
+            # Try to get organization first
+            try:
+                org = github_client.get_organization(self.repo_owner)
+                return list(org.get_repos())
+
+            except GithubException:
+                # If not an org, try as a user
+                user = github_client.get_user(self.repo_owner)
+                return list(user.get_repos())
+        except RateLimitExceededException:
+            sleep_after_rate_limit_exception(github_client)
+            return self.get_all_repos(github_client, attempt_num + 1)
+
+    def _pull_requests_func(
+        self, repo: Repository.Repository
+    ) -> Callable[[], PaginatedList[PullRequest]]:
+        return lambda: repo.get_pulls(
+            state=self.state_filter, sort="updated", direction="desc"
+        )
+
+    def _issues_func(
+        self, repo: Repository.Repository
+    ) -> Callable[[], PaginatedList[Issue]]:
+        return lambda: repo.get_issues(
+            state=self.state_filter, sort="updated", direction="desc"
+        )
+
+    def _fetch_from_github(
+        self,
+        checkpoint: GithubConnectorCheckpoint,
+        start: datetime | None = None,
+        end: datetime | None = None,
+        include_permissions: bool = False,
+    ) -> Generator[Document | ConnectorFailure, None, GithubConnectorCheckpoint]:
+        if self.github_client is None:
+            raise ConnectorMissingCredentialError("GitHub")
+
+        checkpoint = copy.deepcopy(checkpoint)
+
+        # First run of the connector, fetch all repos and store in checkpoint
+        if checkpoint.cached_repo_ids is None:
+            repos = []
+            if self.repositories:
+                if "," in self.repositories:
+                    # Multiple repositories specified
+                    repos = self.get_github_repos(self.github_client)
+                else:
+                    # Single repository (backward compatibility)
+                    repos = [self.get_github_repo(self.github_client)]
+            else:
+                # All repositories
+                repos = self.get_all_repos(self.github_client)
+            if not repos:
+                checkpoint.has_more = False
+                return checkpoint
+
+            curr_repo = repos.pop()
+            checkpoint.cached_repo_ids = [repo.id for repo in repos]
+            checkpoint.cached_repo = SerializedRepository(
+                id=curr_repo.id,
+                headers=curr_repo.raw_headers,
+                raw_data=curr_repo.raw_data,
+            )
+            checkpoint.stage = GithubConnectorStage.PRS
+            checkpoint.curr_page = 0
+            # save checkpoint with repo ids retrieved
+            return checkpoint
+
+        if checkpoint.cached_repo is None:
+            raise ValueError("No repo saved in checkpoint")
+
+        # Deserialize the repository from the checkpoint
+        repo = deserialize_repository(checkpoint.cached_repo, self.github_client)
+
+        cursor_url_callback = make_cursor_url_callback(checkpoint)
+        repo_external_access: ExternalAccess | None = None
+        if include_permissions:
+            repo_external_access = get_external_access_permission(
+                repo, self.github_client
+            )
+        if self.include_prs and checkpoint.stage == GithubConnectorStage.PRS:
+            logging.info(f"Fetching PRs for repo: {repo.name}")
+
+            pr_batch = _get_batch_rate_limited(
+                self._pull_requests_func(repo),
+                checkpoint.curr_page,
+                checkpoint.cursor_url,
+                checkpoint.num_retrieved,
+                cursor_url_callback,
+                self.github_client,
+            )
+            checkpoint.curr_page += 1  # NOTE: not used for cursor-based fallback
+            done_with_prs = False
+            num_prs = 0
+            pr = None
+            print("start: ", start)
+            for pr in pr_batch:
+                num_prs += 1
+                print("-"*40)
+                print("PR name", pr.title)
+                print("updated at", pr.updated_at)
+                print("-"*40)
+                print("\n")
+                # we iterate backwards in time, so at this point we stop processing prs
+                if (
+                    start is not None
+                    and pr.updated_at
+                    and pr.updated_at.replace(tzinfo=timezone.utc) <= start
+                ):
+                    done_with_prs = True
+                    break
+                # Skip PRs updated after the end date
+                if (
+                    end is not None
+                    and pr.updated_at
+                    and pr.updated_at.replace(tzinfo=timezone.utc) > end
+                ):
+                    continue
+                try:
+                    yield _convert_pr_to_document(
+                        cast(PullRequest, pr), repo_external_access
+                    )
+                except Exception as e:
+                    error_msg = f"Error converting PR to document: {e}"
+                    logging.exception(error_msg)
+                    yield ConnectorFailure(
+                        failed_document=DocumentFailure(
+                            document_id=str(pr.id), document_link=pr.html_url
+                        ),
+                        failure_message=error_msg,
+                        exception=e,
+                    )
+                    continue
+
+            # If we reach this point with a cursor url in the checkpoint, we were using
+            # the fallback cursor-based pagination strategy. That strategy tries to get all
+            # PRs, so having curosr_url set means we are done with prs. However, we need to
+            # return AFTER the checkpoint reset to avoid infinite loops.
+
+            # if we found any PRs on the page and there are more PRs to get, return the checkpoint.
+            # In offset mode, while indexing without time constraints, the pr batch
+            # will be empty when we're done.
+            used_cursor = checkpoint.cursor_url is not None
+            if num_prs > 0 and not done_with_prs and not used_cursor:
+                return checkpoint
+
+            # if we went past the start date during the loop or there are no more
+            # prs to get, we move on to issues
+            checkpoint.stage = GithubConnectorStage.ISSUES
+            checkpoint.reset()
+
+            if used_cursor:
+                # save the checkpoint after changing stage; next run will continue from issues
+                return checkpoint
+
+        checkpoint.stage = GithubConnectorStage.ISSUES
+
+        if self.include_issues and checkpoint.stage == GithubConnectorStage.ISSUES:
+            logging.info(f"Fetching issues for repo: {repo.name}")
+
+            issue_batch = list(
+                _get_batch_rate_limited(
+                    self._issues_func(repo),
+                    checkpoint.curr_page,
+                    checkpoint.cursor_url,
+                    checkpoint.num_retrieved,
+                    cursor_url_callback,
+                    self.github_client,
+                )
+            )
+            checkpoint.curr_page += 1
+            done_with_issues = False
+            num_issues = 0
+            for issue in issue_batch:
+                num_issues += 1
+                issue = cast(Issue, issue)
+                # we iterate backwards in time, so at this point we stop processing prs
+                if (
+                    start is not None
+                    and issue.updated_at.replace(tzinfo=timezone.utc) <= start
+                ):
+                    done_with_issues = True
+                    break
+                # Skip PRs updated after the end date
+                if (
+                    end is not None
+                    and issue.updated_at.replace(tzinfo=timezone.utc) > end
+                ):
+                    continue
+
+                if issue.pull_request is not None:
+                    # PRs are handled separately
+                    continue
+
+                try:
+                    yield _convert_issue_to_document(issue, repo_external_access)
+                except Exception as e:
+                    error_msg = f"Error converting issue to document: {e}"
+                    logging.exception(error_msg)
+                    yield ConnectorFailure(
+                        failed_document=DocumentFailure(
+                            document_id=str(issue.id),
+                            document_link=issue.html_url,
+                        ),
+                        failure_message=error_msg,
+                        exception=e,
+                    )
+                    continue
+
+            # if we found any issues on the page, and we're not done, return the checkpoint.
+            # don't return if we're using cursor-based pagination to avoid infinite loops
+            if num_issues > 0 and not done_with_issues and not checkpoint.cursor_url:
+                return checkpoint
+
+            # if we went past the start date during the loop or there are no more
+            # issues to get, we move on to the next repo
+            checkpoint.stage = GithubConnectorStage.PRS
+            checkpoint.reset()
+
+        checkpoint.has_more = len(checkpoint.cached_repo_ids) > 0
+        if checkpoint.cached_repo_ids:
+            next_id = checkpoint.cached_repo_ids.pop()
+            next_repo = self.github_client.get_repo(next_id)
+            checkpoint.cached_repo = SerializedRepository(
+                id=next_id,
+                headers=next_repo.raw_headers,
+                raw_data=next_repo.raw_data,
+            )
+            checkpoint.stage = GithubConnectorStage.PRS
+            checkpoint.reset()
+
+        if checkpoint.cached_repo_ids:
+            logging.info(
+                f"{len(checkpoint.cached_repo_ids)} repos remaining (IDs: {checkpoint.cached_repo_ids})"
+            )
+        else:
+            logging.info("No more repos remaining")
+
+        return checkpoint
+
+    def _load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: GithubConnectorCheckpoint,
+        include_permissions: bool = False,
+    ) -> CheckpointOutput[GithubConnectorCheckpoint]:
+        start_datetime = datetime.fromtimestamp(start, tz=timezone.utc)
+        # add a day for timezone safety
+        end_datetime = datetime.fromtimestamp(end, tz=timezone.utc) + ONE_DAY
+
+        # Move start time back by 3 hours, since some Issues/PRs are getting dropped
+        # Could be due to delayed processing on GitHub side
+        # The non-updated issues since last poll will be shortcut-ed and not embedded
+        # adjusted_start_datetime = start_datetime - timedelta(hours=3)
+
+        adjusted_start_datetime = start_datetime
+
+        epoch = datetime.fromtimestamp(0, tz=timezone.utc)
+        if adjusted_start_datetime < epoch:
+            adjusted_start_datetime = epoch
+
+        return self._fetch_from_github(
+            checkpoint,
+            start=adjusted_start_datetime,
+            end=end_datetime,
+            include_permissions=include_permissions,
+        )
+
+    @override
+    def load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: GithubConnectorCheckpoint,
+    ) -> CheckpointOutput[GithubConnectorCheckpoint]:
+        return self._load_from_checkpoint(
+            start, end, checkpoint, include_permissions=False
+        )
+
+    @override
+    def load_from_checkpoint_with_perm_sync(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: GithubConnectorCheckpoint,
+    ) -> CheckpointOutput[GithubConnectorCheckpoint]:
+        return self._load_from_checkpoint(
+            start, end, checkpoint, include_permissions=True
+        )
+
+    def validate_connector_settings(self) -> None:
+        if self.github_client is None:
+            raise ConnectorMissingCredentialError("GitHub credentials not loaded.")
+
+        if not self.repo_owner:
+            raise ConnectorValidationError(
+                "Invalid connector settings: 'repo_owner' must be provided."
+            )
+
+        try:
+            if self.repositories:
+                if "," in self.repositories:
+                    # Multiple repositories specified
+                    repo_names = [name.strip() for name in self.repositories.split(",")]
+                    if not repo_names:
+                        raise ConnectorValidationError(
+                            "Invalid connector settings: No valid repository names provided."
+                        )
+
+                    # Validate at least one repository exists and is accessible
+                    valid_repos = False
+                    validation_errors = []
+
+                    for repo_name in repo_names:
+                        if not repo_name:
+                            continue
+
+                        try:
+                            test_repo = self.github_client.get_repo(
+                                f"{self.repo_owner}/{repo_name}"
+                            )
+                            logging.info(
+                                f"Successfully accessed repository: {self.repo_owner}/{repo_name}"
+                            )
+                            test_repo.get_contents("")
+                            valid_repos = True
+                            # If at least one repo is valid, we can proceed
+                            break
+                        except GithubException as e:
+                            validation_errors.append(
+                                f"Repository '{repo_name}': {e.data.get('message', str(e))}"
+                            )
+
+                    if not valid_repos:
+                        error_msg = (
+                            "None of the specified repositories could be accessed: "
+                        )
+                        error_msg += ", ".join(validation_errors)
+                        raise ConnectorValidationError(error_msg)
+                else:
+                    # Single repository (backward compatibility)
+                    test_repo = self.github_client.get_repo(
+                        f"{self.repo_owner}/{self.repositories}"
+                    )
+                    test_repo.get_contents("")
+            else:
+                # Try to get organization first
+                try:
+                    org = self.github_client.get_organization(self.repo_owner)
+                    total_count = org.get_repos().totalCount
+                    if total_count == 0:
+                        raise ConnectorValidationError(
+                            f"Found no repos for organization: {self.repo_owner}. "
+                            "Does the credential have the right scopes?"
+                        )
+                except GithubException as e:
+                    # Check for missing SSO
+                    MISSING_SSO_ERROR_MESSAGE = "You must grant your Personal Access token access to this organization".lower()
+                    if MISSING_SSO_ERROR_MESSAGE in str(e).lower():
+                        SSO_GUIDE_LINK = (
+                            "https://docs.github.com/en/enterprise-cloud@latest/authentication/"
+                            "authenticating-with-saml-single-sign-on/"
+                            "authorizing-a-personal-access-token-for-use-with-saml-single-sign-on"
+                        )
+                        raise ConnectorValidationError(
+                            f"Your GitHub token is missing authorization to access the "
+                            f"`{self.repo_owner}` organization. Please follow the guide to "
+                            f"authorize your token: {SSO_GUIDE_LINK}"
+                        )
+                    # If not an org, try as a user
+                    user = self.github_client.get_user(self.repo_owner)
+
+                    # Check if we can access any repos
+                    total_count = user.get_repos().totalCount
+                    if total_count == 0:
+                        raise ConnectorValidationError(
+                            f"Found no repos for user: {self.repo_owner}. "
+                            "Does the credential have the right scopes?"
+                        )
+
+        except RateLimitExceededException:
+            raise UnexpectedValidationError(
+                "Validation failed due to GitHub rate-limits being exceeded. Please try again later."
+            )
+
+        except GithubException as e:
+            if e.status == 401:
+                raise CredentialExpiredError(
+                    "GitHub credential appears to be invalid or expired (HTTP 401)."
+                )
+            elif e.status == 403:
+                raise InsufficientPermissionsError(
+                    "Your GitHub token does not have sufficient permissions for this repository (HTTP 403)."
+                )
+            elif e.status == 404:
+                if self.repositories:
+                    if "," in self.repositories:
+                        raise ConnectorValidationError(
+                            f"None of the specified GitHub repositories could be found for owner: {self.repo_owner}"
+                        )
+                    else:
+                        raise ConnectorValidationError(
+                            f"GitHub repository not found with name: {self.repo_owner}/{self.repositories}"
+                        )
+                else:
+                    raise ConnectorValidationError(
+                        f"GitHub user or organization not found: {self.repo_owner}"
+                    )
+            else:
+                raise ConnectorValidationError(
+                    f"Unexpected GitHub error (status={e.status}): {e.data}"
+                )
+
+        except Exception as exc:
+            raise Exception(
+                f"Unexpected error during GitHub settings validation: {exc}"
+            )
+
+    def validate_checkpoint_json(
+        self, checkpoint_json: str
+    ) -> GithubConnectorCheckpoint:
+        return GithubConnectorCheckpoint.model_validate_json(checkpoint_json)
+
+    def build_dummy_checkpoint(self) -> GithubConnectorCheckpoint:
+        return GithubConnectorCheckpoint(
+            stage=GithubConnectorStage.PRS, curr_page=0, has_more=True, num_retrieved=0
+        )
+
+
+if __name__ == "__main__":
+    # Initialize the connector
+    connector = GithubConnector(
+        repo_owner="EvoAgentX",
+        repositories="EvoAgentX",
+        include_issues=True,
+        include_prs=False,
+    )
+    connector.load_credentials(
+        {"github_access_token": "<Your_GitHub_Access_Token>"}
+    )
+
+    if connector.github_client:
+        get_external_access_permission(
+            connector.get_github_repos(connector.github_client).pop(),
+            connector.github_client,
+        )
+
+    # Create a time range from epoch to now
+    end_time = datetime.now(timezone.utc)
+    start_time = datetime.fromtimestamp(0, tz=timezone.utc)
+    time_range = (start_time, end_time)
+
+    # Initialize the runner with a batch size of 10
+    runner: ConnectorRunner[GithubConnectorCheckpoint] = ConnectorRunner(
+        connector, batch_size=10, include_permissions=False, time_range=time_range
+    )
+
+    # Get initial checkpoint
+    checkpoint = connector.build_dummy_checkpoint()
+
+    # Run the connector
+    while checkpoint.has_more:
+        for doc_batch, failure, next_checkpoint in runner.run(checkpoint):
+            if doc_batch:
+                print(f"Retrieved batch of {len(doc_batch)} documents")
+                for doc in doc_batch:
+                    print(f"Document: {doc.semantic_identifier}")
+            if failure:
+                print(f"Failure: {failure.failure_message}")
+            if next_checkpoint:
+                checkpoint = next_checkpoint

common/data_source/github/models.py

@@ -0,0 +1,17 @@
+from typing import Any
+
+from github import Repository
+from github.Requester import Requester
+from pydantic import BaseModel
+
+
+class SerializedRepository(BaseModel):
+    # id is part of the raw_data as well, just pulled out for convenience
+    id: int
+    headers: dict[str, str | int]
+    raw_data: dict[str, Any]
+
+    def to_Repository(self, requester: Requester) -> Repository.Repository:
+        return Repository.Repository(
+            requester, self.headers, self.raw_data, completed=True
+        )

common/data_source/github/rate_limit_utils.py

@@ -0,0 +1,24 @@
+import time
+import logging
+from datetime import datetime
+from datetime import timedelta
+from datetime import timezone
+
+from github import Github
+
+
+def sleep_after_rate_limit_exception(github_client: Github) -> None:
+    """
+    Sleep until the GitHub rate limit resets.
+
+    Args:
+        github_client: The GitHub client that hit the rate limit
+    """
+    sleep_time = github_client.get_rate_limit().core.reset.replace(
+        tzinfo=timezone.utc
+    ) - datetime.now(tz=timezone.utc)
+    sleep_time += timedelta(minutes=1)  # add an extra minute just to be safe
+    logging.info(
+        "Ran into Github rate-limit. Sleeping %s seconds.", sleep_time.seconds
+    )
+    time.sleep(sleep_time.total_seconds())

common/data_source/github/utils.py

@@ -0,0 +1,44 @@
+import logging
+
+from github import Github
+from github.Repository import Repository
+
+from common.data_source.models import ExternalAccess
+
+from .models import SerializedRepository
+
+
+def get_external_access_permission(
+    repo: Repository, github_client: Github
+) -> ExternalAccess:
+    """
+    Get the external access permission for a repository.
+    This functionality requires Enterprise Edition.
+    """
+    # RAGFlow doesn't implement the Onyx EE external-permissions system.
+    # Default to private/unknown permissions.
+    return ExternalAccess.empty()
+
+
+def deserialize_repository(
+    cached_repo: SerializedRepository, github_client: Github
+) -> Repository:
+    """
+    Deserialize a SerializedRepository back into a Repository object.
+    """
+    # Try to access the requester - different PyGithub versions may use different attribute names
+    try:
+        # Try to get the requester using getattr to avoid linter errors
+        requester = getattr(github_client, "_requester", None)
+        if requester is None:
+            requester = getattr(github_client, "_Github__requester", None)
+        if requester is None:
+            # If we can't find the requester attribute, we need to fall back to recreating the repo
+            raise AttributeError("Could not find requester attribute")
+
+        return cached_repo.to_Repository(requester)
+    except Exception as e:
+        # If all else fails, re-fetch the repo directly
+        logging.warning("Failed to deserialize repository: %s. Attempting to re-fetch.", e)
+        repo_id = cached_repo.id
+        return github_client.get_repo(repo_id)

common/data_source/gitlab_connector.py

@@ -0,0 +1,340 @@
+import fnmatch
+import itertools
+from collections import deque
+from collections.abc import Iterable
+from collections.abc import Iterator
+from datetime import datetime
+from datetime import timezone
+from typing import Any
+from typing import TypeVar
+import gitlab
+from gitlab.v4.objects import Project
+
+from common.data_source.config import DocumentSource, INDEX_BATCH_SIZE
+from common.data_source.exceptions import ConnectorMissingCredentialError
+from common.data_source.exceptions import ConnectorValidationError
+from common.data_source.exceptions import CredentialExpiredError
+from common.data_source.exceptions import InsufficientPermissionsError
+from common.data_source.exceptions import UnexpectedValidationError
+from common.data_source.interfaces import GenerateDocumentsOutput
+from common.data_source.interfaces import LoadConnector
+from common.data_source.interfaces import PollConnector
+from common.data_source.interfaces import SecondsSinceUnixEpoch
+from common.data_source.models import BasicExpertInfo
+from common.data_source.models import Document
+from common.data_source.utils import get_file_ext
+
+T = TypeVar("T")
+
+
+
+# List of directories/Files to exclude
+exclude_patterns = [
+    "logs",
+    ".github/",
+    ".gitlab/",
+    ".pre-commit-config.yaml",
+]
+
+
+def _batch_gitlab_objects(git_objs: Iterable[T], batch_size: int) -> Iterator[list[T]]:
+    it = iter(git_objs)
+    while True:
+        batch = list(itertools.islice(it, batch_size))
+        if not batch:
+            break
+        yield batch
+
+
+def get_author(author: Any) -> BasicExpertInfo:
+    return BasicExpertInfo(
+        display_name=author.get("name"),
+    )
+
+
+def _convert_merge_request_to_document(mr: Any) -> Document:
+    mr_text = mr.description or ""
+    doc = Document(
+        id=mr.web_url,
+        blob=mr_text,
+        source=DocumentSource.GITLAB,
+        semantic_identifier=mr.title,
+        extension=".md",
+        # updated_at is UTC time but is timezone unaware, explicitly add UTC
+        # as there is logic in indexing to prevent wrong timestamped docs
+        # due to local time discrepancies with UTC
+        doc_updated_at=mr.updated_at.replace(tzinfo=timezone.utc),
+        size_bytes=len(mr_text.encode("utf-8")),
+        primary_owners=[get_author(mr.author)],
+        metadata={"state": mr.state, "type": "MergeRequest", "web_url": mr.web_url},
+    )
+    return doc
+
+
+def _convert_issue_to_document(issue: Any) -> Document:
+    issue_text = issue.description or ""
+    doc = Document(
+        id=issue.web_url,
+        blob=issue_text,
+        source=DocumentSource.GITLAB,
+        semantic_identifier=issue.title,
+        extension=".md",
+        # updated_at is UTC time but is timezone unaware, explicitly add UTC
+        # as there is logic in indexing to prevent wrong timestamped docs
+        # due to local time discrepancies with UTC
+        doc_updated_at=issue.updated_at.replace(tzinfo=timezone.utc),
+        size_bytes=len(issue_text.encode("utf-8")),
+        primary_owners=[get_author(issue.author)],
+        metadata={
+            "state": issue.state,
+            "type": issue.type if issue.type else "Issue",
+            "web_url": issue.web_url,
+        },
+    )
+    return doc
+
+
+def _convert_code_to_document(
+    project: Project, file: Any, url: str, projectName: str, projectOwner: str
+) -> Document:
+    
+    # Dynamically get the default branch from the project object
+    default_branch = project.default_branch
+
+    # Fetch the file content using the correct branch
+    file_content_obj = project.files.get(
+        file_path=file["path"], ref=default_branch  # Use the default branch
+    )
+    # BoxConnector uses raw bytes for blob. Keep the same here.
+    file_content_bytes = file_content_obj.decode()
+    file_url = f"{url}/{projectOwner}/{projectName}/-/blob/{default_branch}/{file['path']}"
+
+    # Try to use the last commit timestamp for incremental sync.
+    # Falls back to "now" if the commit lookup fails.
+    last_commit_at = None
+    try:
+        # Query commit history for this file on the default branch.
+        commits = project.commits.list(
+            ref_name=default_branch,
+            path=file["path"],
+            per_page=1,
+        )
+        if commits:
+            # committed_date is ISO string like "2024-01-01T00:00:00.000+00:00"
+            committed_date = commits[0].committed_date
+            if isinstance(committed_date, str):
+                last_commit_at = datetime.strptime(
+                    committed_date, "%Y-%m-%dT%H:%M:%S.%f%z"
+                ).astimezone(timezone.utc)
+            elif isinstance(committed_date, datetime):
+                last_commit_at = committed_date.astimezone(timezone.utc)
+    except Exception:
+        last_commit_at = None
+
+    # Create and return a Document object
+    doc = Document(
+        # Use a stable ID so reruns don't create duplicates.
+        id=file_url,
+        blob=file_content_bytes,
+        source=DocumentSource.GITLAB,
+        semantic_identifier=file.get("name"),
+        extension=get_file_ext(file.get("name")),
+        doc_updated_at=last_commit_at or datetime.now(tz=timezone.utc),
+        size_bytes=len(file_content_bytes) if file_content_bytes is not None else 0,
+        primary_owners=[],  # Add owners if needed
+        metadata={
+            "type": "CodeFile",
+            "path": file.get("path"),
+            "ref": default_branch,
+            "project": f"{projectOwner}/{projectName}",
+            "web_url": file_url,
+        },
+    )
+    return doc
+
+
+def _should_exclude(path: str) -> bool:
+    """Check if a path matches any of the exclude patterns."""
+    return any(fnmatch.fnmatch(path, pattern) for pattern in exclude_patterns)
+
+
+class GitlabConnector(LoadConnector, PollConnector):
+    def __init__(
+        self,
+        project_owner: str,
+        project_name: str,
+        batch_size: int = INDEX_BATCH_SIZE,
+        state_filter: str = "all",
+        include_mrs: bool = True,
+        include_issues: bool = True,
+        include_code_files: bool = False,
+    ) -> None:
+        self.project_owner = project_owner
+        self.project_name = project_name
+        self.batch_size = batch_size
+        self.state_filter = state_filter
+        self.include_mrs = include_mrs
+        self.include_issues = include_issues
+        self.include_code_files = include_code_files
+        self.gitlab_client: gitlab.Gitlab | None = None
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        self.gitlab_client = gitlab.Gitlab(
+            credentials["gitlab_url"], private_token=credentials["gitlab_access_token"]
+        )
+        return None
+
+    def validate_connector_settings(self) -> None:
+        if self.gitlab_client is None:
+            raise ConnectorMissingCredentialError("GitLab")
+
+        try:
+            self.gitlab_client.auth()
+            self.gitlab_client.projects.get(
+                f"{self.project_owner}/{self.project_name}",
+                lazy=True,
+            )
+
+        except gitlab.exceptions.GitlabAuthenticationError as e:
+            raise CredentialExpiredError(
+                "Invalid or expired GitLab credentials."
+            ) from e
+
+        except gitlab.exceptions.GitlabAuthorizationError as e:
+            raise InsufficientPermissionsError(
+                "Insufficient permissions to access GitLab resources."
+            ) from e
+
+        except gitlab.exceptions.GitlabGetError as e:
+            raise ConnectorValidationError(
+                "GitLab project not found or not accessible."
+            ) from e
+
+        except Exception as e:
+            raise UnexpectedValidationError(
+                f"Unexpected error while validating GitLab settings: {e}"
+            ) from e
+
+    def _fetch_from_gitlab(
+        self, start: datetime | None = None, end: datetime | None = None
+    ) -> GenerateDocumentsOutput:
+        if self.gitlab_client is None:
+            raise ConnectorMissingCredentialError("Gitlab")
+        project: Project = self.gitlab_client.projects.get(
+            f"{self.project_owner}/{self.project_name}"
+        )
+
+        start_utc = start.astimezone(timezone.utc) if start else None
+        end_utc = end.astimezone(timezone.utc) if end else None
+
+        # Fetch code files
+        if self.include_code_files:
+            # Fetching using BFS as project.report_tree with recursion causing slow load
+            queue = deque([""])  # Start with the root directory
+            while queue:
+                current_path = queue.popleft()
+                files = project.repository_tree(path=current_path, all=True)
+                for file_batch in _batch_gitlab_objects(files, self.batch_size):
+                    code_doc_batch: list[Document] = []
+                    for file in file_batch:
+                        if _should_exclude(file["path"]):
+                            continue
+
+                        if file["type"] == "blob":
+
+                            doc = _convert_code_to_document(
+                                project,
+                                file,
+                                self.gitlab_client.url,
+                                self.project_name,
+                                self.project_owner,
+                            )
+
+                            # Apply incremental window filtering for code files too.
+                            if start_utc is not None and doc.doc_updated_at <= start_utc:
+                                continue
+                            if end_utc is not None and doc.doc_updated_at > end_utc:
+                                continue
+
+                            code_doc_batch.append(doc)
+                        elif file["type"] == "tree":
+                            queue.append(file["path"])
+
+                    if code_doc_batch:
+                        yield code_doc_batch
+
+        if self.include_mrs:
+            merge_requests = project.mergerequests.list(
+                state=self.state_filter,
+                order_by="updated_at",
+                sort="desc",
+                iterator=True,
+            )
+
+            for mr_batch in _batch_gitlab_objects(merge_requests, self.batch_size):
+                mr_doc_batch: list[Document] = []
+                for mr in mr_batch:
+                    mr.updated_at = datetime.strptime(
+                        mr.updated_at, "%Y-%m-%dT%H:%M:%S.%f%z"
+                    )
+                    if start_utc is not None and mr.updated_at <= start_utc:
+                        yield mr_doc_batch
+                        return
+                    if end_utc is not None and mr.updated_at > end_utc:
+                        continue
+                    mr_doc_batch.append(_convert_merge_request_to_document(mr))
+                yield mr_doc_batch
+
+        if self.include_issues:
+            issues = project.issues.list(state=self.state_filter, iterator=True)
+
+            for issue_batch in _batch_gitlab_objects(issues, self.batch_size):
+                issue_doc_batch: list[Document] = []
+                for issue in issue_batch:
+                    issue.updated_at = datetime.strptime(
+                        issue.updated_at, "%Y-%m-%dT%H:%M:%S.%f%z"
+                    )
+                    # Avoid re-syncing the last-seen item.
+                    if start_utc is not None and issue.updated_at <= start_utc:
+                        yield issue_doc_batch
+                        return
+                    if end_utc is not None and issue.updated_at > end_utc:
+                        continue
+                    issue_doc_batch.append(_convert_issue_to_document(issue))
+                yield issue_doc_batch
+
+    def load_from_state(self) -> GenerateDocumentsOutput:
+        return self._fetch_from_gitlab()
+
+    def poll_source(
+        self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch
+    ) -> GenerateDocumentsOutput:
+        start_datetime = datetime.fromtimestamp(start, tz=timezone.utc)
+        end_datetime = datetime.fromtimestamp(end, tz=timezone.utc)
+        return self._fetch_from_gitlab(start_datetime, end_datetime)
+
+
+if __name__ == "__main__":
+    import os
+
+    connector = GitlabConnector(
+        # gitlab_url="https://gitlab.com/api/v4",
+        project_owner=os.environ["PROJECT_OWNER"],
+        project_name=os.environ["PROJECT_NAME"],
+        batch_size=INDEX_BATCH_SIZE,
+        state_filter="all",
+        include_mrs=True,
+        include_issues=True,
+        include_code_files=True,
+    )
+
+    connector.load_credentials(
+        {
+            "gitlab_access_token": os.environ["GITLAB_ACCESS_TOKEN"],
+            "gitlab_url": os.environ["GITLAB_URL"],
+        }
+    )
+    document_batches = connector.load_from_state()
+    for f in document_batches:
+        print("Batch:", f)
+    print("Finished loading from state.")

common/data_source/gmail_connector.py

@@ -8,10 +8,10 @@ from common.data_source.config import INDEX_BATCH_SIZE, SLIM_BATCH_SIZE, Documen
 from common.data_source.google_util.auth import get_google_creds
 from common.data_source.google_util.constant import DB_CREDENTIALS_PRIMARY_ADMIN_KEY, MISSING_SCOPES_ERROR_STR, SCOPE_INSTRUCTIONS, USER_FIELDS
 from common.data_source.google_util.resource import get_admin_service, get_gmail_service
-from common.data_source.google_util.util import _execute_single_retrieval, execute_paginated_retrieval, sanitize_filename, clean_string
+from common.data_source.google_util.util import _execute_single_retrieval, execute_paginated_retrieval, clean_string
 from common.data_source.interfaces import LoadConnector, PollConnector, SecondsSinceUnixEpoch, SlimConnectorWithPermSync
 from common.data_source.models import BasicExpertInfo, Document, ExternalAccess, GenerateDocumentsOutput, GenerateSlimDocumentOutput, SlimDocument, TextSection
-from common.data_source.utils import build_time_range_query, clean_email_and_extract_name, get_message_body, is_mail_service_disabled_error, gmail_time_str_to_utc
+from common.data_source.utils import build_time_range_query, clean_email_and_extract_name, get_message_body, is_mail_service_disabled_error, gmail_time_str_to_utc, sanitize_filename
 
 # Constants for Gmail API fields
 THREAD_LIST_FIELDS = "nextPageToken, threads(id)"

common/data_source/google_util/util.py

@@ -191,43 +191,6 @@ def get_credentials_from_env(email: str, oauth: bool = False, source="drive") ->
         DB_CREDENTIALS_AUTHENTICATION_METHOD: "uploaded",
     }
 
-def sanitize_filename(name: str) -> str:
-    """
-    Soft sanitize for MinIO/S3:
-    - Replace only prohibited characters with a space.
-    - Preserve readability (no ugly underscores).
-    - Collapse multiple spaces.
-    """
-    if name is None:
-        return "file.txt"
-
-    name = str(name).strip()
-
-    # Characters that MUST NOT appear in S3/MinIO object keys
-    # Replace them with a space (not underscore)
-    forbidden = r'[\\\?\#\%\*\:\|\<\>"]'
-    name = re.sub(forbidden, " ", name)
-
-    # Replace slashes "/" (S3 interprets as folder) with space
-    name = name.replace("/", " ")
-
-    # Collapse multiple spaces into one
-    name = re.sub(r"\s+", " ", name)
-
-    # Trim both ends
-    name = name.strip()
-
-    # Enforce reasonable max length
-    if len(name) > 200:
-        base, ext = os.path.splitext(name)
-        name = base[:180].rstrip() + ext
-
-    # Ensure there is an extension (your original logic)
-    if not os.path.splitext(name)[1]:
-        name += ".txt"
-
-    return name
-
 
 def clean_string(text: str | None) -> str | None:
     """

common/data_source/imap_connector.py

@@ -0,0 +1,724 @@
+import copy
+import email
+from email.header import decode_header
+import imaplib
+import logging
+import os
+import re
+from datetime import datetime, timedelta
+from datetime import timezone
+from email.message import Message
+from email.utils import collapse_rfc2231_value, parseaddr
+from enum import Enum
+from typing import Any
+from typing import cast
+
+import bs4
+from pydantic import BaseModel
+
+from common.data_source.config import IMAP_CONNECTOR_SIZE_THRESHOLD, DocumentSource
+from common.data_source.interfaces import CheckpointOutput, CheckpointedConnectorWithPermSync, CredentialsConnector, CredentialsProviderInterface
+from common.data_source.models import BasicExpertInfo, ConnectorCheckpoint, Document, ExternalAccess, SecondsSinceUnixEpoch
+
+_DEFAULT_IMAP_PORT_NUMBER = int(os.environ.get("IMAP_PORT", 993))
+_IMAP_OKAY_STATUS = "OK"
+_PAGE_SIZE = 100
+_USERNAME_KEY = "imap_username"
+_PASSWORD_KEY = "imap_password"
+
+class Header(str, Enum):
+    SUBJECT_HEADER = "subject"
+    FROM_HEADER = "from"
+    TO_HEADER = "to"
+    CC_HEADER = "cc"
+    DELIVERED_TO_HEADER = (
+        "Delivered-To"  # Used in mailing lists instead of the "to" header.
+    )
+    DATE_HEADER = "date"
+    MESSAGE_ID_HEADER = "Message-ID"
+
+
+class EmailHeaders(BaseModel):
+    """
+    Model for email headers extracted from IMAP messages.
+    """
+
+    id: str
+    subject: str
+    sender: str
+    recipients: str | None
+    cc: str | None
+    date: datetime
+
+    @classmethod
+    def from_email_msg(cls, email_msg: Message) -> "EmailHeaders":
+        def _decode(header: str, default: str | None = None) -> str | None:
+            value = email_msg.get(header, default)
+            if not value:
+                return None
+
+            decoded_fragments = decode_header(value)
+            decoded_strings: list[str] = []
+
+            for decoded_value, encoding in decoded_fragments:
+                if isinstance(decoded_value, bytes):
+                    try:
+                        decoded_strings.append(
+                            decoded_value.decode(encoding or "utf-8", errors="replace")
+                        )
+                    except LookupError:
+                        decoded_strings.append(
+                            decoded_value.decode("utf-8", errors="replace")
+                        )
+                elif isinstance(decoded_value, str):
+                    decoded_strings.append(decoded_value)
+                else:
+                    decoded_strings.append(str(decoded_value))
+
+            return "".join(decoded_strings)
+
+        def _parse_date(date_str: str | None) -> datetime | None:
+            if not date_str:
+                return None
+            try:
+                return email.utils.parsedate_to_datetime(date_str)
+            except (TypeError, ValueError):
+                return None
+
+        message_id = _decode(header=Header.MESSAGE_ID_HEADER)
+        if not message_id:
+            message_id = f"<generated-{uuid.uuid4()}@imap.local>"
+        # It's possible for the subject line to not exist or be an empty string.
+        subject = _decode(header=Header.SUBJECT_HEADER) or "Unknown Subject"
+        from_ = _decode(header=Header.FROM_HEADER)
+        to = _decode(header=Header.TO_HEADER)
+        if not to:
+            to = _decode(header=Header.DELIVERED_TO_HEADER)
+        cc = _decode(header=Header.CC_HEADER)
+        date_str = _decode(header=Header.DATE_HEADER)
+        date = _parse_date(date_str=date_str)
+
+        if not date:
+            date = datetime.now(tz=timezone.utc)
+
+        # If any of the above are `None`, model validation will fail.
+        # Therefore, no guards (i.e.: `if <header> is None: raise RuntimeError(..)`) were written.
+        return cls.model_validate(
+            {
+                "id": message_id,
+                "subject": subject,
+                "sender": from_,
+                "recipients": to,
+                "cc": cc,
+                "date": date,
+            }
+        )
+
+class CurrentMailbox(BaseModel):
+    mailbox: str
+    todo_email_ids: list[str]
+
+
+# An email has a list of mailboxes.
+# Each mailbox has a list of email-ids inside of it.
+#
+# Usage:
+# To use this checkpointer, first fetch all the mailboxes.
+# Then, pop a mailbox and fetch all of its email-ids.
+# Then, pop each email-id and fetch its content (and parse it, etc..).
+# When you have popped all email-ids for this mailbox, pop the next mailbox and repeat the above process until you're done.
+#
+# For initial checkpointing, set both fields to `None`.
+class ImapCheckpoint(ConnectorCheckpoint):
+    todo_mailboxes: list[str] | None = None
+    current_mailbox: CurrentMailbox | None = None
+
+
+class LoginState(str, Enum):
+    LoggedIn = "logged_in"
+    LoggedOut = "logged_out"
+
+
+class ImapConnector(
+    CredentialsConnector,
+    CheckpointedConnectorWithPermSync,
+):
+    def __init__(
+        self,
+        host: str,
+        port: int = _DEFAULT_IMAP_PORT_NUMBER,
+        mailboxes: list[str] | None = None,
+    ) -> None:
+        self._host = host
+        self._port = port
+        self._mailboxes = mailboxes
+        self._credentials: dict[str, Any] | None = None
+
+    @property
+    def credentials(self) -> dict[str, Any]:
+        if not self._credentials:
+            raise RuntimeError(
+                "Credentials have not been initialized; call `set_credentials_provider` first"
+            )
+        return self._credentials
+
+    def _get_mail_client(self) -> imaplib.IMAP4_SSL:
+        """
+        Returns a new `imaplib.IMAP4_SSL` instance.
+
+        The `imaplib.IMAP4_SSL` object is supposed to be an "ephemeral" object; it's not something that you can login,
+        logout, then log back into again. I.e., the following will fail:
+
+        ```py
+        mail_client.login(..)
+        mail_client.logout();
+        mail_client.login(..)
+        ```
+
+        Therefore, you need a fresh, new instance in order to operate with IMAP. This function gives one to you.
+
+        # Notes
+        This function will throw an error if the credentials have not yet been set.
+        """
+
+        def get_or_raise(name: str) -> str:
+            value = self.credentials.get(name)
+            if not value:
+                raise RuntimeError(f"Credential item {name=} was not found")
+            if not isinstance(value, str):
+                raise RuntimeError(
+                    f"Credential item {name=} must be of type str, instead received {type(name)=}"
+                )
+            return value
+
+        username = get_or_raise(_USERNAME_KEY)
+        password = get_or_raise(_PASSWORD_KEY)
+
+        mail_client = imaplib.IMAP4_SSL(host=self._host, port=self._port)
+        status, _data = mail_client.login(user=username, password=password)
+
+        if status != _IMAP_OKAY_STATUS:
+            raise RuntimeError(f"Failed to log into imap server; {status=}")
+
+        return mail_client
+
+    def _load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: ImapCheckpoint,
+        include_perm_sync: bool,
+    ) -> CheckpointOutput[ImapCheckpoint]:
+        checkpoint = cast(ImapCheckpoint, copy.deepcopy(checkpoint))
+        checkpoint.has_more = True
+
+        mail_client = self._get_mail_client()
+
+        if checkpoint.todo_mailboxes is None:
+            # This is the dummy checkpoint.
+            # Fill it with mailboxes first.
+            if self._mailboxes:
+                checkpoint.todo_mailboxes = _sanitize_mailbox_names(self._mailboxes)
+            else:
+                fetched_mailboxes = _fetch_all_mailboxes_for_email_account(
+                    mail_client=mail_client
+                )
+                if not fetched_mailboxes:
+                    raise RuntimeError(
+                        "Failed to find any mailboxes for this email account"
+                    )
+                checkpoint.todo_mailboxes = _sanitize_mailbox_names(fetched_mailboxes)
+
+            return checkpoint
+
+        if (
+            not checkpoint.current_mailbox
+            or not checkpoint.current_mailbox.todo_email_ids
+        ):
+            if not checkpoint.todo_mailboxes:
+                checkpoint.has_more = False
+                return checkpoint
+
+            mailbox = checkpoint.todo_mailboxes.pop()
+            email_ids = _fetch_email_ids_in_mailbox(
+                mail_client=mail_client,
+                mailbox=mailbox,
+                start=start,
+                end=end,
+            )
+            checkpoint.current_mailbox = CurrentMailbox(
+                mailbox=mailbox,
+                todo_email_ids=email_ids,
+            )
+
+        _select_mailbox(
+            mail_client=mail_client, mailbox=checkpoint.current_mailbox.mailbox
+        )
+        current_todos = cast(
+            list, copy.deepcopy(checkpoint.current_mailbox.todo_email_ids[:_PAGE_SIZE])
+        )
+        checkpoint.current_mailbox.todo_email_ids = (
+            checkpoint.current_mailbox.todo_email_ids[_PAGE_SIZE:]
+        )
+
+        for email_id in current_todos:
+            email_msg = _fetch_email(mail_client=mail_client, email_id=email_id)
+            if not email_msg:
+                logging.warning(f"Failed to fetch message {email_id=}; skipping")
+                continue
+
+            email_headers = EmailHeaders.from_email_msg(email_msg=email_msg)
+            msg_dt = email_headers.date
+            if msg_dt.tzinfo is None:
+                msg_dt = msg_dt.replace(tzinfo=timezone.utc)
+            else:
+                msg_dt = msg_dt.astimezone(timezone.utc)
+
+            start_dt = datetime.fromtimestamp(start, tz=timezone.utc)
+            end_dt = datetime.fromtimestamp(end, tz=timezone.utc)
+
+            if not (start_dt < msg_dt <= end_dt):
+                continue
+
+            email_doc = _convert_email_headers_and_body_into_document(
+                email_msg=email_msg,
+                email_headers=email_headers,
+                include_perm_sync=include_perm_sync,
+            )
+            yield email_doc
+            attachments = extract_attachments(email_msg)
+            for att in attachments:
+                yield attachment_to_document(email_doc, att, email_headers)
+
+        return checkpoint
+
+    # impls for BaseConnector
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        self._credentials = credentials
+        return None
+
+    def validate_connector_settings(self) -> None:
+        self._get_mail_client()
+
+    # impls for CredentialsConnector
+
+    def set_credentials_provider(
+        self, credentials_provider: CredentialsProviderInterface
+    ) -> None:
+        self._credentials = credentials_provider.get_credentials()
+
+    # impls for CheckpointedConnector
+
+    def load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: ImapCheckpoint,
+    ) -> CheckpointOutput[ImapCheckpoint]:
+        return self._load_from_checkpoint(
+            start=start, end=end, checkpoint=checkpoint, include_perm_sync=False
+        )
+
+    def build_dummy_checkpoint(self) -> ImapCheckpoint:
+        return ImapCheckpoint(has_more=True)
+
+    def validate_checkpoint_json(self, checkpoint_json: str) -> ImapCheckpoint:
+        return ImapCheckpoint.model_validate_json(json_data=checkpoint_json)
+
+    # impls for CheckpointedConnectorWithPermSync
+
+    def load_from_checkpoint_with_perm_sync(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: ImapCheckpoint,
+    ) -> CheckpointOutput[ImapCheckpoint]:
+        return self._load_from_checkpoint(
+            start=start, end=end, checkpoint=checkpoint, include_perm_sync=True
+        )
+
+
+def _fetch_all_mailboxes_for_email_account(mail_client: imaplib.IMAP4_SSL) -> list[str]:
+    status, mailboxes_data = mail_client.list('""', "*")
+    if status != _IMAP_OKAY_STATUS:
+        raise RuntimeError(f"Failed to fetch mailboxes; {status=}")
+
+    mailboxes = []
+
+    for mailboxes_raw in mailboxes_data:
+        if isinstance(mailboxes_raw, bytes):
+            mailboxes_str = mailboxes_raw.decode()
+        elif isinstance(mailboxes_raw, str):
+            mailboxes_str = mailboxes_raw
+        else:
+            logging.warning(
+                f"Expected the mailbox data to be of type str, instead got {type(mailboxes_raw)=} {mailboxes_raw}; skipping"
+            )
+            continue
+
+        # The mailbox LIST response output can be found here:
+        # https://www.rfc-editor.org/rfc/rfc3501.html#section-7.2.2
+        #
+        # The general format is:
+        # `(<name-attributes>) <hierarchy-delimiter> <mailbox-name>`
+        #
+        # The below regex matches on that pattern; from there, we select the 3rd match (index 2), which is the mailbox-name.
+        match = re.match(r'\([^)]*\)\s+"([^"]+)"\s+"?(.+?)"?$', mailboxes_str)
+        if not match:
+            logging.warning(
+                f"Invalid mailbox-data formatting structure: {mailboxes_str=}; skipping"
+            )
+            continue
+
+        mailbox = match.group(2)
+        mailboxes.append(mailbox)
+    if not mailboxes:
+        logging.warning(
+            "No mailboxes parsed from LIST response; falling back to INBOX"
+        )
+        return ["INBOX"]
+
+    return mailboxes
+
+
+def _select_mailbox(mail_client: imaplib.IMAP4_SSL, mailbox: str) -> bool:
+    try:
+        status, _ = mail_client.select(mailbox=mailbox, readonly=True)
+        if status != _IMAP_OKAY_STATUS:
+            return False
+        return True
+    except Exception:
+        return False
+
+
+def _fetch_email_ids_in_mailbox(
+    mail_client: imaplib.IMAP4_SSL,
+    mailbox: str,
+    start: SecondsSinceUnixEpoch,
+    end: SecondsSinceUnixEpoch,
+) -> list[str]:
+    if not _select_mailbox(mail_client, mailbox):
+        logging.warning(f"Skip mailbox: {mailbox}")
+        return []
+
+    start_dt = datetime.fromtimestamp(start, tz=timezone.utc)
+    end_dt = datetime.fromtimestamp(end, tz=timezone.utc) + timedelta(days=1)
+
+    start_str = start_dt.strftime("%d-%b-%Y")
+    end_str = end_dt.strftime("%d-%b-%Y")
+    search_criteria = f'(SINCE "{start_str}" BEFORE "{end_str}")'
+
+    status, email_ids_byte_array = mail_client.search(None, search_criteria)
+
+    if status != _IMAP_OKAY_STATUS or not email_ids_byte_array:
+        raise RuntimeError(f"Failed to fetch email ids; {status=}")
+
+    email_ids: bytes = email_ids_byte_array[0]
+
+    return [email_id.decode() for email_id in email_ids.split()]
+
+
+def _fetch_email(mail_client: imaplib.IMAP4_SSL, email_id: str) -> Message | None:
+    status, msg_data = mail_client.fetch(message_set=email_id, message_parts="(RFC822)")
+    if status != _IMAP_OKAY_STATUS or not msg_data:
+        return None
+
+    data = msg_data[0]
+    if not isinstance(data, tuple):
+        raise RuntimeError(
+            f"Message data should be a tuple; instead got a {type(data)=} {data=}"
+        )
+
+    _, raw_email = data
+    return email.message_from_bytes(raw_email)
+
+
+def _convert_email_headers_and_body_into_document(
+    email_msg: Message,
+    email_headers: EmailHeaders,
+    include_perm_sync: bool,
+) -> Document:
+    sender_name, sender_addr = _parse_singular_addr(raw_header=email_headers.sender)
+    to_addrs = (
+        _parse_addrs(email_headers.recipients)
+        if email_headers.recipients
+        else []
+    )
+    cc_addrs = (
+        _parse_addrs(email_headers.cc)
+        if email_headers.cc
+        else []
+    )
+    all_participants = to_addrs + cc_addrs
+
+    expert_info_map = {
+        recipient_addr: BasicExpertInfo(
+            display_name=recipient_name, email=recipient_addr
+        )
+        for recipient_name, recipient_addr in all_participants
+    }
+    if sender_addr not in expert_info_map:
+        expert_info_map[sender_addr] = BasicExpertInfo(
+            display_name=sender_name, email=sender_addr
+        )
+
+    email_body = _parse_email_body(email_msg=email_msg, email_headers=email_headers)
+    primary_owners = list(expert_info_map.values())
+    external_access = (
+        ExternalAccess(
+            external_user_emails=set(expert_info_map.keys()),
+            external_user_group_ids=set(),
+            is_public=False,
+        )
+        if include_perm_sync
+        else None
+    )
+    return Document(
+        id=email_headers.id,
+        title=email_headers.subject,
+        blob=email_body,
+        size_bytes=len(email_body),
+        semantic_identifier=email_headers.subject,
+        metadata={},
+        extension='.txt',
+        doc_updated_at=email_headers.date,
+        source=DocumentSource.IMAP,
+        primary_owners=primary_owners,
+        external_access=external_access,
+    )
+
+def extract_attachments(email_msg: Message, max_bytes: int = IMAP_CONNECTOR_SIZE_THRESHOLD):
+    attachments = []
+
+    if not email_msg.is_multipart():
+        return attachments
+
+    for part in email_msg.walk():
+        if part.get_content_maintype() == "multipart":
+            continue
+
+        disposition = (part.get("Content-Disposition") or "").lower()
+        filename = part.get_filename()
+
+        if not (
+            disposition.startswith("attachment")
+            or (disposition.startswith("inline") and filename)
+        ):
+            continue
+
+        payload = part.get_payload(decode=True)
+        if not payload:
+            continue
+
+        if len(payload) > max_bytes:
+            continue
+
+        attachments.append({
+            "filename": filename or "attachment.bin",
+            "content_type": part.get_content_type(),
+            "content_bytes": payload,
+            "size_bytes": len(payload),
+        })
+
+    return attachments
+
+def decode_mime_filename(raw: str | None) -> str | None:
+    if not raw:
+        return None
+
+    try:
+        raw = collapse_rfc2231_value(raw)
+    except Exception:
+        pass
+
+    parts = decode_header(raw)
+    decoded = []
+
+    for value, encoding in parts:
+        if isinstance(value, bytes):
+            decoded.append(value.decode(encoding or "utf-8", errors="replace"))
+        else:
+            decoded.append(value)
+
+    return "".join(decoded)
+
+def attachment_to_document(
+    parent_doc: Document,
+    att: dict,
+    email_headers: EmailHeaders,
+):
+    raw_filename = att["filename"]
+    filename = decode_mime_filename(raw_filename) or "attachment.bin"
+    ext = "." + filename.split(".")[-1] if "." in filename else ""
+
+    return Document(
+        id=f"{parent_doc.id}#att:{filename}",
+        source=DocumentSource.IMAP,
+        semantic_identifier=filename,
+        extension=ext,
+        blob=att["content_bytes"],
+        size_bytes=att["size_bytes"],
+        doc_updated_at=email_headers.date,
+        primary_owners=parent_doc.primary_owners,
+        metadata={
+            "parent_email_id": parent_doc.id,
+            "parent_subject": email_headers.subject,
+            "attachment_filename": filename,
+            "attachment_content_type": att["content_type"],
+        },
+    )
+
+def _parse_email_body(
+    email_msg: Message,
+    email_headers: EmailHeaders,
+) -> str:
+    body = None
+    for part in email_msg.walk():
+        if part.is_multipart():
+            # Multipart parts are *containers* for other parts, not the actual content itself.
+            # Therefore, we skip until we find the individual parts instead.
+            continue
+
+        charset = part.get_content_charset() or "utf-8"
+
+        try:
+            raw_payload = part.get_payload(decode=True)
+            if not isinstance(raw_payload, bytes):
+                logging.warning(
+                    "Payload section from email was expected to be an array of bytes, instead got "
+                    f"{type(raw_payload)=}, {raw_payload=}"
+                )
+                continue
+            body = raw_payload.decode(charset)
+            break
+        except (UnicodeDecodeError, LookupError) as e:
+            logging.warning(f"Could not decode part with charset {charset}. Error: {e}")
+            continue
+
+    if not body:
+        logging.warning(
+            f"Email with {email_headers.id=} has an empty body; returning an empty string"
+        )
+        return ""
+
+    soup = bs4.BeautifulSoup(markup=body, features="html.parser")
+
+    return " ".join(str_section for str_section in soup.stripped_strings)
+
+
+def _sanitize_mailbox_names(mailboxes: list[str]) -> list[str]:
+    """
+    Mailboxes with special characters in them must be enclosed by double-quotes, as per the IMAP protocol.
+    Just to be safe, we wrap *all* mailboxes with double-quotes.
+    """
+    return [f'"{mailbox}"' for mailbox in mailboxes if mailbox]
+
+
+def _parse_addrs(raw_header: str) -> list[tuple[str, str]]:
+    addrs = raw_header.split(",")
+    name_addr_pairs = [parseaddr(addr=addr) for addr in addrs if addr]
+    return [(name, addr) for name, addr in name_addr_pairs if addr]
+
+
+def _parse_singular_addr(raw_header: str) -> tuple[str, str]:
+    addrs = _parse_addrs(raw_header=raw_header)
+    if not addrs:
+        return ("Unknown", "unknown@example.com")
+    elif len(addrs) >= 2:
+        raise RuntimeError(
+            f"Expected a singular address, but instead got multiple; {raw_header=} {addrs=}"
+        )
+
+    return addrs[0]
+
+
+if __name__ == "__main__":
+    import time
+    import uuid
+    from types import TracebackType
+    from common.data_source.utils import load_all_docs_from_checkpoint_connector
+
+
+    class OnyxStaticCredentialsProvider(
+        CredentialsProviderInterface["OnyxStaticCredentialsProvider"]
+    ):
+        """Implementation (a very simple one!) to handle static credentials."""
+
+        def __init__(
+            self,
+            tenant_id: str | None,
+            connector_name: str,
+            credential_json: dict[str, Any],
+        ):
+            self._tenant_id = tenant_id
+            self._connector_name = connector_name
+            self._credential_json = credential_json
+
+            self._provider_key = str(uuid.uuid4())
+
+        def __enter__(self) -> "OnyxStaticCredentialsProvider":
+            return self
+
+        def __exit__(
+            self,
+            exc_type: type[BaseException] | None,
+            exc_value: BaseException | None,
+            traceback: TracebackType | None,
+        ) -> None:
+            pass
+
+        def get_tenant_id(self) -> str | None:
+            return self._tenant_id
+
+        def get_provider_key(self) -> str:
+            return self._provider_key
+
+        def get_credentials(self) -> dict[str, Any]:
+            return self._credential_json
+
+        def set_credentials(self, credential_json: dict[str, Any]) -> None:
+            self._credential_json = credential_json
+
+        def is_dynamic(self) -> bool:
+            return False
+    # from tests.daily.connectors.utils import load_all_docs_from_checkpoint_connector
+    # from onyx.connectors.credentials_provider import OnyxStaticCredentialsProvider
+
+    host = os.environ.get("IMAP_HOST")
+    mailboxes_str = os.environ.get("IMAP_MAILBOXES","INBOX")
+    username = os.environ.get("IMAP_USERNAME")
+    password = os.environ.get("IMAP_PASSWORD")
+
+    mailboxes = (
+        [mailbox.strip() for mailbox in mailboxes_str.split(",")]
+        if mailboxes_str
+        else []
+    )
+
+    if not host:
+        raise RuntimeError("`IMAP_HOST` must be set")
+
+    imap_connector = ImapConnector(
+        host=host,
+        mailboxes=mailboxes,
+    )
+
+    imap_connector.set_credentials_provider(
+        OnyxStaticCredentialsProvider(
+            tenant_id=None,
+            connector_name=DocumentSource.IMAP,
+            credential_json={
+                _USERNAME_KEY: username,
+                _PASSWORD_KEY: password,
+            },
+        )
+    )
+    END = time.time()
+    START = END - 1 * 24 * 60 * 60
+    for doc in load_all_docs_from_checkpoint_connector(
+        connector=imap_connector,
+        start=START,
+        end=END,
+    ):
+        print(doc.id,doc.extension)

common/data_source/interfaces.py

@@ -5,7 +5,7 @@ from abc import ABC, abstractmethod
 from enum import IntFlag, auto
 from types import TracebackType
 from typing import Any, Dict, Generator, TypeVar, Generic, Callable, TypeAlias
-
+from collections.abc import Iterator
 from anthropic import BaseModel
 
 from common.data_source.models import (
@@ -16,6 +16,7 @@ from common.data_source.models import (
     SecondsSinceUnixEpoch, GenerateSlimDocumentOutput
 )
 
+GenerateDocumentsOutput = Iterator[list[Document]]
 
 class LoadConnector(ABC):
     """Load connector interface"""
@@ -236,16 +237,13 @@ class BaseConnector(abc.ABC, Generic[CT]):
 
     def validate_perm_sync(self) -> None:
         """
-        Don't override this; add a function to perm_sync_valid.py in the ee package
-        to do permission sync validation
+        Permission-sync validation hook.
+
+        RAGFlow doesn't ship the Onyx EE permission-sync validation package.
+        Connectors that support permission sync should override
+        `validate_connector_settings()` as needed.
         """
-        """
-        validate_connector_settings_fn = fetch_ee_implementation_or_noop(
-            "onyx.connectors.perm_sync_valid",
-            "validate_perm_sync",
-            noop_return_value=None,
-        )
-        validate_connector_settings_fn(self)"""
+        return None
 
     def set_allow_images(self, value: bool) -> None:
         """Implement if the underlying connector wants to skip/allow image downloading
@@ -344,6 +342,17 @@ class CheckpointOutputWrapper(Generic[CT]):
         yield None, None, self.next_checkpoint
 
 
+class CheckpointedConnectorWithPermSyncGH(CheckpointedConnector[CT]):
+    @abc.abstractmethod
+    def load_from_checkpoint_with_perm_sync(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: CT,
+    ) -> CheckpointOutput[CT]:
+        raise NotImplementedError
+
+
 # Slim connectors retrieve just the ids of documents
 class SlimConnector(BaseConnector):
     @abc.abstractmethod

common/data_source/models.py

@@ -94,7 +94,10 @@ class Document(BaseModel):
     blob: bytes
     doc_updated_at: datetime
     size_bytes: int
+    externale_access: Optional[ExternalAccess] = None
+    primary_owners: Optional[list] = None
     metadata: Optional[dict[str, Any]] = None
+    doc_metadata: Optional[dict[str, Any]] = None
 
 
 class BasicExpertInfo(BaseModel):
@@ -180,6 +183,7 @@ class NotionPage(BaseModel):
     archived: bool
     properties: dict[str, Any]
     url: str
+    parent: Optional[dict[str, Any]] = None  # Parent reference for path reconstruction
     database_name: Optional[str] = None  # Only applicable to database type pages
 
 

common/data_source/notion_connector.py

@@ -66,6 +66,7 @@ class NotionConnector(LoadConnector, PollConnector):
         self.indexed_pages: set[str] = set()
         self.root_page_id = root_page_id
         self.recursive_index_enabled = recursive_index_enabled or bool(root_page_id)
+        self.page_path_cache: dict[str, str] = {}
 
     @retry(tries=3, delay=1, backoff=2)
     def _fetch_child_blocks(self, block_id: str, cursor: Optional[str] = None) -> dict[str, Any] | None:
@@ -242,6 +243,20 @@ class NotionConnector(LoadConnector, PollConnector):
             logging.warning(f"[Notion]: Failed to download Notion file from {url}: {exc}")
             return None
 
+    def _append_block_id_to_name(self, name: str, block_id: Optional[str]) -> str:
+        """Append the Notion block ID to the filename while keeping the extension."""
+        if not block_id:
+            return name
+
+        path = Path(name)
+        stem = path.stem or name
+        suffix = path.suffix
+
+        if not stem:
+            return name
+
+        return f"{stem}_{block_id}{suffix}" if suffix else f"{stem}_{block_id}"
+
     def _extract_file_metadata(self, result_obj: dict[str, Any], block_id: str) -> tuple[str | None, str, str | None]:
         file_source_type = result_obj.get("type")
         file_source = result_obj.get(file_source_type, {}) if file_source_type else {}
@@ -254,6 +269,8 @@ class NotionConnector(LoadConnector, PollConnector):
         elif not name:
             name = f"notion_file_{block_id}"
 
+        name = self._append_block_id_to_name(name, block_id)
+
         caption = self._extract_rich_text(result_obj.get("caption", [])) if "caption" in result_obj else None
 
         return url, name, caption
@@ -265,6 +282,7 @@ class NotionConnector(LoadConnector, PollConnector):
         name: str,
         caption: Optional[str],
         page_last_edited_time: Optional[str],
+        page_path: Optional[str],
     ) -> Document | None:
         file_bytes = self._download_file(url)
         if file_bytes is None:
@@ -277,7 +295,8 @@ class NotionConnector(LoadConnector, PollConnector):
             extension = ".bin"
 
         updated_at = datetime_from_string(page_last_edited_time) if page_last_edited_time else datetime.now(timezone.utc)
-        semantic_identifier = caption or name or f"Notion file {block_id}"
+        base_identifier = name or caption or (f"Notion file {block_id}" if block_id else "Notion file")
+        semantic_identifier = f"{page_path} / {base_identifier}" if page_path else base_identifier
 
         return Document(
             id=block_id,
@@ -289,7 +308,7 @@ class NotionConnector(LoadConnector, PollConnector):
             doc_updated_at=updated_at,
         )
 
-    def _read_blocks(self, base_block_id: str, page_last_edited_time: Optional[str] = None) -> tuple[list[NotionBlock], list[str], list[Document]]:
+    def _read_blocks(self, base_block_id: str, page_last_edited_time: Optional[str] = None, page_path: Optional[str] = None) -> tuple[list[NotionBlock], list[str], list[Document]]:
         result_blocks: list[NotionBlock] = []
         child_pages: list[str] = []
         attachments: list[Document] = []
@@ -370,11 +389,14 @@ class NotionConnector(LoadConnector, PollConnector):
                             name=file_name,
                             caption=caption,
                             page_last_edited_time=page_last_edited_time,
+                            page_path=page_path,
                         )
                         if attachment_doc:
                             attachments.append(attachment_doc)
 
-                        attachment_label = caption or file_name
+                        attachment_label = file_name
+                        if caption:
+                            attachment_label = f"{file_name} ({caption})"
                         if attachment_label:
                             cur_result_text_arr.append(f"{result_type.capitalize()}: {attachment_label}")
 
@@ -383,7 +405,7 @@ class NotionConnector(LoadConnector, PollConnector):
                         child_pages.append(result_block_id)
                     else:
                         logging.debug(f"[Notion]: Entering sub-block: {result_block_id}")
-                        subblocks, subblock_child_pages, subblock_attachments = self._read_blocks(result_block_id, page_last_edited_time)
+                        subblocks, subblock_child_pages, subblock_attachments = self._read_blocks(result_block_id, page_last_edited_time, page_path)
                         logging.debug(f"[Notion]: Finished sub-block: {result_block_id}")
                         result_blocks.extend(subblocks)
                         child_pages.extend(subblock_child_pages)
@@ -423,6 +445,35 @@ class NotionConnector(LoadConnector, PollConnector):
 
         return None
 
+    def _build_page_path(self, page: NotionPage, visited: Optional[set[str]] = None) -> Optional[str]:
+        """Construct a hierarchical path for a page based on its parent chain."""
+        if page.id in self.page_path_cache:
+            return self.page_path_cache[page.id]
+
+        visited = visited or set()
+        if page.id in visited:
+            logging.warning(f"[Notion]: Detected cycle while building path for page {page.id}")
+            return self._read_page_title(page)
+        visited.add(page.id)
+
+        current_title = self._read_page_title(page) or f"Untitled Page {page.id}"
+
+        parent_info = getattr(page, "parent", None) or {}
+        parent_type = parent_info.get("type")
+        parent_id = parent_info.get(parent_type) if parent_type else None
+
+        parent_path = None
+        if parent_type in {"page_id", "database_id"} and isinstance(parent_id, str):
+            try:
+                parent_page = self._fetch_page(parent_id)
+                parent_path = self._build_page_path(parent_page, visited)
+            except Exception as exc:
+                logging.warning(f"[Notion]: Failed to resolve parent {parent_id} for page {page.id}: {exc}")
+
+        full_path = f"{parent_path} / {current_title}" if parent_path else current_title
+        self.page_path_cache[page.id] = full_path
+        return full_path
+
     def _read_pages(self, pages: list[NotionPage], start: SecondsSinceUnixEpoch | None = None, end: SecondsSinceUnixEpoch | None = None) -> Generator[Document, None, None]:
         """Reads pages for rich text content and generates Documents."""
         all_child_page_ids: list[str] = []
@@ -441,13 +492,18 @@ class NotionConnector(LoadConnector, PollConnector):
                     continue
 
             logging.info(f"[Notion]: Reading page with ID {page.id}, with url {page.url}")
-            page_blocks, child_page_ids, attachment_docs = self._read_blocks(page.id, page.last_edited_time)
+            page_path = self._build_page_path(page)
+            page_blocks, child_page_ids, attachment_docs = self._read_blocks(page.id, page.last_edited_time, page_path)
             all_child_page_ids.extend(child_page_ids)
             self.indexed_pages.add(page.id)
 
             raw_page_title = self._read_page_title(page)
             page_title = raw_page_title or f"Untitled Page with ID {page.id}"
 
+            # Append the page id to help disambiguate duplicate names
+            base_identifier = page_path or page_title
+            semantic_identifier = f"{base_identifier}_{page.id}" if base_identifier else page.id
+
             if not page_blocks:
                 if not raw_page_title:
                     logging.warning(f"[Notion]: No blocks OR title found for page with ID {page.id}. Skipping.")
@@ -469,7 +525,7 @@ class NotionConnector(LoadConnector, PollConnector):
             joined_text = "\n".join(sec.text for sec in sections)
             blob = joined_text.encode("utf-8")
             yield Document(
-                id=page.id, blob=blob, source=DocumentSource.NOTION, semantic_identifier=page_title, extension=".txt", size_bytes=len(blob), doc_updated_at=datetime_from_string(page.last_edited_time)
+                id=page.id, blob=blob, source=DocumentSource.NOTION, semantic_identifier=semantic_identifier, extension=".txt", size_bytes=len(blob), doc_updated_at=datetime_from_string(page.last_edited_time)
             )
 
             for attachment_doc in attachment_docs:
@@ -597,4 +653,4 @@ if __name__ == "__main__":
     document_batches = connector.load_from_state()
     for doc_batch in document_batches:
         for doc in doc_batch:
-            print(doc)
+            print(doc)

common/data_source/slack_connector.py

@@ -167,7 +167,6 @@ def get_latest_message_time(thread: ThreadType) -> datetime:
 
 
 def _build_doc_id(channel_id: str, thread_ts: str) -> str:
-    """构建文档ID"""
     return f"{channel_id}__{thread_ts}"
 
 
@@ -179,7 +178,6 @@ def thread_to_doc(
     user_cache: dict[str, BasicExpertInfo | None],
     channel_access: Any | None,
 ) -> Document:
-    """将线程转换为文档"""
     channel_id = channel["id"]
 
     initial_sender_expert_info = expert_info_from_slack_id(
@@ -237,7 +235,6 @@ def filter_channels(
     channels_to_connect: list[str] | None,
     regex_enabled: bool,
 ) -> list[ChannelType]:
-    """过滤频道"""
     if not channels_to_connect:
         return all_channels
 
@@ -381,7 +378,6 @@ def _process_message(
         [MessageType], SlackMessageFilterReason | None
     ] = default_msg_filter,
 ) -> ProcessedSlackMessage:
-    """处理消息"""
     thread_ts = message.get("thread_ts")
     thread_or_message_ts = thread_ts or message["ts"]
     try:
@@ -536,7 +532,6 @@ class SlackConnector(
         end: SecondsSinceUnixEpoch | None = None,
         callback: Any = None,
     ) -> GenerateSlimDocumentOutput:
-        """获取所有简化文档（带权限同步）"""
         if self.client is None:
             raise ConnectorMissingCredentialError("Slack")
 

common/data_source/utils.py

@@ -254,18 +254,21 @@ def create_s3_client(bucket_type: BlobType, credentials: dict[str, Any], europea
     elif bucket_type == BlobType.S3:
         authentication_method = credentials.get("authentication_method", "access_key")
 
+        region_name = credentials.get("region") or None
+
         if authentication_method == "access_key":
             session = boto3.Session(
                 aws_access_key_id=credentials["aws_access_key_id"],
                 aws_secret_access_key=credentials["aws_secret_access_key"],
+                region_name=region_name,
             )
-            return session.client("s3")
+            return session.client("s3", region_name=region_name)
 
         elif authentication_method == "iam_role":
             role_arn = credentials["aws_role_arn"]
 
             def _refresh_credentials() -> dict[str, str]:
-                sts_client = boto3.client("sts")
+                sts_client = boto3.client("sts", region_name=credentials.get("region") or None)
                 assumed_role_object = sts_client.assume_role(
                     RoleArn=role_arn,
                     RoleSessionName=f"onyx_blob_storage_{int(datetime.now().timestamp())}",
@@ -285,11 +288,11 @@ def create_s3_client(bucket_type: BlobType, credentials: dict[str, Any], europea
             )
             botocore_session = get_session()
             botocore_session._credentials = refreshable
-            session = boto3.Session(botocore_session=botocore_session)
-            return session.client("s3")
+            session = boto3.Session(botocore_session=botocore_session, region_name=region_name)
+            return session.client("s3", region_name=region_name)
 
         elif authentication_method == "assume_role":
-            return boto3.client("s3")
+            return boto3.client("s3", region_name=region_name)
 
         else:
             raise ValueError("Invalid authentication method for S3.")
@@ -1146,3 +1149,137 @@ def parallel_yield(gens: list[Iterator[R]], max_workers: int = 10) -> Iterator[R
                     future_to_index[executor.submit(_next_or_none, ind, gens[ind])] = next_ind
                     next_ind += 1
                 del future_to_index[future]
+
+
+def sanitize_filename(name: str, extension: str = "txt") -> str:
+    """
+    Soft sanitize for MinIO/S3:
+    - Replace only prohibited characters with a space.
+    - Preserve readability (no ugly underscores).
+    - Collapse multiple spaces.
+    """
+    if name is None:
+        return f"file.{extension}"
+
+    name = str(name).strip()
+
+    # Characters that MUST NOT appear in S3/MinIO object keys
+    # Replace them with a space (not underscore)
+    forbidden = r'[\\\?\#\%\*\:\|\<\>"]'
+    name = re.sub(forbidden, " ", name)
+
+    # Replace slashes "/" (S3 interprets as folder) with space
+    name = name.replace("/", " ")
+
+    # Collapse multiple spaces into one
+    name = re.sub(r"\s+", " ", name)
+
+    # Trim both ends
+    name = name.strip()
+
+    # Enforce reasonable max length
+    if len(name) > 200:
+        base, ext = os.path.splitext(name)
+        name = base[:180].rstrip() + ext
+
+    if not os.path.splitext(name)[1]:
+        name += f".{extension}"
+
+    return name
+F = TypeVar("F", bound=Callable[..., Any])
+
+class _RateLimitDecorator:
+    """Builds a generic wrapper/decorator for calls to external APIs that
+    prevents making more than `max_calls` requests per `period`
+
+    Implementation inspired by the `ratelimit` library:
+    https://github.com/tomasbasham/ratelimit.
+
+    NOTE: is not thread safe.
+    """
+
+    def __init__(
+        self,
+        max_calls: int,
+        period: float,  # in seconds
+        sleep_time: float = 2,  # in seconds
+        sleep_backoff: float = 2,  # applies exponential backoff
+        max_num_sleep: int = 0,
+    ):
+        self.max_calls = max_calls
+        self.period = period
+        self.sleep_time = sleep_time
+        self.sleep_backoff = sleep_backoff
+        self.max_num_sleep = max_num_sleep
+
+        self.call_history: list[float] = []
+        self.curr_calls = 0
+
+    def __call__(self, func: F) -> F:
+        @wraps(func)
+        def wrapped_func(*args: list, **kwargs: dict[str, Any]) -> Any:
+            # cleanup calls which are no longer relevant
+            self._cleanup()
+
+            # check if we've exceeded the rate limit
+            sleep_cnt = 0
+            while len(self.call_history) == self.max_calls:
+                sleep_time = self.sleep_time * (self.sleep_backoff**sleep_cnt)
+                logging.warning(
+                    f"Rate limit exceeded for function {func.__name__}. "
+                    f"Waiting {sleep_time} seconds before retrying."
+                )
+                time.sleep(sleep_time)
+                sleep_cnt += 1
+                if self.max_num_sleep != 0 and sleep_cnt >= self.max_num_sleep:
+                    raise RateLimitTriedTooManyTimesError(
+                        f"Exceeded '{self.max_num_sleep}' retries for function '{func.__name__}'"
+                    )
+
+                self._cleanup()
+
+            # add the current call to the call history
+            self.call_history.append(time.monotonic())
+            return func(*args, **kwargs)
+
+        return cast(F, wrapped_func)
+
+    def _cleanup(self) -> None:
+        curr_time = time.monotonic()
+        time_to_expire_before = curr_time - self.period
+        self.call_history = [
+            call_time
+            for call_time in self.call_history
+            if call_time > time_to_expire_before
+        ]
+
+rate_limit_builder = _RateLimitDecorator
+
+def retry_builder(
+    tries: int = 20,
+    delay: float = 0.1,
+    max_delay: float | None = 60,
+    backoff: float = 2,
+    jitter: tuple[float, float] | float = 1,
+    exceptions: type[Exception] | tuple[type[Exception], ...] = (Exception,),
+) -> Callable[[F], F]:
+    """Builds a generic wrapper/decorator for calls to external APIs that
+    may fail due to rate limiting, flakes, or other reasons. Applies exponential
+    backoff with jitter to retry the call."""
+
+    def retry_with_default(func: F) -> F:
+        @retry(
+            tries=tries,
+            delay=delay,
+            max_delay=max_delay,
+            backoff=backoff,
+            jitter=jitter,
+            logger=logging.getLogger(__name__),
+            exceptions=exceptions,
+        )
+        def wrapped_func(*args: list, **kwargs: dict[str, Any]) -> Any:
+            return func(*args, **kwargs)
+
+        return cast(F, wrapped_func)
+
+    return retry_with_default

common/data_source/webdav_connector.py

@@ -82,10 +82,6 @@ class WebDAVConnector(LoadConnector, PollConnector):
                 base_url=self.base_url,
                 auth=(username, password)
             )
-            
-            # Test connection
-            self.client.exists(self.remote_path)
-            
         except Exception as e:
             logging.error(f"Failed to connect to WebDAV server: {e}")
             raise ConnectorMissingCredentialError(
@@ -308,60 +304,79 @@ class WebDAVConnector(LoadConnector, PollConnector):
             yield batch
 
     def validate_connector_settings(self) -> None:
-        """Validate WebDAV connector settings
-        
-        Raises:
-            ConnectorMissingCredentialError: If credentials are not loaded
-            ConnectorValidationError: If settings are invalid
+        """Validate WebDAV connector settings.
+
+        Validation should exercise the same code-paths used by the connector
+        (directory listing / PROPFIND), avoiding exists() which may probe with
+        methods that differ across servers.
         """
         if self.client is None:
-            raise ConnectorMissingCredentialError(
-                "WebDAV credentials not loaded."
-            )
+            raise ConnectorMissingCredentialError("WebDAV credentials not loaded.")
 
         if not self.base_url:
-            raise ConnectorValidationError(
-                "No base URL was provided in connector settings."
-            )
+            raise ConnectorValidationError("No base URL was provided in connector settings.")
+
+        # Normalize directory path: for collections, many servers behave better with trailing '/'
+        test_path = self.remote_path or "/"
+        if not test_path.startswith("/"):
+            test_path = f"/{test_path}"
+        if test_path != "/" and not test_path.endswith("/"):
+            test_path = f"{test_path}/"
 
         try:
-            if not self.client.exists(self.remote_path):
-                raise ConnectorValidationError(
-                    f"Remote path '{self.remote_path}' does not exist on WebDAV server."
-                )
+            # Use the same behavior as real sync: list directory with details (PROPFIND)
+            self.client.ls(test_path, detail=True)
 
         except Exception as e:
-            error_message = str(e)
-            
-            if "401" in error_message or "unauthorized" in error_message.lower():
-                raise CredentialExpiredError(
-                    "WebDAV credentials appear invalid or expired."
-                )
-            
-            if "403" in error_message or "forbidden" in error_message.lower():
+            # Prefer structured status codes if present on the exception/response
+            status = None
+            for attr in ("status_code", "code"):
+                v = getattr(e, attr, None)
+                if isinstance(v, int):
+                    status = v
+                    break
+            if status is None:
+                resp = getattr(e, "response", None)
+                v = getattr(resp, "status_code", None)
+                if isinstance(v, int):
+                    status = v
+
+            # If we can classify by status code, do it
+            if status == 401:
+                raise CredentialExpiredError("WebDAV credentials appear invalid or expired.")
+            if status == 403:
                 raise InsufficientPermissionsError(
                     f"Insufficient permissions to access path '{self.remote_path}' on WebDAV server."
                 )
-            
-            if "404" in error_message or "not found" in error_message.lower():
+            if status == 404:
                 raise ConnectorValidationError(
                     f"Remote path '{self.remote_path}' does not exist on WebDAV server."
                 )
 
+            # Fallback: avoid brittle substring matching that caused false positives.
+            # Provide the original exception for diagnosis.
             raise ConnectorValidationError(
-                f"Unexpected WebDAV client error: {e}"
+                f"WebDAV validation failed for path '{test_path}': {repr(e)}"
             )
 
 
+
 if __name__ == "__main__":
     credentials_dict = {
         "username": os.environ.get("WEBDAV_USERNAME"),
         "password": os.environ.get("WEBDAV_PASSWORD"),
     }
 
+    credentials_dict = {
+        "username": "user",
+        "password": "pass",
+    }
+
+
+
     connector = WebDAVConnector(
-        base_url=os.environ.get("WEBDAV_URL") or "https://webdav.example.com",
-        remote_path=os.environ.get("WEBDAV_PATH") or "/",
+        base_url="http://172.17.0.1:8080/",
+        remote_path="/",
     )
 
     try:

common/data_source/zendesk_connector.py

@@ -0,0 +1,667 @@
+import copy
+import logging
+import time
+from collections.abc import Callable
+from collections.abc import Iterator
+from typing import Any
+
+import requests
+from pydantic import BaseModel
+from requests.exceptions import HTTPError
+from typing_extensions import override
+
+from common.data_source.config import ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS, DocumentSource
+from common.data_source.exceptions import ConnectorValidationError, CredentialExpiredError, InsufficientPermissionsError
+from common.data_source.html_utils import parse_html_page_basic
+from common.data_source.interfaces import CheckpointOutput, CheckpointOutputWrapper, CheckpointedConnector, IndexingHeartbeatInterface, SlimConnectorWithPermSync
+from common.data_source.models import BasicExpertInfo, ConnectorCheckpoint, ConnectorFailure, Document, DocumentFailure, GenerateSlimDocumentOutput, SecondsSinceUnixEpoch, SlimDocument
+from common.data_source.utils import retry_builder, time_str_to_utc,rate_limit_builder
+
+MAX_PAGE_SIZE = 30  # Zendesk API maximum
+MAX_AUTHOR_MAP_SIZE = 50_000  # Reset author map cache if it gets too large
+_SLIM_BATCH_SIZE = 1000
+
+
+class ZendeskCredentialsNotSetUpError(PermissionError):
+    def __init__(self) -> None:
+        super().__init__(
+            "Zendesk Credentials are not set up, was load_credentials called?"
+        )
+
+
+class ZendeskClient:
+    def __init__(
+        self,
+        subdomain: str,
+        email: str,
+        token: str,
+        calls_per_minute: int | None = None,
+    ):
+        self.base_url = f"https://{subdomain}.zendesk.com/api/v2"
+        self.auth = (f"{email}/token", token)
+        self.make_request = request_with_rate_limit(self, calls_per_minute)
+
+
+def request_with_rate_limit(
+    client: ZendeskClient, max_calls_per_minute: int | None = None
+) -> Callable[[str, dict[str, Any]], dict[str, Any]]:
+    @retry_builder()
+    @(
+        rate_limit_builder(max_calls=max_calls_per_minute, period=60)
+        if max_calls_per_minute
+        else lambda x: x
+    )
+    def make_request(endpoint: str, params: dict[str, Any]) -> dict[str, Any]:
+        response = requests.get(
+            f"{client.base_url}/{endpoint}", auth=client.auth, params=params
+        )
+
+        if response.status_code == 429:
+            retry_after = response.headers.get("Retry-After")
+            if retry_after is not None:
+                # Sleep for the duration indicated by the Retry-After header
+                time.sleep(int(retry_after))
+
+        elif (
+            response.status_code == 403
+            and response.json().get("error") == "SupportProductInactive"
+        ):
+            return response.json()
+
+        response.raise_for_status()
+        return response.json()
+
+    return make_request
+
+
+class ZendeskPageResponse(BaseModel):
+    data: list[dict[str, Any]]
+    meta: dict[str, Any]
+    has_more: bool
+
+
+def _get_content_tag_mapping(client: ZendeskClient) -> dict[str, str]:
+    content_tags: dict[str, str] = {}
+    params = {"page[size]": MAX_PAGE_SIZE}
+
+    try:
+        while True:
+            data = client.make_request("guide/content_tags", params)
+
+            for tag in data.get("records", []):
+                content_tags[tag["id"]] = tag["name"]
+
+            # Check if there are more pages
+            if data.get("meta", {}).get("has_more", False):
+                params["page[after]"] = data["meta"]["after_cursor"]
+            else:
+                break
+
+        return content_tags
+    except Exception as e:
+        raise Exception(f"Error fetching content tags: {str(e)}")
+
+
+def _get_articles(
+    client: ZendeskClient, start_time: int | None = None, page_size: int = MAX_PAGE_SIZE
+) -> Iterator[dict[str, Any]]:
+    params = {"page[size]": page_size, "sort_by": "updated_at", "sort_order": "asc"}
+    if start_time is not None:
+        params["start_time"] = start_time
+
+    while True:
+        data = client.make_request("help_center/articles", params)
+        for article in data["articles"]:
+            yield article
+
+        if not data.get("meta", {}).get("has_more"):
+            break
+        params["page[after]"] = data["meta"]["after_cursor"]
+
+
+def _get_article_page(
+    client: ZendeskClient,
+    start_time: int | None = None,
+    after_cursor: str | None = None,
+    page_size: int = MAX_PAGE_SIZE,
+) -> ZendeskPageResponse:
+    params = {"page[size]": page_size, "sort_by": "updated_at", "sort_order": "asc"}
+    if start_time is not None:
+        params["start_time"] = start_time
+    if after_cursor is not None:
+        params["page[after]"] = after_cursor
+
+    data = client.make_request("help_center/articles", params)
+    return ZendeskPageResponse(
+        data=data["articles"],
+        meta=data["meta"],
+        has_more=bool(data["meta"].get("has_more", False)),
+    )
+
+
+def _get_tickets(
+    client: ZendeskClient, start_time: int | None = None
+) -> Iterator[dict[str, Any]]:
+    params = {"start_time": start_time or 0}
+
+    while True:
+        data = client.make_request("incremental/tickets.json", params)
+        for ticket in data["tickets"]:
+            yield ticket
+
+        if not data.get("end_of_stream", False):
+            params["start_time"] = data["end_time"]
+        else:
+            break
+
+
+# TODO: maybe these don't need to be their own functions?
+def _get_tickets_page(
+    client: ZendeskClient, start_time: int | None = None
+) -> ZendeskPageResponse:
+    params = {"start_time": start_time or 0}
+
+    # NOTE: for some reason zendesk doesn't seem to be respecting the start_time param
+    # in my local testing with very few tickets. We'll look into it if this becomes an
+    # issue in larger deployments
+    data = client.make_request("incremental/tickets.json", params)
+    if data.get("error") == "SupportProductInactive":
+        raise ValueError(
+            "Zendesk Support Product is not active for this account, No tickets to index"
+        )
+    return ZendeskPageResponse(
+        data=data["tickets"],
+        meta={"end_time": data["end_time"]},
+        has_more=not bool(data.get("end_of_stream", False)),
+    )
+
+
+def _fetch_author(
+    client: ZendeskClient, author_id: str | int
+) -> BasicExpertInfo | None:
+    # Skip fetching if author_id is invalid
+    # cast to str to avoid issues with zendesk changing their types
+    if not author_id or str(author_id) == "-1":
+        return None
+
+    try:
+        author_data = client.make_request(f"users/{author_id}", {})
+        user = author_data.get("user")
+        return (
+            BasicExpertInfo(display_name=user.get("name"), email=user.get("email"))
+            if user and user.get("name") and user.get("email")
+            else None
+        )
+    except requests.exceptions.HTTPError:
+        # Handle any API errors gracefully
+        return None
+
+
+def _article_to_document(
+    article: dict[str, Any],
+    content_tags: dict[str, str],
+    author_map: dict[str, BasicExpertInfo],
+    client: ZendeskClient,
+) -> tuple[dict[str, BasicExpertInfo] | None, Document]:
+    author_id = article.get("author_id")
+    if not author_id:
+        author = None
+    else:
+        author = (
+            author_map.get(author_id)
+            if author_id in author_map
+            else _fetch_author(client, author_id)
+        )
+
+    new_author_mapping = {author_id: author} if author_id and author else None
+
+    updated_at = article.get("updated_at")
+    update_time = time_str_to_utc(updated_at) if updated_at else None
+
+    text = parse_html_page_basic(article.get("body") or "")
+    blob = text.encode("utf-8", errors="replace")
+    # Build metadata
+    metadata: dict[str, str | list[str]] = {
+        "labels": [str(label) for label in article.get("label_names", []) if label],
+        "content_tags": [
+            content_tags[tag_id]
+            for tag_id in article.get("content_tag_ids", [])
+            if tag_id in content_tags
+        ],
+    }
+
+    # Remove empty values
+    metadata = {k: v for k, v in metadata.items() if v}
+
+    return new_author_mapping, Document(
+        id=f"article:{article['id']}",
+        source=DocumentSource.ZENDESK,
+        semantic_identifier=article["title"],
+        extension=".txt",
+        blob=blob,
+        size_bytes=len(blob),
+        doc_updated_at=update_time,
+        primary_owners=[author] if author else None,
+        metadata=metadata,
+    )
+
+
+def _get_comment_text(
+    comment: dict[str, Any],
+    author_map: dict[str, BasicExpertInfo],
+    client: ZendeskClient,
+) -> tuple[dict[str, BasicExpertInfo] | None, str]:
+    author_id = comment.get("author_id")
+    if not author_id:
+        author = None
+    else:
+        author = (
+            author_map.get(author_id)
+            if author_id in author_map
+            else _fetch_author(client, author_id)
+        )
+
+    new_author_mapping = {author_id: author} if author_id and author else None
+
+    comment_text = f"Comment{' by ' + author.display_name if author and author.display_name else ''}"
+    comment_text += f"{' at ' + comment['created_at'] if comment.get('created_at') else ''}:\n{comment['body']}"
+
+    return new_author_mapping, comment_text
+
+
+def _ticket_to_document(
+    ticket: dict[str, Any],
+    author_map: dict[str, BasicExpertInfo],
+    client: ZendeskClient,
+) -> tuple[dict[str, BasicExpertInfo] | None, Document]:
+    submitter_id = ticket.get("submitter")
+    if not submitter_id:
+        submitter = None
+    else:
+        submitter = (
+            author_map.get(submitter_id)
+            if submitter_id in author_map
+            else _fetch_author(client, submitter_id)
+        )
+
+    new_author_mapping = (
+        {submitter_id: submitter} if submitter_id and submitter else None
+    )
+
+    updated_at = ticket.get("updated_at")
+    update_time = time_str_to_utc(updated_at) if updated_at else None
+
+    metadata: dict[str, str | list[str]] = {}
+    if status := ticket.get("status"):
+        metadata["status"] = status
+    if priority := ticket.get("priority"):
+        metadata["priority"] = priority
+    if tags := ticket.get("tags"):
+        metadata["tags"] = tags
+    if ticket_type := ticket.get("type"):
+        metadata["ticket_type"] = ticket_type
+
+    # Fetch comments for the ticket
+    comments_data = client.make_request(f"tickets/{ticket.get('id')}/comments", {})
+    comments = comments_data.get("comments", [])
+
+    comment_texts = []
+    for comment in comments:
+        new_author_mapping, comment_text = _get_comment_text(
+            comment, author_map, client
+        )
+        if new_author_mapping:
+            author_map.update(new_author_mapping)
+        comment_texts.append(comment_text)
+
+    comments_text = "\n\n".join(comment_texts)
+
+    subject = ticket.get("subject")
+    full_text = f"Ticket Subject:\n{subject}\n\nComments:\n{comments_text}"
+
+    blob = full_text.encode("utf-8", errors="replace")
+    return new_author_mapping, Document(
+        id=f"zendesk_ticket_{ticket['id']}",
+        blob=blob,
+        extension=".txt",
+        size_bytes=len(blob),
+        source=DocumentSource.ZENDESK,
+        semantic_identifier=f"Ticket #{ticket['id']}: {subject or 'No Subject'}",
+        doc_updated_at=update_time,
+        primary_owners=[submitter] if submitter else None,
+        metadata=metadata,
+    )
+
+
+class ZendeskConnectorCheckpoint(ConnectorCheckpoint):
+    # We use cursor-based paginated retrieval for articles
+    after_cursor_articles: str | None
+
+    # We use timestamp-based paginated retrieval for tickets
+    next_start_time_tickets: int | None
+
+    cached_author_map: dict[str, BasicExpertInfo] | None
+    cached_content_tags: dict[str, str] | None
+
+
+class ZendeskConnector(
+    SlimConnectorWithPermSync, CheckpointedConnector[ZendeskConnectorCheckpoint]
+):
+    def __init__(
+        self,
+        content_type: str = "articles",
+        calls_per_minute: int | None = None,
+    ) -> None:
+        self.content_type = content_type
+        self.subdomain = ""
+        # Fetch all tags ahead of time
+        self.content_tags: dict[str, str] = {}
+        self.calls_per_minute = calls_per_minute
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        # Subdomain is actually the whole URL
+        subdomain = (
+            credentials["zendesk_subdomain"]
+            .replace("https://", "")
+            .split(".zendesk.com")[0]
+        )
+        self.subdomain = subdomain
+
+        self.client = ZendeskClient(
+            subdomain,
+            credentials["zendesk_email"],
+            credentials["zendesk_token"],
+            calls_per_minute=self.calls_per_minute,
+        )
+        return None
+
+    @override
+    def load_from_checkpoint(
+        self,
+        start: SecondsSinceUnixEpoch,
+        end: SecondsSinceUnixEpoch,
+        checkpoint: ZendeskConnectorCheckpoint,
+    ) -> CheckpointOutput[ZendeskConnectorCheckpoint]:
+        if self.client is None:
+            raise ZendeskCredentialsNotSetUpError()
+        if checkpoint.cached_content_tags is None:
+            checkpoint.cached_content_tags = _get_content_tag_mapping(self.client)
+            return checkpoint  # save the content tags to the checkpoint
+        self.content_tags = checkpoint.cached_content_tags
+
+        if self.content_type == "articles":
+            checkpoint = yield from self._retrieve_articles(start, end, checkpoint)
+            return checkpoint
+        elif self.content_type == "tickets":
+            checkpoint = yield from self._retrieve_tickets(start, end, checkpoint)
+            return checkpoint
+        else:
+            raise ValueError(f"Unsupported content_type: {self.content_type}")
+
+    def _retrieve_articles(
+        self,
+        start: SecondsSinceUnixEpoch | None,
+        end: SecondsSinceUnixEpoch | None,
+        checkpoint: ZendeskConnectorCheckpoint,
+    ) -> CheckpointOutput[ZendeskConnectorCheckpoint]:
+        checkpoint = copy.deepcopy(checkpoint)
+        # This one is built on the fly as there may be more many more authors than tags
+        author_map: dict[str, BasicExpertInfo] = checkpoint.cached_author_map or {}
+        after_cursor = checkpoint.after_cursor_articles
+        doc_batch: list[Document] = []
+
+        response = _get_article_page(
+            self.client,
+            start_time=int(start) if start else None,
+            after_cursor=after_cursor,
+        )
+        articles = response.data
+        has_more = response.has_more
+        after_cursor = response.meta.get("after_cursor")
+        for article in articles:
+            if (
+                article.get("body") is None
+                or article.get("draft")
+                or any(
+                    label in ZENDESK_CONNECTOR_SKIP_ARTICLE_LABELS
+                    for label in article.get("label_names", [])
+                )
+            ):
+                continue
+
+            try:
+                new_author_map, document = _article_to_document(
+                    article, self.content_tags, author_map, self.client
+                )
+            except Exception as e:
+                logging.error(f"Error processing article {article['id']}: {e}")
+                yield ConnectorFailure(
+                    failed_document=DocumentFailure(
+                        document_id=f"{article.get('id')}",
+                        document_link=article.get("html_url", ""),
+                    ),
+                    failure_message=str(e),
+                    exception=e,
+                )
+                continue
+
+            if new_author_map:
+                author_map.update(new_author_map)
+            updated_at = document.doc_updated_at
+            updated_ts = updated_at.timestamp() if updated_at else None
+            if updated_ts is not None:
+                if start is not None and updated_ts <= start:
+                    continue
+                if end is not None and updated_ts > end:
+                    continue
+
+            doc_batch.append(document)
+
+        if not has_more:
+            yield from doc_batch
+            checkpoint.has_more = False
+            return checkpoint
+
+        # Sometimes no documents are retrieved, but the cursor
+        # is still updated so the connector makes progress.
+        yield from doc_batch
+        checkpoint.after_cursor_articles = after_cursor
+
+        last_doc_updated_at = doc_batch[-1].doc_updated_at if doc_batch else None
+        checkpoint.has_more = bool(
+            end is None
+            or last_doc_updated_at is None
+            or last_doc_updated_at.timestamp() <= end
+        )
+        checkpoint.cached_author_map = (
+            author_map if len(author_map) <= MAX_AUTHOR_MAP_SIZE else None
+        )
+        return checkpoint
+
+    def _retrieve_tickets(
+        self,
+        start: SecondsSinceUnixEpoch | None,
+        end: SecondsSinceUnixEpoch | None,
+        checkpoint: ZendeskConnectorCheckpoint,
+    ) -> CheckpointOutput[ZendeskConnectorCheckpoint]:
+        checkpoint = copy.deepcopy(checkpoint)
+        if self.client is None:
+            raise ZendeskCredentialsNotSetUpError()
+
+        author_map: dict[str, BasicExpertInfo] = checkpoint.cached_author_map or {}
+
+        doc_batch: list[Document] = []
+        next_start_time = int(checkpoint.next_start_time_tickets or start or 0)
+        ticket_response = _get_tickets_page(self.client, start_time=next_start_time)
+
+        tickets = ticket_response.data
+        has_more = ticket_response.has_more
+        next_start_time = ticket_response.meta["end_time"]
+        for ticket in tickets:
+            if ticket.get("status") == "deleted":
+                continue
+
+            try:
+                new_author_map, document = _ticket_to_document(
+                    ticket=ticket,
+                    author_map=author_map,
+                    client=self.client,
+                )
+            except Exception as e:
+                logging.error(f"Error processing ticket {ticket['id']}: {e}")
+                yield ConnectorFailure(
+                    failed_document=DocumentFailure(
+                        document_id=f"{ticket.get('id')}",
+                        document_link=ticket.get("url", ""),
+                    ),
+                    failure_message=str(e),
+                    exception=e,
+                )
+                continue
+
+            if new_author_map:
+                author_map.update(new_author_map)
+
+            updated_at = document.doc_updated_at
+            updated_ts = updated_at.timestamp() if updated_at else None
+
+            if updated_ts is not None:
+                if start is not None and updated_ts <= start:
+                    continue
+                if end is not None and updated_ts > end:
+                    continue
+
+            doc_batch.append(document)
+
+        if not has_more:
+            yield from doc_batch
+            checkpoint.has_more = False
+            return checkpoint
+
+        yield from doc_batch
+        checkpoint.next_start_time_tickets = next_start_time
+        last_doc_updated_at = doc_batch[-1].doc_updated_at if doc_batch else None
+        checkpoint.has_more = bool(
+            end is None
+            or last_doc_updated_at is None
+            or last_doc_updated_at.timestamp() <= end
+        )
+        checkpoint.cached_author_map = (
+            author_map if len(author_map) <= MAX_AUTHOR_MAP_SIZE else None
+        )
+        return checkpoint
+
+    def retrieve_all_slim_docs_perm_sync(
+        self,
+        start: SecondsSinceUnixEpoch | None = None,
+        end: SecondsSinceUnixEpoch | None = None,
+        callback: IndexingHeartbeatInterface | None = None,
+    ) -> GenerateSlimDocumentOutput:
+        slim_doc_batch: list[SlimDocument] = []
+        if self.content_type == "articles":
+            articles = _get_articles(
+                self.client, start_time=int(start) if start else None
+            )
+            for article in articles:
+                slim_doc_batch.append(
+                    SlimDocument(
+                        id=f"article:{article['id']}",
+                    )
+                )
+                if len(slim_doc_batch) >= _SLIM_BATCH_SIZE:
+                    yield slim_doc_batch
+                    slim_doc_batch = []
+        elif self.content_type == "tickets":
+            tickets = _get_tickets(
+                self.client, start_time=int(start) if start else None
+            )
+            for ticket in tickets:
+                slim_doc_batch.append(
+                    SlimDocument(
+                        id=f"zendesk_ticket_{ticket['id']}",
+                    )
+                )
+                if len(slim_doc_batch) >= _SLIM_BATCH_SIZE:
+                    yield slim_doc_batch
+                    slim_doc_batch = []
+        else:
+            raise ValueError(f"Unsupported content_type: {self.content_type}")
+        if slim_doc_batch:
+            yield slim_doc_batch
+
+    @override
+    def validate_connector_settings(self) -> None:
+        if self.client is None:
+            raise ZendeskCredentialsNotSetUpError()
+
+        try:
+            _get_article_page(self.client, start_time=0)
+        except HTTPError as e:
+            # Check for HTTP status codes
+            if e.response.status_code == 401:
+                raise CredentialExpiredError(
+                    "Your Zendesk credentials appear to be invalid or expired (HTTP 401)."
+                ) from e
+            elif e.response.status_code == 403:
+                raise InsufficientPermissionsError(
+                    "Your Zendesk token does not have sufficient permissions (HTTP 403)."
+                ) from e
+            elif e.response.status_code == 404:
+                raise ConnectorValidationError(
+                    "Zendesk resource not found (HTTP 404)."
+                ) from e
+            else:
+                raise ConnectorValidationError(
+                    f"Unexpected Zendesk error (status={e.response.status_code}): {e}"
+                ) from e
+
+    @override
+    def validate_checkpoint_json(
+        self, checkpoint_json: str
+    ) -> ZendeskConnectorCheckpoint:
+        return ZendeskConnectorCheckpoint.model_validate_json(checkpoint_json)
+
+    @override
+    def build_dummy_checkpoint(self) -> ZendeskConnectorCheckpoint:
+        return ZendeskConnectorCheckpoint(
+            after_cursor_articles=None,
+            next_start_time_tickets=None,
+            cached_author_map=None,
+            cached_content_tags=None,
+            has_more=True,
+        )
+
+
+if __name__ == "__main__":
+    import os
+
+    connector = ZendeskConnector(content_type="articles")
+    connector.load_credentials(
+        {
+            "zendesk_subdomain": os.environ["ZENDESK_SUBDOMAIN"],
+            "zendesk_email": os.environ["ZENDESK_EMAIL"],
+            "zendesk_token": os.environ["ZENDESK_TOKEN"],
+        }
+    )
+
+    current = time.time()
+    one_day_ago = current - 24 * 60 * 60  # 1 day
+
+    checkpoint = connector.build_dummy_checkpoint()
+
+    while checkpoint.has_more:
+        gen = connector.load_from_checkpoint(
+            one_day_ago, current, checkpoint
+        )
+
+        wrapper = CheckpointOutputWrapper()
+        any_doc = False
+
+        for document, failure, next_checkpoint in wrapper(gen):
+            if document:
+                print("got document:", document.id)
+                any_doc = True
+
+        checkpoint = next_checkpoint
+        if any_doc:
+            break

common/doc_store/doc_store_base.py

@@ -13,7 +13,6 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
 import numpy as np
@@ -22,7 +21,6 @@ DEFAULT_MATCH_VECTOR_TOPN = 10
 DEFAULT_MATCH_SPARSE_TOPN = 10
 VEC = list | np.ndarray
 
-
 @dataclass
 class SparseVector:
     indices: list[int]
@@ -55,14 +53,13 @@ class SparseVector:
     def __repr__(self):
         return str(self)
 
-
-class MatchTextExpr(ABC):
+class MatchTextExpr:
     def __init__(
         self,
         fields: list[str],
         matching_text: str,
         topn: int,
-        extra_options: dict = dict(),
+        extra_options: dict | None = None,
     ):
         self.fields = fields
         self.matching_text = matching_text
@@ -70,7 +67,7 @@ class MatchTextExpr(ABC):
         self.extra_options = extra_options
 
 
-class MatchDenseExpr(ABC):
+class MatchDenseExpr:
     def __init__(
         self,
         vector_column_name: str,
@@ -78,7 +75,7 @@ class MatchDenseExpr(ABC):
         embedding_data_type: str,
         distance_type: str,
         topn: int = DEFAULT_MATCH_VECTOR_TOPN,
-        extra_options: dict = dict(),
+        extra_options: dict | None = None,
     ):
         self.vector_column_name = vector_column_name
         self.embedding_data = embedding_data
@@ -88,7 +85,7 @@ class MatchDenseExpr(ABC):
         self.extra_options = extra_options
 
 
-class MatchSparseExpr(ABC):
+class MatchSparseExpr:
     def __init__(
         self,
         vector_column_name: str,
@@ -104,7 +101,7 @@ class MatchSparseExpr(ABC):
         self.opt_params = opt_params
 
 
-class MatchTensorExpr(ABC):
+class MatchTensorExpr:
     def __init__(
         self,
         column_name: str,
@@ -120,7 +117,7 @@ class MatchTensorExpr(ABC):
         self.extra_option = extra_option
 
 
-class FusionExpr(ABC):
+class FusionExpr:
     def __init__(self, method: str, topn: int, fusion_params: dict | None = None):
         self.method = method
         self.topn = topn
@@ -129,7 +126,8 @@ class FusionExpr(ABC):
 
 MatchExpr = MatchTextExpr | MatchDenseExpr | MatchSparseExpr | MatchTensorExpr | FusionExpr
 
-class OrderByExpr(ABC):
+
+class OrderByExpr:
     def __init__(self):
         self.fields = list()
     def asc(self, field: str):
@@ -141,13 +139,14 @@ class OrderByExpr(ABC):
     def fields(self):
         return self.fields
 
+
 class DocStoreConnection(ABC):
     """
     Database operations
     """
 
     @abstractmethod
-    def dbType(self) -> str:
+    def db_type(self) -> str:
         """
         Return the type of the database.
         """
@@ -165,21 +164,21 @@ class DocStoreConnection(ABC):
     """
 
     @abstractmethod
-    def createIdx(self, indexName: str, knowledgebaseId: str, vectorSize: int):
+    def create_idx(self, index_name: str, dataset_id: str, vector_size: int):
         """
         Create an index with given name
         """
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def deleteIdx(self, indexName: str, knowledgebaseId: str):
+    def delete_idx(self, index_name: str, dataset_id: str):
         """
         Delete an index with given name
         """
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def indexExist(self, indexName: str, knowledgebaseId: str) -> bool:
+    def index_exist(self, index_name: str, dataset_id: str) -> bool:
         """
         Check if an index with given name exists
         """
@@ -191,16 +190,16 @@ class DocStoreConnection(ABC):
 
     @abstractmethod
     def search(
-        self, selectFields: list[str],
-            highlightFields: list[str],
+        self, select_fields: list[str],
+            highlight_fields: list[str],
             condition: dict,
-            matchExprs: list[MatchExpr],
-            orderBy: OrderByExpr,
+            match_expressions: list[MatchExpr],
+            order_by: OrderByExpr,
             offset: int,
             limit: int,
-            indexNames: str|list[str],
-            knowledgebaseIds: list[str],
-            aggFields: list[str] = [],
+            index_names: str|list[str],
+            dataset_ids: list[str],
+            agg_fields: list[str] | None = None,
             rank_feature: dict | None = None
     ):
         """
@@ -209,28 +208,28 @@ class DocStoreConnection(ABC):
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def get(self, chunkId: str, indexName: str, knowledgebaseIds: list[str]) -> dict | None:
+    def get(self, data_id: str, index_name: str, dataset_ids: list[str]) -> dict | None:
         """
         Get single chunk with given id
         """
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def insert(self, rows: list[dict], indexName: str, knowledgebaseId: str = None) -> list[str]:
+    def insert(self, rows: list[dict], index_name: str, dataset_id: str = None) -> list[str]:
         """
         Update or insert a bulk of rows
         """
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def update(self, condition: dict, newValue: dict, indexName: str, knowledgebaseId: str) -> bool:
+    def update(self, condition: dict, new_value: dict, index_name: str, dataset_id: str) -> bool:
         """
         Update rows with given conjunctive equivalent filtering condition
         """
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def delete(self, condition: dict, indexName: str, knowledgebaseId: str) -> int:
+    def delete(self, condition: dict, index_name: str, dataset_id: str) -> int:
         """
         Delete rows with given conjunctive equivalent filtering condition
         """
@@ -245,7 +244,7 @@ class DocStoreConnection(ABC):
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def get_chunk_ids(self, res):
+    def get_doc_ids(self, res):
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
@@ -253,18 +252,18 @@ class DocStoreConnection(ABC):
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def get_highlight(self, res, keywords: list[str], fieldnm: str):
+    def get_highlight(self, res, keywords: list[str], field_name: str):
         raise NotImplementedError("Not implemented")
 
     @abstractmethod
-    def get_aggregation(self, res, fieldnm: str):
+    def get_aggregation(self, res, field_name: str):
         raise NotImplementedError("Not implemented")
 
     """
     SQL
     """
     @abstractmethod
-    def sql(sql: str, fetch_size: int, format: str):
+    def sql(self, sql: str, fetch_size: int, format: str):
         """
         Run the sql generated by text-to-sql
         """

common/doc_store/es_conn_base.py

@@ -0,0 +1,307 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+import logging
+import re
+import json
+import time
+import os
+from abc import abstractmethod
+
+from elasticsearch import NotFoundError
+from elasticsearch_dsl import Index
+from elastic_transport import ConnectionTimeout
+from common.file_utils import get_project_base_directory
+from common.misc_utils import convert_bytes
+from common.doc_store.doc_store_base import DocStoreConnection, OrderByExpr, MatchExpr
+from rag.nlp import is_english, rag_tokenizer
+from common import settings
+
+ATTEMPT_TIME = 2
+
+
+class ESConnectionBase(DocStoreConnection):
+    def __init__(self, mapping_file_name: str="mapping.json", logger_name: str='ragflow.es_conn'):
+        from common.doc_store.es_conn_pool import ES_CONN
+
+        self.logger = logging.getLogger(logger_name)
+
+        self.info = {}
+        self.logger.info(f"Use Elasticsearch {settings.ES['hosts']} as the doc engine.")
+        self.es = ES_CONN.get_conn()
+        fp_mapping = os.path.join(get_project_base_directory(), "conf", mapping_file_name)
+        if not os.path.exists(fp_mapping):
+            msg = f"Elasticsearch mapping file not found at {fp_mapping}"
+            self.logger.error(msg)
+            raise Exception(msg)
+        self.mapping = json.load(open(fp_mapping, "r"))
+        self.logger.info(f"Elasticsearch {settings.ES['hosts']} is healthy.")
+
+    def _connect(self):
+        from common.doc_store.es_conn_pool import ES_CONN
+
+        if self.es.ping():
+            return True
+        self.es = ES_CONN.refresh_conn()
+        return True
+
+    """
+    Database operations
+    """
+
+    def db_type(self) -> str:
+        return "elasticsearch"
+
+    def health(self) -> dict:
+        health_dict = dict(self.es.cluster.health())
+        health_dict["type"] = "elasticsearch"
+        return health_dict
+
+    def get_cluster_stats(self):
+        """
+        curl -XGET "http://{es_host}/_cluster/stats" -H "kbn-xsrf: reporting" to view raw stats.
+        """
+        raw_stats = self.es.cluster.stats()
+        self.logger.debug(f"ESConnection.get_cluster_stats: {raw_stats}")
+        try:
+            res = {
+                'cluster_name': raw_stats['cluster_name'],
+                'status': raw_stats['status']
+            }
+            indices_status = raw_stats['indices']
+            res.update({
+                'indices': indices_status['count'],
+                'indices_shards': indices_status['shards']['total']
+            })
+            doc_info = indices_status['docs']
+            res.update({
+                'docs': doc_info['count'],
+                'docs_deleted': doc_info['deleted']
+            })
+            store_info = indices_status['store']
+            res.update({
+                'store_size': convert_bytes(store_info['size_in_bytes']),
+                'total_dataset_size': convert_bytes(store_info['total_data_set_size_in_bytes'])
+            })
+            mappings_info = indices_status['mappings']
+            res.update({
+                'mappings_fields': mappings_info['total_field_count'],
+                'mappings_deduplicated_fields': mappings_info['total_deduplicated_field_count'],
+                'mappings_deduplicated_size': convert_bytes(mappings_info['total_deduplicated_mapping_size_in_bytes'])
+            })
+            node_info = raw_stats['nodes']
+            res.update({
+                'nodes': node_info['count']['total'],
+                'nodes_version': node_info['versions'],
+                'os_mem': convert_bytes(node_info['os']['mem']['total_in_bytes']),
+                'os_mem_used': convert_bytes(node_info['os']['mem']['used_in_bytes']),
+                'os_mem_used_percent': node_info['os']['mem']['used_percent'],
+                'jvm_versions': node_info['jvm']['versions'][0]['vm_version'],
+                'jvm_heap_used': convert_bytes(node_info['jvm']['mem']['heap_used_in_bytes']),
+                'jvm_heap_max': convert_bytes(node_info['jvm']['mem']['heap_max_in_bytes'])
+            })
+            return res
+
+        except Exception as e:
+            self.logger.exception(f"ESConnection.get_cluster_stats: {e}")
+            return None
+
+    """
+    Table operations
+    """
+
+    def create_idx(self, index_name: str, dataset_id: str, vector_size: int):
+        if self.index_exist(index_name, dataset_id):
+            return True
+        try:
+            from elasticsearch.client import IndicesClient
+            return IndicesClient(self.es).create(index=index_name,
+                                                 settings=self.mapping["settings"],
+                                                 mappings=self.mapping["mappings"])
+        except Exception:
+            self.logger.exception("ESConnection.createIndex error %s" % index_name)
+
+    def delete_idx(self, index_name: str, dataset_id: str):
+        if len(dataset_id) > 0:
+            # The index need to be alive after any kb deletion since all kb under this tenant are in one index.
+            return
+        try:
+            self.es.indices.delete(index=index_name, allow_no_indices=True)
+        except NotFoundError:
+            pass
+        except Exception:
+            self.logger.exception("ESConnection.deleteIdx error %s" % index_name)
+
+    def index_exist(self, index_name: str, dataset_id: str = None) -> bool:
+        s = Index(index_name, self.es)
+        for i in range(ATTEMPT_TIME):
+            try:
+                return s.exists()
+            except ConnectionTimeout:
+                self.logger.exception("ES request timeout")
+                time.sleep(3)
+                self._connect()
+                continue
+            except Exception as e:
+                self.logger.exception(e)
+                break
+        return False
+
+    """
+    CRUD operations
+    """
+
+    def get(self, doc_id: str, index_name: str, dataset_ids: list[str]) -> dict | None:
+        for i in range(ATTEMPT_TIME):
+            try:
+                res = self.es.get(index=index_name,
+                                  id=doc_id, source=True, )
+                if str(res.get("timed_out", "")).lower() == "true":
+                    raise Exception("Es Timeout.")
+                doc = res["_source"]
+                doc["id"] = doc_id
+                return doc
+            except NotFoundError:
+                return None
+            except Exception as e:
+                self.logger.exception(f"ESConnection.get({doc_id}) got exception")
+                raise e
+        self.logger.error(f"ESConnection.get timeout for {ATTEMPT_TIME} times!")
+        raise Exception("ESConnection.get timeout.")
+
+    @abstractmethod
+    def search(
+            self, select_fields: list[str],
+            highlight_fields: list[str],
+            condition: dict,
+            match_expressions: list[MatchExpr],
+            order_by: OrderByExpr,
+            offset: int,
+            limit: int,
+            index_names: str | list[str],
+            dataset_ids: list[str],
+            agg_fields: list[str] | None = None,
+            rank_feature: dict | None = None
+    ):
+        raise NotImplementedError("Not implemented")
+
+    @abstractmethod
+    def insert(self, documents: list[dict], index_name: str, dataset_id: str = None) -> list[str]:
+        raise NotImplementedError("Not implemented")
+
+    @abstractmethod
+    def update(self, condition: dict, new_value: dict, index_name: str, dataset_id: str) -> bool:
+        raise NotImplementedError("Not implemented")
+
+    @abstractmethod
+    def delete(self, condition: dict, index_name: str, dataset_id: str) -> int:
+        raise NotImplementedError("Not implemented")
+
+    """
+    Helper functions for search result
+    """
+
+    def get_total(self, res):
+        if isinstance(res["hits"]["total"], type({})):
+            return res["hits"]["total"]["value"]
+        return res["hits"]["total"]
+
+    def get_doc_ids(self, res):
+        return [d["_id"] for d in res["hits"]["hits"]]
+
+    def _get_source(self, res):
+        rr = []
+        for d in res["hits"]["hits"]:
+            d["_source"]["id"] = d["_id"]
+            d["_source"]["_score"] = d["_score"]
+            rr.append(d["_source"])
+        return rr
+
+    @abstractmethod
+    def get_fields(self, res, fields: list[str]) -> dict[str, dict]:
+        raise NotImplementedError("Not implemented")
+
+    def get_highlight(self, res, keywords: list[str], field_name: str):
+        ans = {}
+        for d in res["hits"]["hits"]:
+            highlights = d.get("highlight")
+            if not highlights:
+                continue
+            txt = "...".join([a for a in list(highlights.items())[0][1]])
+            if not is_english(txt.split()):
+                ans[d["_id"]] = txt
+                continue
+
+            txt = d["_source"][field_name]
+            txt = re.sub(r"[\r\n]", " ", txt, flags=re.IGNORECASE | re.MULTILINE)
+            txt_list = []
+            for t in re.split(r"[.?!;\n]", txt):
+                for w in keywords:
+                    t = re.sub(r"(^|[ .?/'\"\(\)!,:;-])(%s)([ .?/'\"\(\)!,:;-])" % re.escape(w), r"\1<em>\2</em>\3", t,
+                               flags=re.IGNORECASE | re.MULTILINE)
+                if not re.search(r"<em>[^<>]+</em>", t, flags=re.IGNORECASE | re.MULTILINE):
+                    continue
+                txt_list.append(t)
+            ans[d["_id"]] = "...".join(txt_list) if txt_list else "...".join([a for a in list(highlights.items())[0][1]])
+
+        return ans
+
+    def get_aggregation(self, res, field_name: str):
+        agg_field = "aggs_" + field_name
+        if "aggregations" not in res or agg_field not in res["aggregations"]:
+            return list()
+        buckets = res["aggregations"][agg_field]["buckets"]
+        return [(b["key"], b["doc_count"]) for b in buckets]
+
+    """
+    SQL
+    """
+
+    def sql(self, sql: str, fetch_size: int, format: str):
+        self.logger.debug(f"ESConnection.sql get sql: {sql}")
+        sql = re.sub(r"[ `]+", " ", sql)
+        sql = sql.replace("%", "")
+        replaces = []
+        for r in re.finditer(r" ([a-z_]+_l?tks)( like | ?= ?)'([^']+)'", sql):
+            fld, v = r.group(1), r.group(3)
+            match = " MATCH({}, '{}', 'operator=OR;minimum_should_match=30%') ".format(
+                fld, rag_tokenizer.fine_grained_tokenize(rag_tokenizer.tokenize(v)))
+            replaces.append(
+                ("{}{}'{}'".format(
+                    r.group(1),
+                    r.group(2),
+                    r.group(3)),
+                 match))
+
+        for p, r in replaces:
+            sql = sql.replace(p, r, 1)
+        self.logger.debug(f"ESConnection.sql to es: {sql}")
+
+        for i in range(ATTEMPT_TIME):
+            try:
+                res = self.es.sql.query(body={"query": sql, "fetch_size": fetch_size}, format=format,
+                                        request_timeout="2s")
+                return res
+            except ConnectionTimeout:
+                self.logger.exception("ES request timeout")
+                time.sleep(3)
+                self._connect()
+                continue
+            except Exception as e:
+                self.logger.exception(f"ESConnection.sql got exception. SQL:\n{sql}")
+                raise Exception(f"SQL error: {e}\n\nSQL: {sql}")
+        self.logger.error(f"ESConnection.sql timeout for {ATTEMPT_TIME} times!")
+        return None