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

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 \

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:v0.23.0">
+        <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
 
@@ -188,15 +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 `v0.23.0` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.23.0`, 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 v0.23.0
+
+   # 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
 
@@ -233,7 +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.
@@ -303,6 +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:
@@ -387,7 +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:v0.23.0">
+        <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
 
@@ -188,12 +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 v0.23.0 dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.23.0, 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 v0.23.0
+
+   # 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.
 
@@ -233,7 +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.
@@ -277,6 +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:
@@ -359,7 +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:v0.23.0">
+        <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
 
@@ -168,12 +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 v0.23.0 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。v0.23.0 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 v0.23.0
+   # 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。
 
@@ -194,8 +194,8 @@
 
 > `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
    ```
@@ -277,6 +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:
@@ -359,7 +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:v0.23.0">
+        <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
 
@@ -170,12 +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 v0.23.0 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.23.0과 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 v0.23.0
+
+   # 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.
 
@@ -214,7 +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.
@@ -271,6 +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:
@@ -363,7 +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:v0.23.0">
+        <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
 
@@ -188,12 +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`v0.23.0` 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 `v0.23.0`, 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 v0.23.0
+
+   # 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.
 
@@ -232,7 +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.
 
@@ -294,6 +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:
@@ -376,7 +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:v0.23.0">
+        <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
 
@@ -125,7 +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**
 
@@ -187,12 +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 `v0.23.0`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.23.0` 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 v0.23.0
+
+   # 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。
 
@@ -237,7 +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。
 
@@ -303,6 +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：
@@ -390,7 +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:v0.23.0">
+        <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
 
@@ -188,12 +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 `v0.23.0`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.23.0` 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 v0.23.0
+
+   # 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。
 
@@ -204,7 +204,7 @@
    # 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?        |
@@ -238,7 +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）。
@@ -302,6 +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：
@@ -393,7 +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
 

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==0.23.0
+    pip install ragflow-cli==0.23.1
     ```
 3.  Launch the CLI client:
     ```bash

admin/client/admin_client.py

@@ -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)*
@@ -249,6 +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
 
@@ -286,6 +300,43 @@ def encode_to_base64(input_string):
     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__()
@@ -370,7 +421,7 @@ class AdminCLI(Cmd):
                     res_json = response.json()
                     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/0.23.0"})
+                        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:
@@ -566,6 +617,10 @@ class AdminCLI(Cmd):
                 self._show_user_permission(command_dict)
             case "show_version":
                 self._show_version(command_dict)
+            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 _:
@@ -698,6 +753,33 @@ 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"]
         user_name: str = username_tree.children[0].strip("'\"")
@@ -873,36 +955,12 @@ class AdminCLI(Cmd):
         args = command.get("args", [])
 
         if meta_command in ["?", "h", "help"]:
-            self.show_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

admin/client/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ragflow-cli"
-version = "0.23.0"
+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 = "0.23.0"
+version = "0.23.1"
 source = { virtual = "." }
 dependencies = [
     { name = "beartype" },

admin/server/routes.py

@@ -158,6 +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/component/message.py

@@ -33,7 +33,7 @@ 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 save_to_memory
+from api.db.joint_services.memory_message_service import queue_save_to_memory_task
 
 
 class MessageParam(ComponentParamBase):
@@ -437,17 +437,4 @@ class Message(ComponentBase):
             "user_input": self._canvas.get_sys_query(),
             "agent_response": content
         }
-        res = []
-        for memory_id in self._param.memory_ids:
-            success, msg = await save_to_memory(memory_id, message_dict)
-            res.append({
-                "memory_id": memory_id,
-                "success": success,
-                "msg": msg
-            })
-        if all([r["success"] for r in res]):
-            return True, "Successfully added to memories."
-
-        error_text = "Some messages failed to add. " + " ".join([f"Add to memory {r['memory_id']} failed, detail: {r['msg']}" for r in res if not r["success"]])
-        logging.error(error_text)
-        return False, error_text
+        return await queue_save_to_memory_task(self._param.memory_ids, message_dict)

agent/tools/retrieval.py

@@ -202,7 +202,7 @@ class Retrieval(ToolBase, ABC):
             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,
+                ck = await settings.kg_retriever.retrieval(query,
                                                      [kb.tenant_id for kb in kbs],
                                                      kb_ids,
                                                      embd_mdl,
@@ -215,7 +215,7 @@ 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,
+            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

api/apps/chunk_app.py

@@ -381,7 +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

@@ -746,6 +746,7 @@ 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!")
+            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/memories_app.py

@@ -159,7 +159,8 @@ async def delete_memory(memory_id):
         return get_json_result(message=True, code=RetCode.NOT_FOUND)
     try:
         MemoryService.delete_memory(memory_id)
-        MessageService.delete_message({"memory_id": memory_id}, memory.tenant_id, 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)

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

@@ -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/session.py

@@ -60,7 +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.")
@@ -1116,7 +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/db/joint_services/memory_message_service.py

@@ -16,9 +16,14 @@
 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
@@ -82,32 +87,44 @@ async def save_to_memory(memory_id: str, message_dict: dict):
         "forget_at": None,
         "status": True
     } for content in extracted_content]]
-    embedding_model = LLMBundle(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(tenant_id, memory_id):
-        created = MessageService.create_index(tenant_id, memory_id, vector_size=vector_dimension)
-        if not created:
-            return False, "Failed to create message index."
+    return await embed_and_save(memory, message_list)
 
-    new_msg_size = sum([MessageService.calculate_message_size(m) for m in message_list])
-    current_memory_size = get_memory_size_cache(memory_id, tenant_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, tenant_id, size_to_delete)
-            MessageService.delete_message({"message_id": message_ids_to_delete}, 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, 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."
+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,
@@ -136,6 +153,36 @@ async def extract_by_llm(tenant_id: str, llm_id: str, extract_conf: dict, memory
     } 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: {
@@ -231,3 +278,112 @@ def init_memory_size_cache():
 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/services/dialog_service.py

@@ -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)

api/db/services/document_service.py

@@ -342,21 +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_doc_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)
@@ -378,6 +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):

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/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/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/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/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

@@ -129,14 +129,20 @@ class FileSource(StrEnum):
     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"
     DOWNLOAD = "Download"
     RAPTOR = "RAPTOR"
     GRAPH_RAG = "GraphRAG"
     MINDMAP = "Mindmap"
+    MEMORY = "Memory"
 
 
 VALID_PIPELINE_TASK_TYPES = {PipelineTaskType.PARSE, PipelineTaskType.DOWNLOAD, PipelineTaskType.RAPTOR,

common/data_source/__init__.py

@@ -34,9 +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 (
@@ -59,7 +61,6 @@ __all__ = [
     "JiraConnector",
     "SharePointConnector",
     "TeamsConnector",
-    "WebDAVConnector",
     "MoodleConnector",
     "BlobType",
     "DocumentSource",
@@ -73,4 +74,7 @@ __all__ = [
     "InsufficientPermissionsError",
     "UnexpectedValidationError",
     "AirtableConnector",
+    "AsanaConnector",
+    "ImapConnector",
+    "ZendeskConnector",
 ]

common/data_source/airtable_connector.py

@@ -1,6 +1,6 @@
 from datetime import datetime, timezone
 import logging
-from typing import Any
+from typing import Any, Generator
 
 import requests
 
@@ -8,8 +8,8 @@ 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
-from common.data_source.models import Document, GenerateDocumentsOutput
+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):
@@ -19,7 +19,7 @@ class AirtableClientNotSetUpError(PermissionError):
         )
 
 
-class AirtableConnector(LoadConnector):
+class AirtableConnector(LoadConnector, PollConnector):
     """
     Lightweight Airtable connector.
 
@@ -132,6 +132,26 @@ class AirtableConnector(LoadConnector):
         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
 

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/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
@@ -54,6 +57,13 @@ class DocumentSource(str, Enum):
     DROPBOX = "dropbox"
     BOX = "box"
     AIRTABLE = "airtable"
+    ASANA = "asana"
+    GITHUB = "github"
+    GITLAB = "gitlab"
+    IMAP = "imap"
+    BITBUCKET = "bitbucket"
+    ZENDESK = "zendesk"
+
 
 class FileOrigin(str, Enum):
     """File origins"""
@@ -231,6 +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"
@@ -256,6 +268,18 @@ 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/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/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

@@ -94,8 +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):

common/data_source/utils.py

@@ -1149,3 +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/metadata_utils.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import ast
 import logging
 from typing import Any, Callable, Dict
 
@@ -49,8 +50,8 @@ def meta_filter(metas: dict, filters: list[dict], logic: str = "and"):
                 try:
                     if isinstance(input, list):
                         input = input[0]
-                    input = float(input)
-                    value = float(value)
+                    input = ast.literal_eval(input)
+                    value = ast.literal_eval(value)
                 except Exception:
                     pass
             if isinstance(input, str):
@@ -58,28 +59,41 @@ def meta_filter(metas: dict, filters: list[dict], logic: str = "and"):
             if isinstance(value, str):
                 value = value.lower()
 
-            for conds in [
-                (operator == "contains", input in value if not isinstance(input, list) else all([i in value for i in input])),
-                (operator == "not contains", input not in value if not isinstance(input, list) else all([i not in value for i in input])),
-                (operator == "in", input in value if not isinstance(input, list) else all([i in value for i in input])),
-                (operator == "not in", input not in value if not isinstance(input, list) else all([i not in value for i in input])),
-                (operator == "start with", str(input).lower().startswith(str(value).lower())  if not isinstance(input, list) else "".join([str(i).lower() for i in input]).startswith(str(value).lower())),
-                (operator == "end with", str(input).lower().endswith(str(value).lower())  if not isinstance(input, list) else "".join([str(i).lower() for i in input]).endswith(str(value).lower())),
-                (operator == "empty", not input),
-                (operator == "not empty", input),
-                (operator == "=", input == value),
-                (operator == "≠", input != value),
-                (operator == ">", input > value),
-                (operator == "<", input < value),
-                (operator == "≥", input >= value),
-                (operator == "≤", input <= value),
-            ]:
-                try:
-                    if all(conds):
-                        ids.extend(docids)
-                        break
-                except Exception:
-                    pass
+            matched = False
+            try:
+                if operator == "contains":
+                    matched = input in value if not isinstance(input, list) else all(i in value for i in input)
+                elif operator == "not contains":
+                    matched = input not in value if not isinstance(input, list) else all(i not in value for i in input)
+                elif operator == "in":
+                    matched = input in value if not isinstance(input, list) else all(i in value for i in input)
+                elif operator == "not in":
+                    matched = input not in value if not isinstance(input, list) else all(i not in value for i in input)
+                elif operator == "start with":
+                    matched = str(input).lower().startswith(str(value).lower()) if not isinstance(input, list) else "".join([str(i).lower() for i in input]).startswith(str(value).lower())
+                elif operator == "end with":
+                    matched = str(input).lower().endswith(str(value).lower()) if not isinstance(input, list) else "".join([str(i).lower() for i in input]).endswith(str(value).lower())
+                elif operator == "empty":
+                    matched = not input
+                elif operator == "not empty":
+                    matched = bool(input)
+                elif operator == "=":
+                    matched = input == value
+                elif operator == "≠":
+                    matched = input != value
+                elif operator == ">":
+                    matched = input > value
+                elif operator == "<":
+                    matched = input < value
+                elif operator == "≥":
+                    matched = input >= value
+                elif operator == "≤":
+                    matched = input <= value
+            except Exception:
+                pass
+
+            if matched:
+                ids.extend(docids)
         return ids
 
     for k, v2docs in metas.items():

common/settings.py

@@ -170,7 +170,7 @@ def init_settings():
     global DATABASE_TYPE, DATABASE
     DATABASE_TYPE = os.getenv("DB_TYPE", "mysql")
     DATABASE = decrypt_database_config(name=DATABASE_TYPE)
-
+    
     global ALLOWED_LLM_FACTORIES, LLM_FACTORY, LLM_BASE_URL
     llm_settings = get_base_config("user_default_llm", {}) or {}
     llm_default_models = llm_settings.get("default_models", {}) or {}
@@ -334,6 +334,9 @@ def init_settings():
     DOC_BULK_SIZE = int(os.environ.get("DOC_BULK_SIZE", 4))
     EMBEDDING_BATCH_SIZE = int(os.environ.get("EMBEDDING_BATCH_SIZE", 16))
 
+    os.environ["DOTNET_SYSTEM_GLOBALIZATION_INVARIANT"] = "1"
+
+
 def check_and_install_torch():
     global PARALLEL_DEVICES
     try:

deepdoc/parser/docling_parser.py

@@ -78,14 +78,21 @@ class DoclingParser(RAGFlowPdfParser):
     def __images__(self, fnm, zoomin: int = 1, page_from=0, page_to=600, callback=None):
         self.page_from = page_from
         self.page_to = page_to
+        bytes_io = None
         try:
-            opener = pdfplumber.open(fnm) if isinstance(fnm, (str, PathLike)) else pdfplumber.open(BytesIO(fnm))
+            if not isinstance(fnm, (str, PathLike)):
+                bytes_io = BytesIO(fnm)
+
+            opener = pdfplumber.open(fnm) if isinstance(fnm, (str, PathLike)) else pdfplumber.open(bytes_io)
             with opener as pdf:
                 pages = pdf.pages[page_from:page_to]
                 self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).original for p in pages]
         except Exception as e:
             self.page_images = []
             self.logger.exception(e)
+        finally:
+            if bytes_io:
+                bytes_io.close()
 
     def _make_line_tag(self,bbox: _BBox) -> str:
         if bbox is None:

deepdoc/parser/pdf_parser.py

@@ -476,11 +476,13 @@ class RAGFlowPdfParser:
         self.boxes = bxs
 
     def _naive_vertical_merge(self, zoomin=3):
-        bxs = self._assign_column(self.boxes, zoomin)
+        #bxs = self._assign_column(self.boxes, zoomin)
+        bxs = self.boxes
 
         grouped = defaultdict(list)
         for b in bxs:
-            grouped[(b["page_number"], b.get("col_id", 0))].append(b)
+            # grouped[(b["page_number"], b.get("col_id", 0))].append(b)
+            grouped[(b["page_number"], "x")].append(b)
 
         merged_boxes = []
         for (pg, col), bxs in grouped.items():
@@ -551,7 +553,7 @@ class RAGFlowPdfParser:
 
             merged_boxes.extend(bxs)
 
-        self.boxes = sorted(merged_boxes, key=lambda x: (x["page_number"], x.get("col_id", 0), x["top"]))
+        #self.boxes = sorted(merged_boxes, key=lambda x: (x["page_number"], x.get("col_id", 0), x["top"]))
 
     def _final_reading_order_merge(self, zoomin=3):
         if not self.boxes:
@@ -1061,8 +1063,8 @@ class RAGFlowPdfParser:
 
                     self.total_page = len(self.pdf.pages)
 
-        except Exception:
-            logging.exception("RAGFlowPdfParser __images__")
+        except Exception as e:
+            logging.exception(f"RAGFlowPdfParser __images__, exception: {e}")
         logging.info(f"__images__ dedupe_chars cost {timer() - start}s")
 
         self.outlines = []
@@ -1206,7 +1208,7 @@ class RAGFlowPdfParser:
         start = timer()
         self._text_merge()
         self._concat_downward()
-        #self._naive_vertical_merge(zoomin)
+        self._naive_vertical_merge(zoomin)
         if callback:
             callback(0.92, "Text merged ({:.2f}s)".format(timer() - start))
 

deepdoc/vision/operators.py

@@ -16,6 +16,7 @@
 
 import logging
 import sys
+import ast
 import six
 import cv2
 import numpy as np
@@ -108,7 +109,14 @@ class NormalizeImage:
 
     def __init__(self, scale=None, mean=None, std=None, order='chw', **kwargs):
         if isinstance(scale, str):
-            scale = eval(scale)
+            try:
+                scale = float(scale)
+            except ValueError:
+                if '/' in scale:
+                    parts = scale.split('/')
+                    scale = ast.literal_eval(parts[0]) / ast.literal_eval(parts[1])
+                else:
+                    scale = ast.literal_eval(scale)
         self.scale = np.float32(scale if scale is not None else 1.0 / 255.0)
         mean = mean if mean is not None else [0.485, 0.456, 0.406]
         std = std if std is not None else [0.229, 0.224, 0.225]

docker/.env

@@ -1,3 +1,10 @@
+# -----------------------------------------------------------------------------
+# SECURITY WARNING: DO NOT DEPLOY WITH DEFAULT PASSWORDS
+# For non-local deployments, please change all passwords (ELASTIC_PASSWORD, 
+# MYSQL_PASSWORD, MINIO_PASSWORD, etc.) to strong, unique values.
+# You can generate a random string using: openssl rand -hex 32
+# -----------------------------------------------------------------------------
+
 # ------------------------------
 # docker env var for specifying vector db type at startup
 # (based on the vector db type, the corresponding docker
@@ -30,6 +37,7 @@ ES_HOST=es01
 ES_PORT=1200
 
 # The password for Elasticsearch.
+# WARNING: Change this for production!
 ELASTIC_PASSWORD=infini_rag_flow
 
 # the hostname where OpenSearch service is exposed, set it not the same as elasticsearch
@@ -85,6 +93,7 @@ OB_DATAFILE_SIZE=${OB_DATAFILE_SIZE:-20G}
 OB_LOG_DISK_SIZE=${OB_LOG_DISK_SIZE:-20G}
 
 # The password for MySQL.
+# WARNING: Change this for production!
 MYSQL_PASSWORD=infini_rag_flow
 # The hostname where the MySQL service is exposed
 MYSQL_HOST=mysql
@@ -128,11 +137,11 @@ ADMIN_SVR_HTTP_PORT=9381
 SVR_MCP_PORT=9382
 
 # The RAGFlow Docker image to download. v0.22+ doesn't include embedding models.
-RAGFLOW_IMAGE=infiniflow/ragflow:v0.23.0
+RAGFLOW_IMAGE=infiniflow/ragflow:v0.23.1
 
 # If you cannot download the RAGFlow Docker image:
-# RAGFLOW_IMAGE=swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow:v0.23.0
-# RAGFLOW_IMAGE=registry.cn-hangzhou.aliyuncs.com/infiniflow/ragflow:v0.23.0
+# RAGFLOW_IMAGE=swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow:v0.23.1
+# RAGFLOW_IMAGE=registry.cn-hangzhou.aliyuncs.com/infiniflow/ragflow:v0.23.1
 #
 # - For the `nightly` edition, uncomment either of the following:
 # RAGFLOW_IMAGE=swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow:nightly

docker/README.md

@@ -77,7 +77,7 @@ The [.env](./.env) file contains important environment variables for Docker.
 - `SVR_HTTP_PORT`  
   The port used to expose RAGFlow's HTTP API service to the host machine, allowing **external** access to the service running inside the Docker container. Defaults to `9380`.
 - `RAGFLOW-IMAGE`  
-  The Docker image edition. Defaults to `infiniflow/ragflow:v0.23.0`. The RAGFlow Docker image does not include embedding models.
+  The Docker image edition. Defaults to `infiniflow/ragflow:v0.23.1`. The RAGFlow Docker image does not include embedding models.
 
   
 > [!TIP]  

docs/basics/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Basics",
+  "position": 2,
+  "link": {
+    "type": "generated-index",
+    "description": "Basic concepts."
+  }
+}

docs/basics/agent_context_engine.md

@@ -0,0 +1,61 @@
+---
+sidebar_position: 2
+slug: /what-is-agent-context-engine
+---
+
+# What is Agent context engine?
+
+From 2025, a silent revolution began beneath the dazzling surface of AI Agents. While the world marveled at agents that could write code, analyze data, and automate workflows, a fundamental bottleneck emerged: why do even the most advanced agents still stumble on simple questions, forget previous conversations, or misuse available tools?
+
+The answer lies not in the intelligence of the Large Language Model (LLM) itself, but in the quality of the Context it receives. An LLM, no matter how powerful, is only as good as the information we feed it. Today’s cutting-edge agents are often crippled by a cumbersome, manual, and error-prone process of context assembly—a process known as Context Engineering.
+
+This is where the Agent Context Engine comes in. It is not merely an incremental improvement but a foundational shift, representing the evolution of RAG from a singular technique into the core data and intelligence substrate for the entire Agent ecosystem.
+
+## Beyond the hype: The reality of today's "intelligent" Agents
+Today, the “intelligence” behind most AI Agents hides a mountain of human labor. Developers must:
+
+- Hand-craft elaborate prompt templates
+- Hard-code document-retrieval logic for every task
+- Juggle tool descriptions, conversation history, and knowledge snippets inside a tiny context window
+- Repeat the whole process for each new scenario
+
+This pattern is called Context Engineering. It is deeply tied to expert know-how, almost impossible to scale, and prohibitively expensive to maintain. When an enterprise needs to keep dozens of distinct agents alive, the artisanal workshop model collapses under its own weight.
+
+The mission of an Agent Context Engine is to turn Context Engineering from an “art” into an industrial-grade science.
+
+Deconstructing the Agent Context Engine
+So, what exactly is an Agent Context Engine? It is a unified, intelligent, and automated platform responsible for the end-to-end process of assembling the optimal context for an LLM or Agent at the moment of inference. It moves from artisanal crafting to industrialized production.
+At its core, an Agent Context Engine is built on a triumvirate of next-generation retrieval capabilities, seamlessly integrated into a single service layer:
+
+1. The Knowledge Core (Advanced RAG): This is the evolution of traditional RAG. It moves beyond simple chunk-and-embed to intelligently process static, private enterprise knowledge. Techniques like TreeRAG (building LLM-generated document outlines for "locate-then-expand" retrieval) and GraphRAG (extracting entity networks to find semantically distant connections) work to close the "semantic gap." The engine’s Ingestion Pipeline acts as the ETL for unstructured data, parsing multi-format documents and using LLMs to enrich content with summaries, metadata, and structure before indexing.
+
+2. The Memory Layer: An Agent’s intelligence is defined by its ability to learn from interaction. The Memory Layer is a specialized retrieval system for dynamic, episodic data: conversation history, user preferences, and the agent’s own internal state (e.g., "waiting for human input"). It manages the lifecycle of this data—storing raw dialogue, triggering summarization into semantic memory, and retrieving relevant past interactions to provide continuity and personalization. Technologically, it is a close sibling to RAG, but focused on a temporal stream of data.
+
+3. The Tool Orchestrator: As MCP (Model Context Protocol) enables the connection of hundreds of internal services as tools, a new problem arises: tool selection. The Context Engine solves this with Tool Retrieval. Instead of dumping all tool descriptions into the prompt, it maintains an index of tools and—critically—an index of Playbooks or Guidelines (best practices on when and how to use tools). For a given task, it retrieves only the most relevant tools and instructions, transforming the LLM’s job from "searching a haystack" to "following a recipe."
+
+## Why we need a dedicated engine? The case for a unified substrate
+
+The necessity of an Agent Context Engine becomes clear when we examine the alternative: siloed, manually wired components.
+
+- The Data Silo Problem: Knowledge, memory, and tools reside in separate systems, requiring complex integration for each new agent.
+- The Assembly Line Bottleneck: Developers spend more time on context plumbing than on agent logic, slowing innovation to a crawl.
+- The "Context Ownership" Dilemma: In manually engineered systems, context logic is buried in code, owned by developers, and opaque to business users. An Engine makes context a configurable, observable, and customer-owned asset.
+
+The shift from Context Engineering to a Context Platform/Engine marks the maturation of enterprise AI, as summarized in the table below:
+
+| Dimension           | Context engineering (present)                                              | Context engineering/Platform (future)                                                               |
+| ------------------- | -------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------- |
+| Context creation    | Manual, artisanal work by developers and prompt engineers.                 | Automated, driven by intelligent ingestion pipelines and configurable rules.                        |
+| Context delivery    | Hard-coded prompts and static retrieval logic embedded in agent workflows. | Dynamic, real-time retrieval and assembly based on the agent's live state and intent.               |
+| Context maintenance | A development and operational burden, logic locked in code.                | A manageable platform function, with visibility and control returned to the business.               |
+
+
+## RAGFlow: A resolute march toward the context engine of Agents
+
+This is the future RAGFlow is forging.
+
+We left behind the label of “yet another RAG system” long ago. From DeepDoc—our deeply-optimized, multimodal document parser—to the bleeding-edge architectures that bridge semantic chasms in complex RAG scenarios, all the way to a full-blown, enterprise-grade ingestion pipeline, every evolutionary step RAGFlow takes is a deliberate stride toward the ultimate form: an Agentic Context Engine.
+
+We believe tomorrow’s enterprise AI advantage will hinge not on who owns the largest model, but on who can feed that model the highest-quality, most real-time, and most relevant context. An Agentic Context Engine is the critical infrastructure that turns this vision into reality.
+
+In the paradigm shift from “hand-crafted prompts” to “intelligent context,” RAGFlow is determined to be the most steadfast propeller and enabler. We invite every developer, enterprise, and researcher who cares about the future of AI agents to follow RAGFlow’s journey—so together we can witness and build the cornerstone of the next-generation AI stack.

docs/basics/rag.md

@@ -0,0 +1,107 @@
+---
+sidebar_position: 1
+slug: /what-is-rag
+---
+
+# What is Retreival-Augmented-Generation (RAG)？
+
+Since large language models (LLMs) became the focus of technology, their ability to handle general knowledge has been astonishing. However, when questions shift to internal corporate documents, proprietary knowledge bases, or real-time data, the limitations of LLMs become glaringly apparent: they cannot access private information outside their training data. Retrieval-Augmented Generation (RAG) was born precisely to address this core need. Before an LLM generates an answer, it first retrieves the most relevant context from an external knowledge base and inputs it as "reference material" to the LLM, thereby guiding it to produce accurate answers. In short, RAG elevates LLMs from "relying on memory" to "having evidence to rely on," significantly improving their accuracy and trustworthiness in specialized fields and real-time information queries.
+
+## Why RAG is important?
+
+Although LLMs excel in language understanding and generation, they have inherent limitations:
+
+- Static Knowledge: The model's knowledge is based on a data snapshot from its training time and cannot be automatically updated, making it difficult to perceive the latest information.
+- Blind Spot to External Data: They cannot directly access corporate private documents, real-time information streams, or domain-specific content.
+- Hallucination Risk: When lacking accurate evidence, they may still fabricate plausible-sounding but false answers to maintain conversational fluency.
+
+The introduction of RAG provides LLMs with real-time, credible "factual grounding." Its core mechanism is divided into two stages:
+
+- Retrieval Stage: Based on the user's question, quickly retrieve the most relevant documents or data fragments from an external knowledge base.
+- Generation Stage: The LLM organizes and generates the final answer by incorporating the retrieved information as context, combined with its own linguistic capabilities.
+
+This upgrades LLMs from "speaking from memory" to "speaking with documentation," significantly enhancing reliability in professional and enterprise-level applications.
+
+## How RAG works?
+
+Retrieval-Augmented Generation enables LLMs to generate higher-quality responses by leveraging real-time, external, or private data sources through the introduction of an information retrieval mechanism. Its workflow can be divided into following key steps:
+
+### Data processing and vectorization
+
+The knowledge required by RAG comes from unstructured data in various formats, such as documents, database records, or API return content. This data typically needs to be chunked, then transformed into vectors via an embedding model, and stored in a vector database.
+
+Why is Chunking Needed? Indexing entire documents directly faces the following problems:
+
+- Decreased Retrieval Precision: Vectorizing long documents leads to semantic "averaging," losing details.
+- Context Length Limitation: LLMs have a finite context window, requiring filtering of the most relevant parts for input.
+- Cost and Efficiency: Embedding computation and retrieval costs are higher for long texts.
+
+Therefore, an intelligent chunking strategy is key to balancing information integrity, retrieval granularity, and computational efficiency.
+
+### Retrieve relevant information
+
+The user's query is also converted into a vector to perform semantic relevance searches (e.g., calculating cosine similarity) in the vector database, matching and recalling the most relevant text fragments.
+
+### Context construction and answer generation
+
+The retrieved relevant content is added to the LLM's context as factual grounding, and the LLM finally generates the answer. Therefore, RAG can be seen as Context Engineering 1.0 for automated context construction.
+
+## Deep dive into existing RAG architecture: beyond vector retrieval
+
+An industrial-grade RAG system is far from being as simple as "vector search + LLM"; its complexity and challenges are primarily embedded in the retrieval process.
+
+### Data complexity: multimodal document processing
+
+Core Challenge: Corporate knowledge mostly exists in the form of multimodal documents containing text, charts, tables, and formulas. Simple OCR extraction loses a large amount of semantic information.
+
+Advanced Practice: Leading solutions, such as RAGFlow, tend to use Visual Language Models (VLM) or specialized parsing models like DeepDoc to "translate" multimodal documents into unimodal text rich in structural and semantic information. Converting multimodal information into high-quality unimodal text has become standard practice for advanced RAG.
+
+### The complexity of chunking: the trade-off between precision and context
+
+A simple "chunk-embed-retrieve" pipeline has an inherent contradiction:
+- Semantic Matching requires small text chunks to ensure clear semantic focus.
+- Context Understanding requires large text chunks to ensure complete and coherent information.
+
+This forces system design into a difficult trade-off between "precise but fragmented" and "complete but vague."
+
+Advanced Practice: Leading solutions, such as RAGFlow, employ semantic enhancement techniques like constructing semantic tables of contents and knowledge graphs. These not only address semantic fragmentation caused by physical chunking but also enable the discovery of relevant content across documents based on entity-relationship networks.
+
+### Why is a vector database insufficient for serving RAG?
+
+Vector databases excel at semantic similarity search, but RAG requires precise and reliable answers, demanding more capabilities from the retrieval system:
+- Hybrid Search: Relying solely on vector retrieval may miss exact keyword matches (e.g., product codes, regulation numbers). Hybrid search, combining vector retrieval with keyword retrieval (BM25), ensures both semantic breadth and keyword precision.
+- Tensor or Multi-Vector Representation: To support cross-modal data, employing tensor or multi-vector representation has become an important trend.
+- Metadata Filtering: Filtering based on attributes like date, department, and type is a rigid requirement in business scenarios.
+
+Therefore, the retrieval layer of RAG is a composite system based on vector search but must integrate capabilities like full-text search, re-ranking, and metadata filtering.
+
+## RAG and memory: Retrieval from the same source but different streams
+
+Within the agent framework, the essence of the memory mechanism is the same as RAG: both retrieve relevant information from storage based on current needs. The key difference lies in the data source:
+- RAG: Targets pre-existing static or dynamic private data provided by the user in advance (e.g., documents, databases).
+- Memory: Targets dynamic data generated or perceived by the agent in real-time during interaction (e.g., conversation history, environmental state, tool execution results).
+They are highly consistent at the technical base (e.g., vector retrieval, keyword matching) and can be seen as the same retrieval capability applied in different scenarios ("existing knowledge" vs. "interaction memory"). A complete agent system often includes both an RAG module for inherent knowledge and a Memory module for interaction history.
+
+## RAG applications
+
+RAG has demonstrated clear value in several typical scenarios:
+
+1. Enterprise Knowledge Q&A and Internal Search  
+   By vectorizing corporate private data and combining it with an LLM, RAG can directly return natural language answers based on authoritative sources, rather than document lists. While meeting intelligent Q&A needs, it inherently aligns with corporate requirements for data security, access control, and compliance.
+2. Complex Document Understanding and Professional Q&A  
+   For structurally complex documents like contracts and regulations, the value of RAG lies in its ability to generate accurate, verifiable answers while maintaining context integrity. Its system accuracy largely depends on text chunking and semantic understanding strategies.
+3. Dynamic Knowledge Fusion and Decision Support  
+   In business scenarios requiring the synthesis of information from multiple sources, RAG evolves into a knowledge orchestration and reasoning support system for business decisions. Through a multi-path recall mechanism, it fuses knowledge from different systems and formats, maintaining factual consistency and logical controllability during the generation phase.
+
+## The future of RAG
+
+The evolution of RAG is unfolding along several clear paths:
+
+1. RAG as the data foundation for Agents  
+   RAG and agents have an architecture vs. scenario relationship. For agents to achieve autonomous and reliable decision-making and execution, they must rely on accurate and timely knowledge. RAG provides them with a standardized capability to access private domain knowledge and is an inevitable choice for building knowledge-aware agents.
+2. Advanced RAG: Using LLMs to optimize retrieval itself  
+   The core feature of next-generation RAG is fully utilizing the reasoning capabilities of LLMs to optimize the retrieval process, such as rewriting queries, summarizing or fusing results, or implementing intelligent routing. Empowering every aspect of retrieval with LLMs is key to breaking through current performance bottlenecks.
+3. Towards context engineering 2.0  
+   Current RAG can be viewed as Context Engineering 1.0, whose core is assembling static knowledge context for single Q&A tasks. The forthcoming Context Engineering 2.0 will extend with RAG technology at its core, becoming a system that automatically and dynamically assembles comprehensive context for agents. The context fused by this system will come not only from documents but also include interaction memory, available tools/skills, and real-time environmental information. This marks the transition of agent development from a "handicraft workshop" model to the industrial starting point of automated context engineering.
+
+The essence of RAG is to build a dedicated, efficient, and trustworthy external data interface for large language models; its core is Retrieval, not Generation. Starting from the practical need to solve private data access, its technical depth is reflected in the optimization of retrieval for complex unstructured data. With its deep integration into agent architectures and its development towards automated context engineering, RAG is evolving from a technology that improves Q&A quality into the core infrastructure for building the next generation of trustworthy, controllable, and scalable intelligent applications.

docs/configurations.md

@@ -99,7 +99,7 @@ RAGFlow utilizes MinIO as its object storage solution, leveraging its scalabilit
 - `SVR_HTTP_PORT`
   The port used to expose RAGFlow's HTTP API service to the host machine, allowing **external** access to the service running inside the Docker container. Defaults to `9380`.
 - `RAGFLOW-IMAGE`
-  The Docker image edition. Defaults to `infiniflow/ragflow:v0.23.0` (the RAGFlow Docker image without embedding models).
+  The Docker image edition. Defaults to `infiniflow/ragflow:v0.23.1` (the RAGFlow Docker image without embedding models).
 
 :::tip NOTE
 If you cannot download the RAGFlow Docker image, try the following mirrors.

docs/develop/build_docker_image.mdx

@@ -47,7 +47,7 @@ After building the infiniflow/ragflow:nightly image, you are ready to launch a f
 
 1. Edit Docker Compose Configuration
 
-Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.23.0` to `infiniflow/ragflow:nightly` to use the pre-built image.
+Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.23.1` to `infiniflow/ragflow:nightly` to use the pre-built image.
 
 
 2. Launch the Service

docs/guides/admin/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Administration",
+  "position": 6,
+  "link": {
+    "type": "generated-index",
+    "description": "RAGFlow administration"
+  }
+}

docs/guides/admin/admin_cli.md

@@ -1,43 +1,11 @@
 ---
-sidebar_position: 6
-slug: /manage_users_and_services
+sidebar_position: 2
+slug: /admin_cli
 ---
 
+# Admin CLI
 
-# Admin CLI and Admin Service
-
-
-
-The Admin CLI and Admin Service form a client-server architectural suite for RAGFlow system administration. The Admin CLI serves as an interactive command-line interface that receives instructions and displays execution results from the Admin Service in real-time. This duo enables real-time monitoring of system operational status, supporting visibility into RAGFlow Server services and dependent components including MySQL, Elasticsearch, Redis, and MinIO. In administrator mode, they provide user management capabilities that allow viewing users and performing critical operations—such as user creation, password updates, activation status changes, and comprehensive user data deletion—even when corresponding web interface functionalities are disabled.
-
-
-
-## Starting the Admin Service
-
-### Launching from source code
-
-1. Before start Admin Service, please make sure RAGFlow system is already started.
-
-2. Launch from source code:
-
-   ```bash
-   python admin/server/admin_server.py
-   ```
-
-   The service will start and listen for incoming connections from the CLI on the configured port. 
-
-### Using docker image
-
-1. Before startup, please configure the `docker_compose.yml`  file to enable admin server:
-
-   ```bash
-   command:
-     - --enable-adminserver
-   ```
-
-2. Start the containers, the service will start and listen for incoming connections from the CLI on the configured port.
-
-
+The RAGFlow Admin CLI is a command-line-based system administration tool that offers administrators an efficient and flexible method for system interaction and control. Operating on a client-server architecture, it communicates in real-time with the Admin Service, receiving administrator commands and dynamically returning execution results.
 
 ## Using the Admin CLI
 
@@ -46,7 +14,7 @@ The Admin CLI and Admin Service form a client-server architectural suite for RAG
 2. Install ragflow-cli.
 
    ```bash
-   pip install ragflow-cli==0.23.0
+   pip install ragflow-cli==0.23.1
    ```
 
 3. Launch the CLI client:

docs/guides/admin/admin_service.md

@@ -0,0 +1,39 @@
+---
+sidebar_position: 0
+slug: /admin_service
+---
+
+
+# Admin Service
+
+The Admin Service is the core backend management service of the RAGFlow system, providing comprehensive system administration capabilities through centralized API interfaces for managing and controlling the entire platform. Adopting a client-server architecture, it supports access and operations via both a Web UI and an Admin CLI, ensuring flexible and efficient execution of administrative tasks.
+
+The core functions of the Admin Service include real-time monitoring of the operational status of the RAGFlow server and its critical dependent components—such as MySQL, Elasticsearch, Redis, and MinIO—along with full-featured user management. In administrator mode, it enables key operations such as viewing user information, creating users, updating passwords, modifying activation status, and performing complete user data deletion. These functions remain accessible via the Admin CLI even when the web management interface is disabled, ensuring the system stays under control at all times.
+
+With its unified interface design, the Admin Service combines the convenience of visual administration with the efficiency and stability of command-line operations, serving as a crucial foundation for the reliable operation and secure management of the RAGFlow system.
+
+## Starting the Admin Service
+
+### Launching from source code
+
+1. Before start Admin Service, please make sure RAGFlow system is already started.
+
+2. Launch from source code:
+
+   ```bash
+   python admin/server/admin_server.py
+   ```
+
+   The service will start and listen for incoming connections from the CLI on the configured port. 
+
+### Using docker image
+
+1. Before startup, please configure the `docker_compose.yml`  file to enable admin server:
+
+   ```bash
+   command:
+     - --enable-adminserver
+   ```
+
+2. Start the containers, the service will start and listen for incoming connections from the CLI on the configured port.
+

docs/guides/admin/admin_ui.md

@@ -1,6 +1,6 @@
 ---
-sidebar_position: 7
-slug: /accessing_admin_ui
+sidebar_position: 1
+slug: /admin_ui
 ---
 
 # Admin UI

docs/guides/dataset/auto_metadata.md

@@ -34,7 +34,7 @@ Enabling TOC extraction requires significant memory, computational resources, an
 
 ![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/auto_metadata_settings.png)
 
-3. Click **+** to add new fields and enter the congiruation page.
+3. Click **+** to add new fields and enter the configuration page.
 
 ![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/metadata_field_settings.png)
 

docs/guides/dataset/configure_knowledge_base.md

@@ -133,7 +133,7 @@ See [Run retrieval test](./run_retrieval_test.md) for details.
 
 ## Search for dataset
 
-As of RAGFlow v0.23.0, the search feature is still in a rudimentary form, supporting only dataset search by name.
+As of RAGFlow v0.23.1, the search feature is still in a rudimentary form, supporting only dataset search by name.
 
 ![search dataset](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/search_datasets.jpg)
 

docs/guides/manage_files.md

@@ -87,4 +87,4 @@ RAGFlow's file management allows you to download an uploaded file:
 
 ![download_file](https://github.com/infiniflow/ragflow/assets/93570324/cf3b297f-7d9b-4522-bf5f-4f45743e4ed5)
 
-> As of RAGFlow v0.23.0, bulk download is not supported, nor can you download an entire folder. 
+> As of RAGFlow v0.23.1, bulk download is not supported, nor can you download an entire folder. 

docs/guides/migration/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Migration",
+  "position": 5,
+  "link": {
+    "type": "generated-index",
+    "description": "RAGFlow migration guide"
+  }
+}

docs/guides/models/deploy_local_llm.mdx

@@ -340,13 +340,13 @@ Application startup complete.
 
 setting->model providers->search->vllm->add ,configure as follow:
 
-![add vllm](https://github.com/user-attachments/assets/6f1d9f1a-3507-465b-87a3-4427254fff86)
+![add vllm](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/ragflow_vllm.png)
 
 select vllm chat model as default llm model as follow:
-![chat](https://github.com/user-attachments/assets/05efbd4b-2c18-4c6b-8d1c-52bae712372d)
+![chat](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/ragflow_vllm1.png)
 ### 5.3 chat with vllm chat model
 create chat->create conversations-chat as follow:
-![chat](https://github.com/user-attachments/assets/dc1885f6-23a9-48f1-8850-d5f59b5e8f67)
+![chat](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/ragflow_vllm2.png)
 
 
 

docs/guides/run_health_check.md

@@ -1,109 +0,0 @@
----
-sidebar_position: 8
-slug: /run_health_check
----
-
-# Monitoring
-
-Double-check the health status of RAGFlow's dependencies.
-
----
-
-The operation of RAGFlow depends on four services:
-
-- **Elasticsearch** (default) or [Infinity](https://github.com/infiniflow/infinity) as the document engine
-- **MySQL**
-- **Redis**
-- **MinIO** for object storage
-
-If an exception or error occurs related to any of the above services, such as `Exception: Can't connect to ES cluster`, refer to this document to check their health status.
-
-You can also click you avatar in the top right corner of the page **>** System to view the visualized health status of RAGFlow's core services. The following screenshot shows that all services are 'green' (running healthily). The task executor displays the *cumulative* number of completed and failed document parsing tasks from the past 30 minutes:
-
-![system_status_page](https://github.com/user-attachments/assets/b0c1a11e-93e3-4947-b17a-1bfb4cdab6e4)
-
-Services with a yellow or red light are not running properly. The following is a screenshot of the system page after running `docker stop ragflow-es-10`:
-
-![es_failed](https://github.com/user-attachments/assets/06056540-49f5-48bf-9cc9-a7086bc75790)
-
-You can click on a specific 30-second time interval to view the details of completed and failed tasks:
-
-![done_tasks](https://github.com/user-attachments/assets/49b25ec4-03af-48cf-b2e5-c892f6eaa261)
-
-![done_vs_failed](https://github.com/user-attachments/assets/eaa928d0-a31c-4072-adea-046091e04599)
-
-## API Health Check
-
-In addition to checking the system dependencies from the **avatar > System** page in the UI, you can directly query the backend health check endpoint:
-
-```bash
-http://IP_OF_YOUR_MACHINE/v1/system/healthz
-```
-
-Here `<port>` refers to the actual port of your backend service (e.g., `7897`, `9222`, etc.).
-
-Key points:
-- **No login required** (no `@login_required` decorator)
-- Returns results in JSON format
-- If all dependencies are healthy → HTTP **200 OK**
-- If any dependency fails → HTTP **500 Internal Server Error**
-
-### Example 1: All services healthy (HTTP 200)
-
-```bash
-http://127.0.0.1/v1/system/healthz
-```
-
-Response:
-
-```http
-HTTP/1.1 200 OK
-Content-Type: application/json
-Content-Length: 120
-
-```
-
-Explanation:
-- Database (MySQL/Postgres), Redis, document engine (Elasticsearch/Infinity), and object storage (MinIO) are all healthy.
-- The `status` field returns `"ok"`.
-
-### Example 2: One service unhealthy (HTTP 500)
-
-For example, if Redis is down:
-
-Response:
-
-```http
-HTTP/1.1 500 INTERNAL SERVER ERROR
-Content-Type: application/json
-Content-Length: 300
-
-```
-
-Explanation:
-- `redis` is marked as `"nok"`, with detailed error info under `_meta.redis.error`.
-- The overall `status` is `"nok"`, so the endpoint returns 500.
-
----
-
-This endpoint allows you to monitor RAGFlow’s core dependencies programmatically in scripts or external monitoring systems, without relying on the frontend UI.
-  "redis": "nok",
-  "doc_engine": "ok",
-  "storage": "ok",
-  "status": "nok",
-  "_meta": {
-    "redis": {
-      "elapsed": "5.2",
-      "error": "Lost connection!"
-    }
-  }
-}
-```
-
-Explanation:
-- `redis` is marked as `"nok"`, with detailed error info under `_meta.redis.error`.
-- The overall `status` is `"nok"`, so the endpoint returns 500.
-
----
-
-This endpoint allows you to monitor RAGFlow’s core dependencies programmatically in scripts or external monitoring systems, without relying on the frontend UI.

docs/guides/upgrade_ragflow.mdx

@@ -60,16 +60,16 @@ To upgrade RAGFlow, you must upgrade **both** your code **and** your Docker imag
    git pull
    ```
 
-3. Switch to the latest, officially published release, e.g., `v0.23.0`:
+3. Switch to the latest, officially published release, e.g., `v0.23.1`:
 
    ```bash
-   git checkout -f v0.23.0
+   git checkout -f v0.23.1
    ```
 
 4. Update **ragflow/docker/.env**:
 
    ```bash
-   RAGFLOW_IMAGE=infiniflow/ragflow:v0.23.0
+   RAGFLOW_IMAGE=infiniflow/ragflow:v0.23.1
    ```
 
 5. Update the RAGFlow image and restart RAGFlow:
@@ -90,10 +90,10 @@ No, you do not need to. Upgrading RAGFlow in itself will *not* remove your uploa
 1. From an environment with Internet access, pull the required Docker image.
 2. Save the Docker image to a **.tar** file.
    ```bash
-   docker save -o ragflow.v0.23.0.tar infiniflow/ragflow:v0.23.0
+   docker save -o ragflow.v0.23.1.tar infiniflow/ragflow:v0.23.1
    ```
 3. Copy the **.tar** file to the target server.
 4. Load the **.tar** file into Docker:
    ```bash
-   docker load -i ragflow.v0.23.0.tar
+   docker load -i ragflow.v0.23.1.tar
    ```

docs/quickstart.mdx

@@ -46,7 +46,7 @@ This section provides instructions on setting up the RAGFlow server on Linux. If
 
    `vm.max_map_count`. This value sets the maximum number of memory map areas a process may have. Its default value is 65530. While most applications require fewer than a thousand maps, reducing this value can result in abnormal behaviors, and the system will throw out-of-memory errors when a process reaches the limitation.
 
-   RAGFlow v0.23.0 uses Elasticsearch or [Infinity](https://github.com/infiniflow/infinity) for multiple recall. Setting the value of `vm.max_map_count` correctly is crucial to the proper functioning of the Elasticsearch component.
+   RAGFlow v0.23.1 uses Elasticsearch or [Infinity](https://github.com/infiniflow/infinity) for multiple recall. Setting the value of `vm.max_map_count` correctly is crucial to the proper functioning of the Elasticsearch component.
 
 <Tabs
   defaultValue="linux"
@@ -186,7 +186,7 @@ This section provides instructions on setting up the RAGFlow server on Linux. If
    ```bash
    $ git clone https://github.com/infiniflow/ragflow.git
    $ cd ragflow/docker
-   $ git checkout -f v0.23.0
+   $ git checkout -f v0.23.1
    ```
 
 3. Use the pre-built Docker images and start up the server:
@@ -202,7 +202,7 @@ This section provides instructions on setting up the RAGFlow server on Linux. If
 
    | RAGFlow image tag   | Image size (GB) | Stable?                  |
    | ------------------- | --------------- | ------------------------ |
-   | v0.23.0             | &approx;2       | Stable release           |
+   | v0.23.1             | &approx;2       | Stable release           |
    | nightly             | &approx;2       | _Unstable_ nightly build |
 
    ```mdx-code-block

docs/references/python_api_reference.md

@@ -1603,7 +1603,7 @@ In streaming mode, not all responses include a reference, as this depends on the
 
 ##### question: `str`
 
-The question to start an AI-powered conversation. Ifthe **Begin** component takes parameters, a question is not required.
+The question to start an AI-powered conversation. If the **Begin** component takes parameters, a question is not required.
 
 ##### stream: `bool`
 

docs/references/supported_models.mdx

@@ -13,61 +13,58 @@ A complete list of models supported by RAGFlow, which will continue to expand.
 <APITable>
 ```
 
-| Provider              | Chat               | Embedding          | Rerank             | Img2txt            | Speech2txt         | TTS                |
-| --------------------- | ------------------ | ------------------ | ------------------ | ------------------ | ------------------ | ------------------ |
-| Anthropic             | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Azure-OpenAI          | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
-| BAAI                  |                    | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |
-| BaiChuan              | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| BaiduYiyan            | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| Bedrock               | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| Cohere                | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| DeepSeek              | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| FastEmbed             |                    | :heavy_check_mark: |                    |                    |                    |                    |
-| Fish Audio            |                    |                    |                    |                    |                    | :heavy_check_mark: |
-| Gemini                | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
-| Google Cloud          | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| GPUStack              | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: |
-| Groq                  | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| HuggingFace           | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| Jina                  |                    | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |
-| LeptonAI              | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| LocalAI               | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
-| LM-Studio             | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
-| MiniMax               | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Mistral               | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| ModelScope            | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Moonshot              | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
-| Novita AI             | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| NVIDIA                | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| Ollama                | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
-| OpenAI                | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
-| OpenAI-API-Compatible | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| OpenRouter            | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
-| PerfXCloud            | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| Replicate             | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| PPIO                  | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| SILICONFLOW           | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| StepFun               | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Tencent Hunyuan       | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Tencent Cloud         |                    |                    |                    |                    | :heavy_check_mark: |                    |
-| TogetherAI            | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| Tongyi-Qianwen        | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
-| Upstage               | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| VLLM                  | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| VolcEngine            | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| Voyage AI             |                    | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| Xinference            | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |
-| XunFei Spark          | :heavy_check_mark: |                    |                    |                    |                    | :heavy_check_mark: |
-| xAI                   | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
-| Youdao                |                    | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |
-| ZHIPU-AI              | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
-| 01.AI                 | :heavy_check_mark: |                    |                    |                    |                    |                    |
-| DeepInfra             | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |
-| 302.AI                | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
-| CometAPI              | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |
-| DeerAPI               | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    | :heavy_check_mark: |
-| Jiekou.AI             | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |
+| Provider              | LLM                | Image2Text         | Speech2text        | TTS                | Embedding          | Rerank             | OCR                |
+| --------------------- | ------------------ | ------------------ | ------------------ | ------------------ | ------------------ | ------------------ | ------------------ |
+| Anthropic             | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Azure-OpenAI          | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: |                    |                    |
+| BaiChuan              | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| BaiduYiyan            | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| Bedrock               | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| Cohere                | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| DeepSeek              | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Fish Audio            |                    |                    |                    | :heavy_check_mark: |                    |                    |                    |
+| Gemini                | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
+| Google Cloud          | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| GPUStack              | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |
+| Groq                  | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| HuggingFace           | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| Jina                  |                    |                    |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| LocalAI               | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
+| LongCat               | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| LM-Studio             | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
+| MiniMax               | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| MinerU                |                    |                    |                    |                    |                    |                    | :heavy_check_mark: |
+| Mistral               | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| ModelScope            | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Moonshot              | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |                    |
+| NovitaAI              | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| NVIDIA                | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| Ollama                | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
+| OpenAI                | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
+| OpenAI-API-Compatible | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| OpenRouter            | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |                    |
+| Replicate             | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| PPIO                  | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| SILICONFLOW           | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| StepFun               | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Tencent Hunyuan       | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Tencent Cloud         |                    |                    | :heavy_check_mark: |                    |                    |                    |                    |
+| TogetherAI            | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| TokenPony             | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Tongyi-Qianwen        | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |
+| Upstage               | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| VLLM                  | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| VolcEngine            | :heavy_check_mark: |                    |                    |                    |                    |                    |                    |
+| Voyage AI             |                    | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| Xinference            | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |
+| XunFei Spark          | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |                    |
+| xAI                   | :heavy_check_mark: | :heavy_check_mark: |                    |                    |                    |                    |                    |
+| ZHIPU-AI              | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: |                    |                    |
+| DeepInfra             | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
+| 302.AI                | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
+| CometAPI              | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: |                    |                    |
+| DeerAPI               | :heavy_check_mark: | :heavy_check_mark: |                    | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
+| Jiekou.AI             | :heavy_check_mark: |                    |                    |                    | :heavy_check_mark: | :heavy_check_mark: |                    |
 
 ```mdx-code-block
 </APITable>

docs/release_notes.md

@@ -7,9 +7,34 @@ slug: /release_notes
 
 Key features, improvements and bug fixes in the latest releases.
 
+
+## v0.23.1
+
+Released on December 31, 2025.
+
+### Improvements
+
+- Memory: Enhances the stability of memory extraction when all memory types are selected.
+- RAG: Refines the context window extraction strategy for images and tables.
+
+
+### Fixed issues
+
+- Memory: 
+  - The RAGFlow server failed to start if an empty memory object existed.
+  - Unable to delete a newly created empty Memory.
+- RAG: MDX file parsing was not supported.
+
+### Data sources
+
+- GitHub
+- Gitlab
+- Asana
+- IMAP
+
 ## v0.23.0
 
-Released on December 29, 2025.
+Released on December 27, 2025.
 
 ### New features
 
@@ -32,7 +57,8 @@ Released on December 29, 2025.
 
 ### Improvements
 
-- Bumps RAGFlow's document engine, [Infinity](https://github.com/infiniflow/infinity) to v0.6.13 (backward compatible).
+- RAG: Accelerates GraphRAG generation significantly.
+- Bumps RAGFlow's document engine, [Infinity](https://github.com/infiniflow/infinity) to v0.6.15 (backward compatible).
 
 ### Data sources
 

graphrag/search.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import asyncio
 import json
 import logging
 from collections import defaultdict
@@ -32,21 +33,21 @@ from common.doc_store.doc_store_base import OrderByExpr
 
 
 class KGSearch(Dealer):
-    def _chat(self, llm_bdl, system, history, gen_conf):
+    async def _chat(self, llm_bdl, system, history, gen_conf):
         response = get_llm_cache(llm_bdl.llm_name, system, history, gen_conf)
         if response:
             return response
-        response = llm_bdl.chat(system, history, gen_conf)
+        response = await llm_bdl.async_chat(system, history, gen_conf)
         if response.find("**ERROR**") >= 0:
             raise Exception(response)
         set_llm_cache(llm_bdl.llm_name, system, response, history, gen_conf)
         return response
 
-    def query_rewrite(self, llm, question, idxnms, kb_ids):
+    async def query_rewrite(self, llm, question, idxnms, kb_ids):
         ty2ents = get_entity_type2samples(idxnms, kb_ids)
         hint_prompt = PROMPTS["minirag_query2kwd"].format(query=question,
                                                           TYPE_POOL=json.dumps(ty2ents, ensure_ascii=False, indent=2))
-        result = self._chat(llm, hint_prompt, [{"role": "user", "content": "Output:"}], {})
+        result = await self._chat(llm, hint_prompt, [{"role": "user", "content": "Output:"}], {})
         try:
             keywords_data = json_repair.loads(result)
             type_keywords = keywords_data.get("answer_type_keywords", [])
@@ -138,7 +139,7 @@ class KGSearch(Dealer):
                                        idxnms, kb_ids)
         return self._ent_info_from_(es_res, 0)
 
-    def retrieval(self, question: str,
+    async def retrieval(self, question: str,
                tenant_ids: str | list[str],
                kb_ids: list[str],
                emb_mdl,
@@ -158,7 +159,7 @@ class KGSearch(Dealer):
         idxnms = [index_name(tid) for tid in tenant_ids]
         ty_kwds = []
         try:
-            ty_kwds, ents = self.query_rewrite(llm, qst, [index_name(tid) for tid in tenant_ids], kb_ids)
+            ty_kwds, ents = await self.query_rewrite(llm, qst, [index_name(tid) for tid in tenant_ids], kb_ids)
             logging.info(f"Q: {qst}, Types: {ty_kwds}, Entities: {ents}")
         except Exception as e:
             logging.exception(e)
@@ -334,5 +335,5 @@ if __name__ == "__main__":
     embed_bdl = LLMBundle(args.tenant_id, LLMType.EMBEDDING, kb.embd_id)
 
     kg = KGSearch(settings.docStoreConn)
-    print(kg.retrieval({"question": args.question, "kb_ids": [kb_id]},
-                    search.index_name(kb.tenant_id), [kb_id], embed_bdl, llm_bdl))
+    print(asyncio.run(kg.retrieval({"question": args.question, "kb_ids": [kb_id]},
+                    search.index_name(kb.tenant_id), [kb_id], embed_bdl, llm_bdl)))

helm/README.md

@@ -0,0 +1,133 @@
+# RAGFlow Helm Chart
+
+A Helm chart to deploy RAGFlow and its dependencies on Kubernetes.
+
+- Components: RAGFlow (web/api) and optional dependencies (Infinity/Elasticsearch/OpenSearch, MySQL, MinIO, Redis)
+- Requirements: Kubernetes >= 1.24, Helm >= 3.10
+
+## Install
+
+```bash
+helm upgrade --install ragflow ./ \
+  --namespace ragflow --create-namespace
+```
+
+Uninstall:
+```bash
+helm uninstall ragflow -n ragflow
+```
+
+## Global Settings
+
+- `global.repo`: Prepend a global image registry prefix for all images.
+  - Behavior: Replaces the registry part and keeps the image path (e.g., `quay.io/minio/minio` -> `registry.example.com/myproj/minio/minio`).
+  - Example: `global.repo: "registry.example.com/myproj"`
+- `global.imagePullSecrets`: List of image pull secrets applied to all Pods.
+  - Example:
+    ```yaml
+    global:
+      imagePullSecrets:
+        - name: regcred
+    ```
+
+## External Services (MySQL / MinIO / Redis)
+
+The chart can deploy in-cluster services or connect to external ones. Toggle with `*.enabled`. When disabled, provide host/port via `env.*`.
+
+- MySQL
+  - `mysql.enabled`: default `true`
+  - If `false`, set:
+    - `env.MYSQL_HOST` (required), `env.MYSQL_PORT` (default `3306`)
+    - `env.MYSQL_DBNAME` (default `rag_flow`), `env.MYSQL_PASSWORD` (required)
+    - `env.MYSQL_USER` (default `root` if omitted)
+- MinIO
+  - `minio.enabled`: default `true`
+  - Configure:
+    - `env.MINIO_HOST` (optional external host), `env.MINIO_PORT` (default `9000`)
+    - `env.MINIO_ROOT_USER` (default `rag_flow`), `env.MINIO_PASSWORD` (optional)
+- Redis (Valkey)
+  - `redis.enabled`: default `true`
+  - If `false`, set:
+    - `env.REDIS_HOST` (required), `env.REDIS_PORT` (default `6379`)
+    - `env.REDIS_PASSWORD` (optional; empty disables auth if server allows)
+
+Notes:
+- When `*.enabled=true`, the chart renders in-cluster resources and injects corresponding `*_HOST`/`*_PORT` automatically.
+- Sensitive variables like `MYSQL_PASSWORD` are required; `MINIO_PASSWORD` and `REDIS_PASSWORD` are optional. All secrets are stored in a Secret.
+
+### Example: use external MySQL, MinIO, and Redis
+
+```yaml
+# values.override.yaml
+mysql:
+  enabled: false  # use external MySQL
+minio:
+  enabled: false  # use external MinIO (S3 compatible)
+redis:
+  enabled: false  # use external Redis/Valkey
+
+env:
+  # MySQL
+  MYSQL_HOST: mydb.example.com
+  MYSQL_PORT: "3306"
+  MYSQL_USER: root
+  MYSQL_DBNAME: rag_flow
+  MYSQL_PASSWORD: "<your-mysql-password>"
+
+  # MinIO
+  MINIO_HOST: s3.example.com
+  MINIO_PORT: "9000"
+  MINIO_ROOT_USER: rag_flow
+  MINIO_PASSWORD: "<your-minio-secret>"
+
+  # Redis
+  REDIS_HOST: redis.example.com
+  REDIS_PORT: "6379"
+  REDIS_PASSWORD: "<your-redis-pass>"
+```
+
+Apply:
+```bash
+helm upgrade --install ragflow ./helm -n ragflow -f values.override.yaml
+```
+
+## Document Engine Selection
+
+Choose one of `infinity` (default), `elasticsearch`, or `opensearch` via `env.DOC_ENGINE`. The chart renders only the selected engine and sets the appropriate host variables.
+
+```yaml
+env:
+  DOC_ENGINE: infinity   # or: elasticsearch | opensearch
+  # For elasticsearch
+  ELASTIC_PASSWORD: "<es-pass>"
+  # For opensearch
+  OPENSEARCH_PASSWORD: "<os-pass>"
+```
+
+## Ingress
+
+Expose the web UI via Ingress:
+
+```yaml
+ingress:
+  enabled: true
+  className: nginx
+  hosts:
+    - host: ragflow.example.com
+      paths:
+        - path: /
+          pathType: Prefix
+```
+
+## Validate the Chart
+
+```bash
+helm lint ./helm
+helm template ragflow ./helm > rendered.yaml
+```
+
+## Notes
+
+- By default, the chart uses `DOC_ENGINE: infinity` and deploys in-cluster MySQL, MinIO, and Redis.
+- The chart injects derived `*_HOST`/`*_PORT` and required secrets into a single Secret (`<release>-ragflow-env-config`).
+- `global.repo` and `global.imagePullSecrets` apply to all Pods; per-component `*.image.pullSecrets` still work and are merged with global settings.

helm/templates/_helpers.tpl

@@ -42,6 +42,31 @@ app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
 app.kubernetes.io/managed-by: {{ .Release.Service }}
 {{- end }}
 
+{{/*
+Resolve image repository with optional global repo prefix.
+If .Values.global.repo is set, replace registry part and keep image path.
+Detect existing registry by first segment containing '.' or ':' or being 'localhost'.
+Usage: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.foo.image.repository) }}
+*/}}
+{{- define "ragflow.imageRepo" -}}
+{{- $root := .root -}}
+{{- $repo := .repo -}}
+{{- $global := $root.Values.global -}}
+{{- if and $global $global.repo }}
+  {{- $parts := splitList "/" $repo -}}
+  {{- $first := index $parts 0 -}}
+  {{- $hasRegistry := or (regexMatch "\\." $first) (regexMatch ":" $first) (eq $first "localhost") -}}
+  {{- if $hasRegistry -}}
+    {{- $path := join "/" (rest $parts) -}}
+    {{- printf "%s/%s" $global.repo $path -}}
+  {{- else -}}
+    {{- printf "%s/%s" $global.repo $repo -}}
+  {{- end -}}
+{{- else -}}
+  {{- $repo -}}
+{{- end -}}
+{{- end }}
+
 {{/*
 Selector labels
 */}}

helm/templates/elasticsearch.yaml

@@ -32,7 +32,7 @@ spec:
       {{- include "ragflow.selectorLabels" . | nindent 6 }}
       app.kubernetes.io/component: elasticsearch
   {{- with .Values.elasticsearch.deployment.strategy }}
-  strategy:
+  updateStrategy:
     {{- . | toYaml | nindent 4 }}
   {{- end }}
   template:
@@ -44,9 +44,9 @@ spec:
         checksum/config-es: {{ include (print $.Template.BasePath "/elasticsearch-config.yaml") . | sha256sum }}
         checksum/config-env: {{ include (print $.Template.BasePath "/env.yaml") . | sha256sum }}
     spec:
-      {{- if or .Values.imagePullSecrets .Values.elasticsearch.image.pullSecrets }}
+      {{- if or .Values.global.imagePullSecrets .Values.elasticsearch.image.pullSecrets }}
       imagePullSecrets:
-        {{- with .Values.imagePullSecrets }}
+        {{- with .Values.global.imagePullSecrets }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
         {{- with .Values.elasticsearch.image.pullSecrets }}
@@ -55,7 +55,7 @@ spec:
       {{- end }}
       initContainers:
       - name: fix-data-volume-permissions
-        image: {{ .Values.elasticsearch.initContainers.alpine.repository }}:{{ .Values.elasticsearch.initContainers.alpine.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.elasticsearch.initContainers.alpine.repository) }}:{{ .Values.elasticsearch.initContainers.alpine.tag }}
         {{- with .Values.elasticsearch.initContainers.alpine.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -67,7 +67,7 @@ spec:
           - mountPath: /usr/share/elasticsearch/data
             name: es-data
       - name: sysctl
-        image: {{ .Values.elasticsearch.initContainers.busybox.repository }}:{{ .Values.elasticsearch.initContainers.busybox.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.elasticsearch.initContainers.busybox.repository) }}:{{ .Values.elasticsearch.initContainers.busybox.tag }}
         {{- with .Values.elasticsearch.initContainers.busybox.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -77,7 +77,7 @@ spec:
         command: ["sysctl", "-w", "vm.max_map_count=262144"]
       containers:
       - name: elasticsearch
-        image: {{ .Values.elasticsearch.image.repository }}:{{ .Values.elasticsearch.image.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.elasticsearch.image.repository) }}:{{ .Values.elasticsearch.image.tag }}
         {{- with .Values.elasticsearch.image.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}

helm/templates/env.yaml

@@ -9,20 +9,39 @@ metadata:
 type: Opaque
 stringData:
   {{- range $key, $val := .Values.env }}
-  {{- if $val }}
+  {{- if and $val (ne $key "MYSQL_HOST") (ne $key "MYSQL_PORT") (ne $key "MYSQL_USER") (ne $key "MINIO_HOST") (ne $key "MINIO_PORT") (ne $key "REDIS_HOST") (ne $key "REDIS_PORT") }}
   {{ $key }}: {{ quote $val }}
   {{- end }}
   {{- end }}
   {{- /*
   Use host names derived from internal cluster DNS
   */}}
+  {{- if .Values.redis.enabled }}
   REDIS_HOST: {{ printf "%s-redis.%s.svc" (include "ragflow.fullname" .) .Release.Namespace }}
+  REDIS_PORT: "6379"
+  {{- else }}
+  REDIS_HOST: {{ required "env.REDIS_HOST is required when redis.enabled=false" .Values.env.REDIS_HOST | quote }}
+  REDIS_PORT: {{ default "6379" .Values.env.REDIS_PORT | quote }}
+  {{- end }}
+  {{- if .Values.mysql.enabled }}
   MYSQL_HOST: {{ printf "%s-mysql.%s.svc" (include "ragflow.fullname" .) .Release.Namespace }}
+  MYSQL_PORT: "3306"
+  {{- else }}
+  MYSQL_HOST: {{ required "env.MYSQL_HOST is required when mysql.enabled=false" .Values.env.MYSQL_HOST | quote }}
+  MYSQL_PORT: {{ default "3306" .Values.env.MYSQL_PORT | quote }}
+  MYSQL_USER: {{ default "root" .Values.env.MYSQL_USER | quote }}
+  {{- end }}
+  {{- if .Values.minio.enabled }}
   MINIO_HOST: {{ printf "%s-minio.%s.svc" (include "ragflow.fullname" .) .Release.Namespace }}
+  MINIO_PORT: "9000"
+  {{- else }}
+  MINIO_HOST: {{ default "" .Values.env.MINIO_HOST | quote }}
+  MINIO_PORT: {{ default "9000" .Values.env.MINIO_PORT | quote }}
+  {{- end }}
   {{- /*
   Fail if passwords are not provided in release values
   */}}
-  REDIS_PASSWORD: {{ .Values.env.REDIS_PASSWORD | required "REDIS_PASSWORD is required" }}
+  REDIS_PASSWORD: {{ default "" .Values.env.REDIS_PASSWORD }}
   {{- /*
   NOTE: MySQL uses MYSQL_ROOT_PASSWORD env var but Ragflow container expects
   MYSQL_PASSWORD so we need to define both as the same value here.
@@ -31,10 +50,9 @@ stringData:
   MYSQL_PASSWORD: {{ . }}
   MYSQL_ROOT_PASSWORD: {{ . }}
   {{- end }}
-  {{- with .Values.env.MINIO_PASSWORD | required "MINIO_PASSWORD is required" }}
-  MINIO_PASSWORD: {{ . }}
-  MINIO_ROOT_PASSWORD: {{ . }}
-  {{- end }}
+  {{- $minioPass := default "" .Values.env.MINIO_PASSWORD }}
+  MINIO_PASSWORD: {{ $minioPass }}
+  MINIO_ROOT_PASSWORD: {{ $minioPass }}
   {{- /*
   Only provide env vars for enabled doc engine
   */}}

helm/templates/infinity.yaml

@@ -32,7 +32,7 @@ spec:
       {{- include "ragflow.selectorLabels" . | nindent 6 }}
       app.kubernetes.io/component: infinity
   {{- with .Values.infinity.deployment.strategy }}
-  strategy:
+  updateStrategy:
     {{- . | toYaml | nindent 4 }}
   {{- end }}
   template:
@@ -43,9 +43,9 @@ spec:
       annotations:
         checksum/config: {{ include (print $.Template.BasePath "/env.yaml") . | sha256sum }}
     spec:
-      {{- if or .Values.imagePullSecrets .Values.infinity.image.pullSecrets }}
+      {{- if or .Values.global.imagePullSecrets .Values.infinity.image.pullSecrets }}
       imagePullSecrets:
-        {{- with .Values.imagePullSecrets }}
+        {{- with .Values.global.imagePullSecrets }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
         {{- with .Values.infinity.image.pullSecrets }}
@@ -54,7 +54,7 @@ spec:
       {{- end }}
       containers:
       - name: infinity
-        image: {{ .Values.infinity.image.repository }}:{{ .Values.infinity.image.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.infinity.image.repository) }}:{{ .Values.infinity.image.tag }}
         {{- with .Values.infinity.image.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}

helm/templates/ingress.yaml

@@ -35,7 +35,7 @@ spec:
             {{- end }}
             backend:
               service:
-                name: {{ $.Release.Name }}
+                name: {{ include "ragflow.fullname" $ }}
                 port:
                   name: http
           {{- end }}

helm/templates/minio.yaml

@@ -1,3 +1,4 @@
+{{- if .Values.minio.enabled }}
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
@@ -43,9 +44,9 @@ spec:
         {{- include "ragflow.labels" . | nindent 8 }}
         app.kubernetes.io/component: minio
     spec:
-      {{- if or .Values.imagePullSecrets .Values.minio.image.pullSecrets }}
+      {{- if or .Values.global.imagePullSecrets .Values.minio.image.pullSecrets }}
       imagePullSecrets:
-        {{- with .Values.imagePullSecrets }}
+        {{- with .Values.global.imagePullSecrets }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
         {{- with .Values.minio.image.pullSecrets }}
@@ -54,7 +55,7 @@ spec:
       {{- end }}
       containers:
       - name: minio
-        image: {{ .Values.minio.image.repository }}:{{ .Values.minio.image.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.minio.image.repository) }}:{{ .Values.minio.image.tag }}
         {{- with .Values.minio.image.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -103,3 +104,4 @@ spec:
       port: 9001
       targetPort: console
   type: {{ .Values.minio.service.type }}
+{{- end }}

helm/templates/mysql-config.yaml

@@ -1,3 +1,4 @@
+{{- if .Values.mysql.enabled }}
 ---
 apiVersion: v1
 kind: ConfigMap
@@ -7,3 +8,4 @@ data:
   init.sql: |-
     CREATE DATABASE IF NOT EXISTS rag_flow;
     USE rag_flow;
+{{- end }}

helm/templates/mysql.yaml

@@ -1,3 +1,4 @@
+{{- if .Values.mysql.enabled }}
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
@@ -32,7 +33,7 @@ spec:
       {{- include "ragflow.selectorLabels" . | nindent 6 }}
       app.kubernetes.io/component: mysql
   {{- with .Values.mysql.deployment.strategy }}
-  strategy:
+  updateStrategy:
     {{- . | toYaml | nindent 4 }}
   {{- end }}
   template:
@@ -44,9 +45,9 @@ spec:
         checksum/config-mysql: {{ include (print $.Template.BasePath "/mysql-config.yaml") . | sha256sum }}
         checksum/config-env: {{ include (print $.Template.BasePath "/env.yaml") . | sha256sum }}
     spec:
-      {{- if or .Values.imagePullSecrets .Values.mysql.image.pullSecrets }}
+      {{- if or .Values.global.imagePullSecrets .Values.mysql.image.pullSecrets }}
       imagePullSecrets:
-        {{- with .Values.imagePullSecrets }}
+        {{- with .Values.global.imagePullSecrets }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
         {{- with .Values.mysql.image.pullSecrets }}
@@ -55,7 +56,7 @@ spec:
       {{- end }}
       containers:
       - name: mysql
-        image: {{ .Values.mysql.image.repository }}:{{ .Values.mysql.image.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.mysql.image.repository) }}:{{ .Values.mysql.image.tag }}
         {{- with .Values.mysql.image.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -108,3 +109,4 @@ spec:
       port: 3306
       targetPort: mysql
   type: {{ .Values.mysql.service.type }}
+{{- end }}

helm/templates/opensearch.yaml

@@ -32,7 +32,7 @@ spec:
       {{- include "ragflow.selectorLabels" . | nindent 6 }}
       app.kubernetes.io/component: opensearch
   {{- with .Values.opensearch.deployment.strategy }}
-  strategy:
+  updateStrategy:
     {{- . | toYaml | nindent 4 }}
   {{- end }}
   template:
@@ -44,9 +44,9 @@ spec:
         checksum/config-opensearch: {{ include (print $.Template.BasePath "/opensearch-config.yaml") . | sha256sum }}
         checksum/config-env: {{ include (print $.Template.BasePath "/env.yaml") . | sha256sum }}
     spec:
-      {{- if or .Values.imagePullSecrets .Values.opensearch.image.pullSecrets }}
+      {{- if or .Values.global.imagePullSecrets .Values.opensearch.image.pullSecrets }}
       imagePullSecrets:
-        {{- with .Values.imagePullSecrets }}
+        {{- with .Values.global.imagePullSecrets }}
         {{- toYaml . | nindent 8 }}
         {{- end }}
         {{- with .Values.opensearch.image.pullSecrets }}
@@ -55,7 +55,7 @@ spec:
       {{- end }}
       initContainers:
       - name: fix-data-volume-permissions
-        image: {{ .Values.opensearch.initContainers.alpine.repository }}:{{ .Values.opensearch.initContainers.alpine.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.opensearch.initContainers.alpine.repository) }}:{{ .Values.opensearch.initContainers.alpine.tag }}
         {{- with .Values.opensearch.initContainers.alpine.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -67,7 +67,7 @@ spec:
           - mountPath: /usr/share/opensearch/data
             name: opensearch-data
       - name: sysctl
-        image: {{ .Values.opensearch.initContainers.busybox.repository }}:{{ .Values.opensearch.initContainers.busybox.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.opensearch.initContainers.busybox.repository) }}:{{ .Values.opensearch.initContainers.busybox.tag }}
         {{- with .Values.opensearch.initContainers.busybox.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}
@@ -77,7 +77,7 @@ spec:
         command: ["sysctl", "-w", "vm.max_map_count=262144"]
       containers:
       - name: opensearch
-        image: {{ .Values.opensearch.image.repository }}:{{ .Values.opensearch.image.tag }}
+        image: {{ include "ragflow.imageRepo" (dict "root" . "repo" .Values.opensearch.image.repository) }}:{{ .Values.opensearch.image.tag }}
         {{- with .Values.opensearch.image.pullPolicy }}
         imagePullPolicy: {{ . }}
         {{- end }}