Refa: knowledge_graph chunk method is deprecated (#7220)

### What problem does this PR solve?

The knowledge_graph chunk method is deprecated and should no longer be
used. #7184.

### Type of change

- [x] Refactoring

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -1,7 +1,7 @@
-name: Bug Report
+name: "🐞 Bug Report"
 description: Create a bug issue for RAGFlow
 title: "[Bug]: "
-labels: [bug]
+labels: ["🐞 bug"]
 body:
 - type: checkboxes
   attributes:

.github/ISSUE_TEMPLATE/feature_request.md

@@ -1,10 +0,0 @@
----
-name: Feature request
-title: '[Feature Request]: '
-about: Suggest an idea for RAGFlow
-labels: ''
----
-
-**Summary**
-
-Description for this feature.

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -1,7 +1,7 @@
-name: Feature request
+name: "💞 Feature request"
 description: Propose a feature request for RAGFlow.
 title: "[Feature Request]: "
-labels: [feature request]
+labels: ["💞 feature"]
 body:
   - type: checkboxes
     attributes:

.github/ISSUE_TEMPLATE/question.yml

@@ -1,7 +1,7 @@
-name: Question
+name: "🙋‍♀️ Question"
 description: Ask questions on RAGFlow
 title: "[Question]: "
-labels: [question]
+labels: ["🙋‍♀️ question"]
 body:
 - type: checkboxes
   attributes:

.github/workflows/tests.yml

@@ -51,7 +51,7 @@ jobs:
         uses: astral-sh/ruff-action@v2
         with:
           version: ">=0.8.2"
-          args: "check --ignore E402"
+          args: "check"
 
       - name: Build ragflow:nightly-slim
         run: |
@@ -145,7 +145,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          cd sdk/python && uv sync --python 3.10 --frozen && uv pip install . && source .venv/bin/activate && cd test/test_http_api && pytest -s --tb=short -m "not slow"
+          cd sdk/python && uv sync --python 3.10 --frozen && uv pip install . && source .venv/bin/activate && cd test/test_http_api && DOC_ENGINE=infinity pytest -s --tb=short -m "not slow"
 
       - name: Stop ragflow:nightly
         if: always()  # always run this step even if previous steps failed

Dockerfile

@@ -21,9 +21,7 @@ RUN --mount=type=bind,from=infiniflow/ragflow_deps:latest,source=/huggingface.co
     if [ "$LIGHTEN" != "1" ]; then \
         (tar -cf - \
             /huggingface.co/BAAI/bge-large-zh-v1.5 \
-            /huggingface.co/BAAI/bge-reranker-v2-m3 \
             /huggingface.co/maidalun1020/bce-embedding-base_v1 \
-            /huggingface.co/maidalun1020/bce-reranker-base_v1 \
             | tar -xf - --strip-components=2 -C /root/.ragflow) \
     fi
 
@@ -46,7 +44,8 @@ ENV DEBIAN_FRONTEND=noninteractive
 # 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 \
     if [ "$NEED_MIRROR" == "1" ]; then \
-        sed -i 's|http://archive.ubuntu.com|https://mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list; \
+        sed -i 's|http://ports.ubuntu.com|http://mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list; \
+        sed -i 's|http://archive.ubuntu.com|http://mirrors.tuna.tsinghua.edu.cn|g' /etc/apt/sources.list; \
     fi; \
     rm -f /etc/apt/apt.conf.d/docker-clean && \
     echo 'Binary::apt::APT::Keep-Downloaded-Packages "true";' > /etc/apt/apt.conf.d/keep-cache && \
@@ -199,9 +198,10 @@ COPY agent agent
 COPY graphrag graphrag
 COPY agentic_reasoning agentic_reasoning
 COPY pyproject.toml uv.lock ./
+COPY mcp mcp
 
 COPY docker/service_conf.yaml.template ./conf/service_conf.yaml.template
-COPY docker/entrypoint.sh docker/entrypoint-parser.sh ./
+COPY docker/entrypoint.sh ./
 RUN chmod +x ./entrypoint*.sh
 
 # Copy compiled web pages

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Document</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -78,12 +78,10 @@ Try our demo at [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Latest Updates
 
+- 2025-03-19 Supports using a multi-modal model to make sense of images within PDF or DOCX files.
 - 2025-02-28 Combined with Internet search (Tavily), supports reasoning like Deep Research for any LLMs.
-- 2025-02-05 Updates the model list of 'SILICONFLOW' and adds support for Deepseek-R1/DeepSeek-V3.
 - 2025-01-26 Optimizes knowledge graph extraction and application, offering various configuration options.
 - 2024-12-18 Upgrades Document Layout Analysis model in DeepDoc.
-- 2024-12-04 Adds support for pagerank score in knowledge base.
-- 2024-11-22 Adds more variables to Agent.
 - 2024-11-01 Adds keyword extraction and related question generation to the parsed chunks to improve the accuracy of retrieval.
 - 2024-08-22 Support text to SQL statements through RAG.
 
@@ -178,17 +176,21 @@ 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 `v0.17.2-slim` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.17.2-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2` for the full edition `v0.17.2`.
+   > The command below downloads the `v0.18.0-slim` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.18.0-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0` for the full edition `v0.18.0`.
 
    ```bash
    $ cd ragflow/docker
+   # Use CPU for embedding and DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
+
+   # To use GPU to accelerate embedding and DeepDoc tasks:
+   # docker compose -f docker-compose-gpu.yml up -d
    ```
 
    | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
    |-------------------|-----------------|-----------------------|--------------------------|
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                   | Stable release            |
+   | v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.18.0-slim      | &approx;2       | ❌                   | Stable release            |
    | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
    | nightly-slim      | &approx;2       | ❌                   | _Unstable_ nightly build  |
 
@@ -276,7 +278,7 @@ This image is approximately 2 GB in size and relies on external LLM and embeddin
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 Build a Docker image including embedding models
@@ -286,7 +288,7 @@ This image is approximately 9 GB in size. As it includes embedding models, it re
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 Launch service from source for development
@@ -349,9 +351,12 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 ## 📚 Documentation
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [User guide](https://ragflow.io/docs/dev/category/guides)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
 - [References](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 Roadmap
 
@@ -359,7 +364,7 @@ See the [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214
 
 ## 🏄 Community
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/NjYzJD3GM3)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Dokumentasi</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Peta Jalan</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -75,12 +75,10 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Pembaruan Terbaru
 
+- 2025-03-19 Mendukung penggunaan model multi-modal untuk memahami gambar di dalam file PDF atau DOCX.
 - 2025-02-28 dikombinasikan dengan pencarian Internet (TAVILY), mendukung penelitian mendalam untuk LLM apa pun.
-- 2025-02-05 Memperbarui daftar model 'SILICONFLOW' dan menambahkan dukungan untuk Deepseek-R1/DeepSeek-V3.
 - 2025-01-26 Optimalkan ekstraksi dan penerapan grafik pengetahuan dan sediakan berbagai opsi konfigurasi.
 - 2024-12-18 Meningkatkan model Analisis Tata Letak Dokumen di DeepDoc.
-- 2024-12-04 Mendukung skor pagerank ke basis pengetahuan.
-- 2024-11-22 Peningkatan definisi dan penggunaan variabel di Agen.
 - 2024-11-01 Penambahan ekstraksi kata kunci dan pembuatan pertanyaan terkait untuk meningkatkan akurasi pengambilan.
 - 2024-08-22 Dukungan untuk teks ke pernyataan SQL melalui RAG.
 
@@ -171,19 +169,23 @@ 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 v0.17.2-slim dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.17.2-slim, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server. Misalnya, atur RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2 untuk edisi lengkap v0.17.2.
+> Perintah di bawah ini mengunduh edisi v0.18.0-slim dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.18.0-slim, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server. Misalnya, atur RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0 untuk edisi lengkap v0.18.0.
 
-   ```bash
-   $ cd ragflow/docker
-   $ docker compose -f docker-compose.yml up -d
-   ```
+```bash
+$ cd ragflow/docker
+# Use CPU for embedding and DeepDoc tasks:
+$ docker compose -f docker-compose.yml up -d
 
-   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
-   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                    | Stable release           |
-   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
-   | nightly-slim      | &approx;2       | ❌                    | _Unstable_ nightly build |
+# To use GPU to accelerate embedding and DeepDoc tasks:
+# docker compose -f docker-compose-gpu.yml up -d
+```
+
+| RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
+| ----------------- | --------------- | --------------------- | ------------------------ |
+| v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+| v0.18.0-slim      | &approx;2       | ❌                    | Stable release           |
+| nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
+| nightly-slim      | &approx;2       | ❌                    | _Unstable_ nightly build |
 
 1. Periksa status server setelah server aktif dan berjalan:
 
@@ -242,7 +244,7 @@ Image ini berukuran sekitar 2 GB dan bergantung pada aplikasi LLM eksternal dan
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 Membangun Docker Image Termasuk Model Embedding
@@ -252,7 +254,7 @@ Image ini berukuran sekitar 9 GB. Karena sudah termasuk model embedding, ia hany
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 Menjalankan Aplikasi dari untuk Pengembangan
@@ -315,9 +317,12 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 ## 📚 Dokumentasi
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [Panduan Pengguna](https://ragflow.io/docs/dev/category/guides)
-- [Referensi](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
+- [References](https://ragflow.io/docs/dev/category/references)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 Roadmap
 
@@ -325,7 +330,7 @@ Lihat [Roadmap RAGFlow 2025](https://github.com/infiniflow/ragflow/issues/4214)
 
 ## 🏄 Komunitas
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/NjYzJD3GM3)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Document</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -55,12 +55,10 @@
 
 ## 🔥 最新情報
 
+- 2025-03-19 PDFまたはDOCXファイル内の画像を理解するために、多モーダルモデルを使用することをサポートします。
 - 2025-02-28 インターネット検索 (TAVILY) と組み合わせて、あらゆる LLM の詳細な調査をサポートします。
-- 2025-02-05 シリコン フローの St およびモデル リストを更新し、Deep Seek-R1/Deep Seek-V3 のサポートを追加しました。
 - 2025-01-26 ナレッジ グラフの抽出と適用を最適化し、さまざまな構成オプションを提供します。
 - 2024-12-18 DeepDoc のドキュメント レイアウト分析モデルをアップグレードします。
-- 2024-12-04 ナレッジ ベースへのページランク スコアをサポートしました。
-- 2024-11-22 エージェントでの変数の定義と使用法を改善しました。
 - 2024-11-01 再現の精度を向上させるために、解析されたチャンクにキーワード抽出と関連質問の生成を追加しました。
 - 2024-08-22 RAG を介して SQL ステートメントへのテキストをサポートします。
 
@@ -151,17 +149,21 @@
 > 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 v0.17.2-slim 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。v0.17.2-slim 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。例えば、完全版 v0.17.2 をダウンロードするには、RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2 と設定します。
+   > 以下のコマンドは、RAGFlow Docker イメージの v0.18.0-slim 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。v0.18.0-slim 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。例えば、完全版 v0.18.0 をダウンロードするには、RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0 と設定します。
 
    ```bash
    $ cd ragflow/docker
+   # Use CPU for embedding and DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
+
+   # To use GPU to accelerate embedding and DeepDoc tasks:
+   # docker compose -f docker-compose-gpu.yml up -d
    ```
 
    | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
    | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.18.0-slim      | &approx;2       | ❌                    | Stable release           |
    | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
    | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |
 
@@ -238,7 +240,7 @@ RAGFlow はデフォルトで Elasticsearch を使用して全文とベクトル
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 ソースコードをコンパイルした Docker イメージ（埋め込みモデルを含む）
@@ -248,7 +250,7 @@ docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-s
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 -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
@@ -311,9 +313,12 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 ## 📚 ドキュメンテーション
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [User guide](https://ragflow.io/docs/dev/category/guides)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
 - [References](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 ロードマップ
 
@@ -321,7 +326,7 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 
 ## 🏄 コミュニティ
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/NjYzJD3GM3)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Document</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -55,13 +55,10 @@
 
 ## 🔥 업데이트
 
+- 2025-03-19 PDF 또는 DOCX 파일 내의 이미지를 이해하기 위해 다중 모드 모델을 사용하는 것을 지원합니다.
 - 2025-02-28 인터넷 검색(TAVILY)과 결합되어 모든 LLM에 대한 심층 연구를 지원합니다.
-- 2025-02-05 'SILICONFLOW' 모델 목록을 업데이트하고 Deepseek-R1/DeepSeek-V3에 대한 지원을 추가합니다.
 - 2025-01-26 지식 그래프 추출 및 적용을 최적화하고 다양한 구성 옵션을 제공합니다.
 - 2024-12-18 DeepDoc의 문서 레이아웃 분석 모델 업그레이드.
-- 2024-12-04 지식베이스에 대한 페이지랭크 점수를 지원합니다.
-
-- 2024-11-22 에이전트의 변수 정의 및 사용을 개선했습니다.
 - 2024-11-01 파싱된 청크에 키워드 추출 및 관련 질문 생성을 추가하여 재현율을 향상시킵니다.
 - 2024-08-22 RAG를 통해 SQL 문에 텍스트를 지원합니다.
 
@@ -152,17 +149,21 @@
 > 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 v0.17.2-slim 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.17.2-slim과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오. 예를 들어, 전체 버전인 v0.17.2을 다운로드하려면 RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2로 설정합니다.
+   > 아래 명령어는 RAGFlow Docker 이미지의 v0.18.0-slim 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.18.0-slim과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오. 예를 들어, 전체 버전인 v0.18.0을 다운로드하려면 RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0로 설정합니다.
 
    ```bash
    $ cd ragflow/docker
+   # Use CPU for embedding and DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
+
+   # To use GPU to accelerate embedding and DeepDoc tasks:
+   # docker compose -f docker-compose-gpu.yml up -d
    ```
 
    | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
    | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.18.0-slim      | &approx;2       | ❌                    | Stable release           |
    | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
    | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |
 
@@ -238,7 +239,7 @@ RAGFlow 는 기본적으로 Elasticsearch 를 사용하여 전체 텍스트 및
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 소스 코드로 Docker 이미지를 컴파일합니다(임베딩 모델 포함)
@@ -248,7 +249,7 @@ docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-s
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 소스 코드로 서비스를 시작합니다.
@@ -311,9 +312,12 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 ## 📚 문서
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [User guide](https://ragflow.io/docs/dev/category/guides)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
 - [References](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 로드맵
 
@@ -321,7 +325,7 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 
 ## 🏄 커뮤니티
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/NjYzJD3GM3)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Documentação</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -75,12 +75,10 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
 
 ## 🔥 Últimas Atualizações
 
-- 28/02/2025 combinado com a pesquisa na Internet (T AVI LY), suporta pesquisas profundas para qualquer LLM.
-- 05-02-2025 Atualiza a lista de modelos de 'SILICONFLOW' e adiciona suporte para Deepseek-R1/DeepSeek-V3.
+- 19-03-2025 Suporta o uso de um modelo multi-modal para entender imagens dentro de arquivos PDF ou DOCX.
+- 28-02-2025 combinado com a pesquisa na Internet (T AVI LY), suporta pesquisas profundas para qualquer LLM.
 - 26-01-2025 Otimize a extração e aplicação de gráficos de conhecimento e forneça uma variedade de opções de configuração.
 - 18-12-2024 Atualiza o modelo de Análise de Layout de Documentos no DeepDoc.
-- 04-12-2024 Adiciona suporte para pontuação de pagerank na base de conhecimento.
-- 22-11-2024 Adiciona mais variáveis para o Agente.
 - 01-11-2024 Adiciona extração de palavras-chave e geração de perguntas relacionadas aos blocos analisados para melhorar a precisão da recuperação.
 - 22-08-2024 Suporta conversão de texto para comandos SQL via RAG.
 
@@ -171,17 +169,21 @@ 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 `v0.17.2-slim` 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 `v0.17.2-slim`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor. Por exemplo: defina `RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2` para a edição completa `v0.17.2`.
+    > O comando abaixo baixa a edição `v0.18.0-slim` 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 `v0.18.0-slim`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor. Por exemplo: defina `RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0` para a edição completa `v0.18.0`.
 
     ```bash
     $ cd ragflow/docker
+    # Use CPU for embedding and DeepDoc tasks:
     $ docker compose -f docker-compose.yml up -d
+
+    # To use GPU to accelerate embedding and DeepDoc tasks:
+    # docker compose -f docker-compose-gpu.yml up -d
     ```
 
     | Tag da imagem RAGFlow | Tamanho da imagem (GB) | Possui modelos de incorporação? | Estável?                 |
     | --------------------- | ---------------------- | ------------------------------- | ------------------------ |
-    | v0.17.2               | ~9                     | :heavy_check_mark:              | Lançamento estável       |
-    | v0.17.2-slim          | ~2                     | ❌                              | Lançamento estável       |
+    | v0.18.0               | ~9                     | :heavy_check_mark:              | Lançamento estável       |
+    | v0.18.0-slim          | ~2                     | ❌                              | Lançamento estável       |
     | nightly               | ~9                     | :heavy_check_mark:              | _Instável_ build noturno |
     | nightly-slim          | ~2                     | ❌                               | _Instável_ build noturno |
 
@@ -261,7 +263,7 @@ Esta imagem tem cerca de 2 GB de tamanho e depende de serviços externos de LLM
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 Criar uma imagem Docker incluindo modelos de incorporação
@@ -271,7 +273,7 @@ Esta imagem tem cerca de 9 GB de tamanho. Como inclui modelos de incorporação,
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 Lançar o serviço a partir do código-fonte para desenvolvimento
@@ -335,10 +337,13 @@ docker build -f Dockerfile -t infiniflow/ragflow:nightly .
 
 ## 📚 Documentação
 
-- [Início rápido](https://ragflow.io/docs/dev/)
-- [Guia do usuário](https://ragflow.io/docs/dev/category/guides)
-- [Referências](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [Quickstart](https://ragflow.io/docs/dev/)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
+- [References](https://ragflow.io/docs/dev/category/references)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 Roadmap
 
@@ -346,7 +351,7 @@ Veja o [RAGFlow Roadmap 2025](https://github.com/infiniflow/ragflow/issues/4214)
 
 ## 🏄 Comunidade
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/NjYzJD3GM3)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

README_tzh.md

@@ -21,7 +21,7 @@
         <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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -35,7 +35,7 @@
   <a href="https://ragflow.io/docs/dev/">Document</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -54,12 +54,10 @@
 
 ## 🔥 近期更新
 
+- 2025-03-19 PDF和DOCX中的圖支持用多模態大模型去解析得到描述.
 - 2025-02-28 結合網路搜尋（Tavily），對於任意大模型實現類似 Deep Research 的推理功能.
-- 2025-02-05 更新「SILICONFLOW」的型號清單並新增 Deepseek-R1/DeepSeek-V3 的支援。
 - 2025-01-26 最佳化知識圖譜的擷取與應用，提供了多種配置選擇。
 - 2024-12-18 升級了 DeepDoc 的文檔佈局分析模型。
-- 2024-12-04 支援知識庫的 Pagerank 分數。
-- 2024-11-22 完善了 Agent 中的變數定義和使用。
 - 2024-11-01 對解析後的 chunk 加入關鍵字抽取和相關問題產生以提高回想的準確度。
 - 2024-08-22 支援用 RAG 技術實現從自然語言到 SQL 語句的轉換。
 
@@ -150,17 +148,21 @@
 > 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 slim Docker 映像 `v0.17.2-slim`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.17.2-slim` 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。例如，你可以透過設定 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2` 來下載 RAGFlow 鏡像的 `v0.17.2` 完整發行版。
+   > 執行以下指令會自動下載 RAGFlow slim Docker 映像 `v0.18.0-slim`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.18.0-slim` 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。例如，你可以透過設定 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0` 來下載 RAGFlow 鏡像的 `v0.18.0` 完整發行版。
 
    ```bash
    $ cd ragflow/docker
+   # Use CPU for embedding and DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
+
+   # To use GPU to accelerate embedding and DeepDoc tasks:
+   # docker compose -f docker-compose-gpu.yml up -d
    ```
 
    | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
    | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.18.0-slim      | &approx;2       | ❌                    | Stable release           |
    | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
    | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |
 
@@ -249,7 +251,7 @@ RAGFlow 預設使用 Elasticsearch 儲存文字和向量資料. 如果要切換
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 原始碼編譯 Docker 映像（包含 embedding 模型）
@@ -259,7 +261,7 @@ docker build --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t in
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 以原始碼啟動服務
@@ -325,9 +327,12 @@ npm install
 ## 📚 技術文檔
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [User guide](https://ragflow.io/docs/dev/category/guides)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
 - [References](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 路線圖
 
@@ -335,7 +340,7 @@ npm install
 
 ## 🏄 開源社群
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/zd4qPW6t)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

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/badge/docker_pull-ragflow:v0.17.2-brightgreen" alt="docker pull infiniflow/ragflow:v0.17.2">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.18.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.18.0">
     </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">
@@ -36,7 +36,7 @@
   <a href="https://ragflow.io/docs/dev/">Document</a> |
   <a href="https://github.com/infiniflow/ragflow/issues/4214">Roadmap</a> |
   <a href="https://twitter.com/infiniflowai">Twitter</a> |
-  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://discord.gg/NjYzJD3GM3">Discord</a> |
   <a href="https://demo.ragflow.io">Demo</a>
 </h4>
 
@@ -55,12 +55,10 @@
 
 ## 🔥 近期更新
 
+- 2025-03-19 PDF和DOCX中的图支持用多模态大模型去解析得到描述.
 - 2025-02-28 结合互联网搜索（Tavily），对于任意大模型实现类似 Deep Research 的推理功能.
-- 2025-02-05 更新硅基流动的模型列表，增加了对 Deepseek-R1/DeepSeek-V3 的支持。
 - 2025-01-26 优化知识图谱的提取和应用，提供了多种配置选择。
 - 2024-12-18 升级了 DeepDoc 的文档布局分析模型。
-- 2024-12-04 支持知识库的 Pagerank 分数。
-- 2024-11-22 完善了 Agent 中的变量定义和使用。
 - 2024-11-01 对解析后的 chunk 加入关键词抽取和相关问题生成以提高召回的准确度。
 - 2024-08-22 支持用 RAG 技术实现从自然语言到 SQL 语句的转换。
 
@@ -151,17 +149,21 @@
 > 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 slim Docker 镜像 `v0.17.2-slim`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.17.2-slim` 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。比如，你可以通过设置 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2` 来下载 RAGFlow 镜像的 `v0.17.2` 完整发行版。
+   > 运行以下命令会自动下载 RAGFlow slim Docker 镜像 `v0.18.0-slim`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.18.0-slim` 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。比如，你可以通过设置 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0` 来下载 RAGFlow 镜像的 `v0.18.0` 完整发行版。
 
    ```bash
    $ cd ragflow/docker
+   # Use CPU for embedding and DeepDoc tasks:
    $ docker compose -f docker-compose.yml up -d
+
+   # To use GPU to accelerate embedding and DeepDoc tasks:
+   # docker compose -f docker-compose-gpu.yml up -d
    ```
 
    | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
    | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.17.2           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.17.2-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.18.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.18.0-slim      | &approx;2       | ❌                    | Stable release           |
    | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
    | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |
 
@@ -250,7 +252,7 @@ RAGFlow 默认使用 Elasticsearch 存储文本和向量数据. 如果要切换
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
+docker build --platform linux/amd64 --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly-slim .
 ```
 
 ## 🔧 源码编译 Docker 镜像（包含 embedding 模型）
@@ -260,7 +262,7 @@ docker build --build-arg LIGHTEN=1 --build-arg NEED_MIRROR=1 -f Dockerfile -t in
 ```bash
 git clone https://github.com/infiniflow/ragflow.git
 cd ragflow/
-docker build --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly .
+docker build --platform linux/amd64 --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:nightly .
 ```
 
 ## 🔨 以源代码启动服务
@@ -286,12 +288,11 @@ docker build --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:night
    docker compose -f docker/docker-compose-base.yml up -d
    ```
 
-   在 `/etc/hosts` 中添加以下代码，将 **conf/service_conf.yaml** 文件中的所有 host 地址都解析为 `127.0.0.1`：
+   在 `/etc/hosts` 中添加以下代码，目的是将 **conf/service_conf.yaml** 文件中的所有 host 地址都解析为 `127.0.0.1`：
 
    ```
    127.0.0.1       es01 infinity mysql minio redis
    ```
-
 4. 如果无法访问 HuggingFace，可以把环境变量 `HF_ENDPOINT` 设成相应的镜像站点：
 
    ```bash
@@ -320,13 +321,21 @@ docker build --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:night
    _以下界面说明系统已经成功启动：_
 
    ![](https://github.com/user-attachments/assets/0daf462c-a24d-4496-a66f-92533534e187)
+8. 开发完成后停止 RAGFlow 服务
+   停止 RAGFlow 前端和后端服务：
+   ```bash
+   pkill -f "ragflow_server.py|task_executor.py"
+   ```
 
 ## 📚 技术文档
 
 - [Quickstart](https://ragflow.io/docs/dev/)
-- [User guide](https://ragflow.io/docs/dev/category/guides)
+- [Configuration](https://ragflow.io/docs/dev/configurations)
+- [Release notes](https://ragflow.io/docs/dev/release_notes)
+- [User guides](https://ragflow.io/docs/dev/category/guides)
+- [Developer guides](https://ragflow.io/docs/dev/category/developers)
 - [References](https://ragflow.io/docs/dev/category/references)
-- [FAQ](https://ragflow.io/docs/dev/faq)
+- [FAQs](https://ragflow.io/docs/dev/faq)
 
 ## 📜 路线图
 
@@ -334,7 +343,7 @@ docker build --build-arg NEED_MIRROR=1 -f Dockerfile -t infiniflow/ragflow:night
 
 ## 🏄 开源社区
 
-- [Discord](https://discord.gg/4XxujFgUN7)
+- [Discord](https://discord.gg/zd4qPW6t)
 - [Twitter](https://twitter.com/infiniflowai)
 - [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 

agent/canvas.py

@@ -235,7 +235,7 @@ class Canvas:
             pid = self.components[cid]["parent_id"]
             o, _ = self.components[cid]["obj"].output(allow_partial=False)
             oo, _ = self.components[pid]["obj"].output(allow_partial=False)
-            self.components[pid]["obj"].set(pd.concat([oo, o], ignore_index=True))
+            self.components[pid]["obj"].set_output(pd.concat([oo, o], ignore_index=True).dropna())
             downstream = [pid]
 
         for m in prepare2run(downstream):
@@ -252,20 +252,20 @@ class Canvas:
             if loop:
                 raise OverflowError(f"Too much loops: {loop}")
 
+            downstream = []
             if cpn["obj"].component_name.lower() in ["switch", "categorize", "relevant"]:
                 switch_out = cpn["obj"].output()[1].iloc[0, 0]
                 assert switch_out in self.components, \
                     "{}'s output: {} not valid.".format(cpn_id, switch_out)
-                for m in prepare2run([switch_out]):
-                    yield {"content": m, "running_status": True}
-                continue
+                downstream = [switch_out]
+            else:
+                downstream = cpn["downstream"]
 
-            downstream = cpn["downstream"]
             if not downstream and cpn.get("parent_id"):
                 pid = cpn["parent_id"]
                 _, o = cpn["obj"].output(allow_partial=False)
                 _, oo = self.components[pid]["obj"].output(allow_partial=False)
-                self.components[pid]["obj"].set_output(pd.concat([oo.dropna(axis=1), o.dropna(axis=1)], ignore_index=True))
+                self.components[pid]["obj"].set_output(pd.concat([oo.dropna(axis=1), o.dropna(axis=1)], ignore_index=True).dropna())
                 downstream = [pid]
 
             for m in prepare2run(downstream):

agent/component/base.py

@@ -384,6 +384,11 @@ class ComponentBase(ABC):
             "params": {}
         }
         """
+        out = getattr(self._param, self._param.output_var_name)
+        if isinstance(out, pd.DataFrame) and "chunks" in out:
+            del out["chunks"]
+            setattr(self._param, self._param.output_var_name, out)
+
         return """{{
             "component_name": "{}",
             "params": {},
@@ -396,6 +401,8 @@ class ComponentBase(ABC):
         )
 
     def __init__(self, canvas, id, param: ComponentParamBase):
+        from agent.canvas import Canvas  # Local import to avoid cyclic dependency
+        assert isinstance(canvas, Canvas), "canvas must be an instance of Canvas"
         self._canvas = canvas
         self._id = id
         self._param = param
@@ -429,7 +436,7 @@ class ComponentBase(ABC):
         if not isinstance(o, partial):
             if not isinstance(o, pd.DataFrame):
                 if isinstance(o, list):
-                    return self._param.output_var_name, pd.DataFrame(o)
+                    return self._param.output_var_name, pd.DataFrame(o).dropna()
                 if o is None:
                     return self._param.output_var_name, pd.DataFrame()
                 return self._param.output_var_name, pd.DataFrame([{"content": str(o)}])
@@ -437,15 +444,15 @@ class ComponentBase(ABC):
 
         if allow_partial or not isinstance(o, partial):
             if not isinstance(o, partial) and not isinstance(o, pd.DataFrame):
-                return pd.DataFrame(o if isinstance(o, list) else [o])
+                return pd.DataFrame(o if isinstance(o, list) else [o]).dropna()
             return self._param.output_var_name, o
 
         outs = None
         for oo in o():
             if not isinstance(oo, pd.DataFrame):
-                outs = pd.DataFrame(oo if isinstance(oo, list) else [oo])
+                outs = pd.DataFrame(oo if isinstance(oo, list) else [oo]).dropna()
             else:
-                outs = oo
+                outs = oo.dropna()
         return self._param.output_var_name, outs
 
     def reset(self):
@@ -463,6 +470,8 @@ class ComponentBase(ABC):
         if len(self._canvas.path) > 1:
             reversed_cpnts.extend(self._canvas.path[-2])
         reversed_cpnts.extend(self._canvas.path[-1])
+        up_cpns = self.get_upstream()
+        reversed_up_cpnts = [cpn for cpn in reversed_cpnts if cpn in up_cpns]
 
         if self._param.query:
             self._param.inputs = []
@@ -484,7 +493,7 @@ class ComponentBase(ABC):
                     if q["component_id"].lower().find("answer") == 0:
                         txt = []
                         for r, c in self._canvas.history[::-1][:self._param.message_history_window_size][::-1]:
-                            txt.append(f"{r.upper()}: {c}")
+                            txt.append(f"{r.upper()}:{c}")
                         txt = "\n".join(txt)
                         self._param.inputs.append({"content": txt, "component_id": q["component_id"]})
                         outs.append(pd.DataFrame([{"content": txt}]))
@@ -505,7 +514,7 @@ class ComponentBase(ABC):
 
         upstream_outs = []
 
-        for u in reversed_cpnts[::-1]:
+        for u in reversed_up_cpnts[::-1]:
             if self.get_component_name(u) in ["switch", "concentrator"]:
                 continue
             if self.component_name.lower() == "generate" and self.get_component_name(u) == "retrieval":
@@ -545,7 +554,7 @@ class ComponentBase(ABC):
         return df
 
     def get_input_elements(self):
-        assert self._param.query, "Please identify input parameters firstly."
+        assert self._param.query, "Please verify the input parameters first."
         eles = []
         for q in self._param.query:
             if q.get("component_id"):
@@ -565,8 +574,10 @@ class ComponentBase(ABC):
         if len(self._canvas.path) > 1:
             reversed_cpnts.extend(self._canvas.path[-2])
         reversed_cpnts.extend(self._canvas.path[-1])
+        up_cpns = self.get_upstream()
+        reversed_up_cpnts = [cpn for cpn in reversed_cpnts if cpn in up_cpns]
 
-        for u in reversed_cpnts[::-1]:
+        for u in reversed_up_cpnts[::-1]:
             if self.get_component_name(u) in ["switch", "answer"]:
                 continue
             return self._canvas.get_component(u)["obj"].output()[1]
@@ -584,3 +595,7 @@ class ComponentBase(ABC):
     def get_parent(self):
         pid = self._canvas.get_component(self._id)["parent_id"]
         return self._canvas.get_component(pid)["obj"]
+
+    def get_upstream(self):
+        cpn_nms = self._canvas.get_component(self._id)['upstream']
+        return cpn_nms

agent/component/categorize.py

@@ -50,26 +50,29 @@ class CategorizeParam(GenerateParam):
         for c, desc in self.category_description.items():
             if desc.get("description"):
                 descriptions.append(
-                    "--------------------\nCategory: {}\nDescription: {}\n".format(c, desc["description"]))
+                    "\nCategory: {}\nDescription: {}".format(c, desc["description"]))
 
         self.prompt = """
-        You're a text classifier. You need to categorize the user’s questions into {} categories, 
-        namely: {}
-        Here's description of each category:
-        {}
+Role: You're a text classifier. 
+Task: You need to categorize the user’s questions into {} categories, namely: {}
 
-        You could learn from the following examples:
-        {}
-        You could learn from the above examples.
-        Just mention the category names, no need for any additional words.
+Here's description of each category:
+{}
 
-        ---- Real Data ----
-        {}
+You could learn from the following examples:
+{}
+You could learn from the above examples.
+
+Requirements:
+- Just mention the category names, no need for any additional words.
+
+---- Real Data ----
+USER: {}\n
         """.format(
             len(self.category_description.keys()),
             "/".join(list(self.category_description.keys())),
             "\n".join(descriptions),
-            "- ".join(cate_lines),
+            "\n\n- ".join(cate_lines),
             chat_hist
         )
         return self.prompt
@@ -85,9 +88,16 @@ class Categorize(Generate, ABC):
         ans = chat_mdl.chat(self._param.get_prompt(input), [{"role": "user", "content": "\nCategory: "}],
                             self._param.gen_conf())
         logging.debug(f"input: {input}, answer: {str(ans)}")    
+        # Count the number of times each category appears in the answer.
+        category_counts = {}
         for c in self._param.category_description.keys():
-            if ans.lower().find(c.lower()) >= 0:
-                return Categorize.be_output(self._param.category_description[c]["to"])
+            count = ans.lower().count(c.lower())
+            category_counts[c] = count
+            
+        # If a category is found, return the category with the highest count.
+        if any(category_counts.values()):
+            max_category = max(category_counts.items(), key=lambda x: x[1])
+            return Categorize.be_output(self._param.category_description[max_category[0]]["to"])
 
         return Categorize.be_output(list(self._param.category_description.items())[-1][1]["to"])
 

agent/component/email.py

@@ -82,7 +82,10 @@ class Email(ComponentBase, ABC):
             logging.info(f"Connecting to SMTP server {self._param.smtp_server}:{self._param.smtp_port}")
             
             context = smtplib.ssl.create_default_context()
-            with smtplib.SMTP_SSL(self._param.smtp_server, self._param.smtp_port, context=context) as server:
+            with smtplib.SMTP(self._param.smtp_server, self._param.smtp_port) as server:
+                server.ehlo()
+                server.starttls(context=context)
+                server.ehlo()
                 # Login
                 logging.info(f"Attempting to login with email: {self._param.email}")
                 server.login(self._param.email, self._param.password)

agent/component/generate.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import json
 import re
 from functools import partial
 import pandas as pd
@@ -74,29 +75,30 @@ class Generate(ComponentBase):
         return list(cpnts)
 
     def set_cite(self, retrieval_res, answer):
-        retrieval_res = retrieval_res.dropna(subset=["vector", "content_ltks"]).reset_index(drop=True)
         if "empty_response" in retrieval_res.columns:
             retrieval_res["empty_response"].fillna("", inplace=True)
+        chunks = json.loads(retrieval_res["chunks"][0])
         answer, idx = settings.retrievaler.insert_citations(answer,
-                                                            [ck["content_ltks"] for _, ck in retrieval_res.iterrows()],
-                                                            [ck["vector"] for _, ck in retrieval_res.iterrows()],
+                                                            [ck["content_ltks"] for ck in chunks],
+                                                            [ck["vector"] for ck in chunks],
                                                             LLMBundle(self._canvas.get_tenant_id(), LLMType.EMBEDDING,
                                                                       self._canvas.get_embedding_model()), tkweight=0.7,
                                                             vtweight=0.3)
         doc_ids = set([])
         recall_docs = []
         for i in idx:
-            did = retrieval_res.loc[int(i), "doc_id"]
+            did = chunks[int(i)]["doc_id"]
             if did in doc_ids:
                 continue
             doc_ids.add(did)
-            recall_docs.append({"doc_id": did, "doc_name": retrieval_res.loc[int(i), "docnm_kwd"]})
+            recall_docs.append({"doc_id": did, "doc_name": chunks[int(i)]["docnm_kwd"]})
 
-        del retrieval_res["vector"]
-        del retrieval_res["content_ltks"]
+        for c in chunks:
+            del c["vector"]
+            del c["content_ltks"]
 
         reference = {
-            "chunks": [ck.to_dict() for _, ck in retrieval_res.iterrows()],
+            "chunks": chunks,
             "doc_aggs": recall_docs
         }
 
@@ -200,7 +202,7 @@ class Generate(ComponentBase):
         ans = chat_mdl.chat(msg[0]["content"], msg[1:], self._param.gen_conf())
         ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
 
-        if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
+        if self._param.cite and "chunks" in retrieval_res.columns:
             res = self.set_cite(retrieval_res, ans)
             return pd.DataFrame([res])
 
@@ -229,7 +231,7 @@ class Generate(ComponentBase):
             answer = ans
             yield res
 
-        if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
+        if self._param.cite and "chunks" in retrieval_res.columns:
             res = self.set_cite(retrieval_res, answer)
             yield res
 

agent/component/keyword.py

@@ -57,6 +57,7 @@ class KeywordExtract(Generate, ABC):
         ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": query}],
                             self._param.gen_conf())
 
+        ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
         ans = re.sub(r".*keyword:", "", ans).strip()
         logging.debug(f"ans: {ans}")
         return KeywordExtract.be_output(ans)

agent/component/retrieval.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import json
 import logging
 from abc import ABC
 
@@ -24,6 +25,7 @@ from api.db.services.llm_service import LLMBundle
 from api import settings
 from agent.component.base import ComponentBase, ComponentParamBase
 from rag.app.tag import label_question
+from rag.prompts import kb_prompt
 from rag.utils.tavily_conn import Tavily
 
 
@@ -56,9 +58,6 @@ class Retrieval(ComponentBase, ABC):
     def _run(self, history, **kwargs):
         query = self.get_input()
         query = str(query["content"][0]) if "content" in query else ""
-        lines = query.split('\n')
-        user_queries = [line.split("USER:", 1)[1] for line in lines if line.startswith("USER:")]
-        query = user_queries[-1] if user_queries else ""
         kbs = KnowledgebaseService.get_by_ids(self._param.kb_ids)
         if not kbs:
             return Retrieval.be_output("")
@@ -66,19 +65,25 @@ class Retrieval(ComponentBase, ABC):
         embd_nms = list(set([kb.embd_id for kb in kbs]))
         assert len(embd_nms) == 1, "Knowledge bases use different embedding models."
 
-        embd_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.EMBEDDING, embd_nms[0])
-        self._canvas.set_embedding_model(embd_nms[0])
+        embd_mdl = None
+        if embd_nms:
+            embd_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.EMBEDDING, embd_nms[0])
+            self._canvas.set_embedding_model(embd_nms[0])
 
         rerank_mdl = None
         if self._param.rerank_id:
             rerank_mdl = LLMBundle(kbs[0].tenant_id, LLMType.RERANK, self._param.rerank_id)
 
-        kbinfos = settings.retrievaler.retrieval(query, embd_mdl, kbs[0].tenant_id, self._param.kb_ids,
+        if kbs:
+            kbinfos = settings.retrievaler.retrieval(query, embd_mdl, kbs[0].tenant_id, self._param.kb_ids,
                                         1, self._param.top_n,
                                         self._param.similarity_threshold, 1 - self._param.keywords_similarity_weight,
                                         aggs=False, rerank_mdl=rerank_mdl,
                                         rank_feature=label_question(query, kbs))
-        if self._param.use_kg:
+        else:
+            kbinfos = {"chunks": [], "doc_aggs": []}
+
+        if self._param.use_kg and kbs:
             ck = settings.kg_retrievaler.retrieval(query,
                                                    [kbs[0].tenant_id],
                                                    self._param.kb_ids,
@@ -99,10 +104,8 @@ class Retrieval(ComponentBase, ABC):
                 df["empty_response"] = self._param.empty_response
             return df
 
-        df = pd.DataFrame(kbinfos["chunks"])
-        df["content"] = df["content_with_weight"]
-        del df["content_with_weight"]
+        df = pd.DataFrame({"content": kb_prompt(kbinfos, 200000), "chunks": json.dumps(kbinfos["chunks"])})
         logging.debug("{} {}".format(query, df))
-        return df
+        return df.dropna()
 
 

agent/component/switch.py

@@ -54,7 +54,7 @@ class Switch(ComponentBase, ABC):
             for item in cond["items"]:
                 if not item["cpn_id"]:
                     continue
-                if item["cpn_id"].find("begin") >= 0:
+                if item["cpn_id"].lower().find("begin") >= 0 or item["cpn_id"].lower().find("answer") >= 0:
                     continue
                 cid = item["cpn_id"].split("@")[0]
                 res.append(cid)
@@ -75,7 +75,7 @@ class Switch(ComponentBase, ABC):
                             res.append(self.process_operator(p.get("value",""), item["operator"], item.get("value", "")))
                             break
                 else:
-                    out = self._canvas.get_component(cid)["obj"].output()[1]
+                    out = self._canvas.get_component(cid)["obj"].output(allow_partial=False)[1]
                     cpn_input = "" if "content" not in out.columns else " ".join([str(s) for s in out["content"]])
                     res.append(self.process_operator(cpn_input, item["operator"], item.get("value", "")))
 

agent/component/template.py

@@ -109,16 +109,14 @@ class Template(ComponentBase):
             pass
 
         for n, v in kwargs.items():
-            try:
-                v = json.dumps(v, ensure_ascii=False)
-            except Exception:
-                pass
+            if not isinstance(v, str):
+                try:
+                    v = json.dumps(v, ensure_ascii=False)
+                except Exception:
+                    pass
             content = re.sub(
                 r"\{%s\}" % re.escape(n), v, content
             )
-            content = re.sub(
-                r"(\\\"|\")", "", content
-            )
             content = re.sub(
                 r"(#+)", r" \1 ", content
             )

api/apps/__init__.py

@@ -83,7 +83,7 @@ app.errorhandler(Exception)(server_error_response)
 app.config["SESSION_PERMANENT"] = False
 app.config["SESSION_TYPE"] = "filesystem"
 app.config["MAX_CONTENT_LENGTH"] = int(
-    os.environ.get("MAX_CONTENT_LENGTH", 128 * 1024 * 1024)
+    os.environ.get("MAX_CONTENT_LENGTH", 1024 * 1024 * 1024)
 )
 
 Session(app)

api/apps/api_app.py

@@ -479,7 +479,7 @@ def upload():
                 doc = doc.to_dict()
                 doc["tenant_id"] = tenant_id
                 bucket, name = File2DocumentService.get_storage_address(doc_id=doc["id"])
-                queue_tasks(doc, bucket, name)
+                queue_tasks(doc, bucket, name, 0)
             except Exception as e:
                 return server_error_response(e)
 

api/apps/canvas_app.py

@@ -18,13 +18,16 @@ import traceback
 from flask import request, Response
 from flask_login import login_required, current_user
 from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService
+from api.db.services.user_service import TenantService
+from api.db.services.user_canvas_version import UserCanvasVersionService
 from api.settings import RetCode
 from api.utils import get_uuid
 from api.utils.api_utils import get_json_result, server_error_response, validate_request, get_data_error_result
 from agent.canvas import Canvas
 from peewee import MySQLDatabase, PostgresqlDatabase
 from api.db.db_models import APIToken
-
+import logging
+import time
 
 @manager.route('/templates', methods=['GET'])  # noqa: F821
 @login_required
@@ -61,7 +64,6 @@ def save():
     req["user_id"] = current_user.id
     if not isinstance(req["dsl"], str):
         req["dsl"] = json.dumps(req["dsl"], ensure_ascii=False)
-
     req["dsl"] = json.loads(req["dsl"])
     if "id" not in req:
         if UserCanvasService.query(user_id=current_user.id, title=req["title"].strip()):
@@ -75,16 +77,22 @@ def save():
                 data=False, message='Only owner of canvas authorized for this operation.',
                 code=RetCode.OPERATING_ERROR)
         UserCanvasService.update_by_id(req["id"], req)
+    # save version    
+    UserCanvasVersionService.insert( user_canvas_id=req["id"], dsl=req["dsl"], title="{0}_{1}".format(req["title"], time.strftime("%Y_%m_%d_%H_%M_%S")))
+    UserCanvasVersionService.delete_all_versions(req["id"])
     return get_json_result(data=req)
 
  
+
+
 @manager.route('/get/<canvas_id>', methods=['GET'])  # noqa: F821
 @login_required
 def get(canvas_id):
-    e, c = UserCanvasService.get_by_id(canvas_id)
+    e, c = UserCanvasService.get_by_tenant_id(canvas_id)
+    logging.info(f"get canvas_id: {canvas_id} c: {c}")
     if not e:
         return get_data_error_result(message="canvas not found.")
-    return get_json_result(data=c.to_dict())
+    return get_json_result(data=c)
 
 @manager.route('/getsse/<canvas_id>', methods=['GET'])  # type: ignore # noqa: F821
 def getsse(canvas_id):
@@ -283,4 +291,62 @@ def test_db_connect():
         return get_json_result(data="Database Connection Successful!")
     except Exception as e:
         return server_error_response(e)
+#api get list version dsl of canvas
+@manager.route('/getlistversion/<canvas_id>', methods=['GET'])  # noqa: F821
+@login_required
+def getlistversion(canvas_id):
+    try:
+        list =sorted([c.to_dict() for c in UserCanvasVersionService.list_by_canvas_id(canvas_id)], key=lambda x: x["update_time"]*-1)
+        return get_json_result(data=list)
+    except Exception as e:
+        return get_data_error_result(message=f"Error getting history files: {e}")
+#api get version dsl of canvas
+@manager.route('/getversion/<version_id>', methods=['GET'])  # noqa: F821
+@login_required
+def getversion( version_id):
+    try:
       
+        e, version = UserCanvasVersionService.get_by_id(version_id)
+        if version:
+            return get_json_result(data=version.to_dict())
+    except Exception as e:
+        return get_json_result(data=f"Error getting history file: {e}")
+@manager.route('/listteam', methods=['GET'])  # noqa: F821
+@login_required
+def list_kbs():
+    keywords = request.args.get("keywords", "")
+    page_number = int(request.args.get("page", 1))
+    items_per_page = int(request.args.get("page_size", 150))
+    orderby = request.args.get("orderby", "create_time")
+    desc = request.args.get("desc", True)
+    try:
+        tenants = TenantService.get_joined_tenants_by_user_id(current_user.id)
+        kbs, total = UserCanvasService.get_by_tenant_ids(
+            [m["tenant_id"] for m in tenants], current_user.id, page_number,
+            items_per_page, orderby, desc, keywords)
+        return get_json_result(data={"kbs": kbs, "total": total})
+    except Exception as e:
+        return server_error_response(e)
+@manager.route('/setting', methods=['POST'])  # noqa: F821
+@validate_request("id", "title", "permission")
+@login_required
+def setting():
+    req = request.json
+    req["user_id"] = current_user.id
+    e,flow = UserCanvasService.get_by_id(req["id"])
+    if not e:
+        return get_data_error_result(message="canvas not found.")
+    flow = flow.to_dict()
+    flow["title"] = req["title"]
+    if req["description"]:
+        flow["description"] = req["description"]
+    if req["permission"]:
+        flow["permission"] = req["permission"]
+    if req["avatar"]:
+        flow["avatar"] = req["avatar"]
+    if not UserCanvasService.query(user_id=current_user.id, id=req["id"]):
+        return get_json_result(
+            data=False, message='Only owner of canvas authorized for this operation.',
+            code=RetCode.OPERATING_ERROR)
+    num= UserCanvasService.update_by_id(req["id"], flow)
+    return get_json_result(data=num)

api/apps/conversation_app.py

@@ -17,26 +17,25 @@ import json
 import re
 import traceback
 from copy import deepcopy
+
 import trio
-from api.db.db_models import APIToken
-
-from api.db.services.conversation_service import ConversationService, structure_answer
-from api.db.services.user_service import UserTenantService
-from flask import request, Response
-from flask_login import login_required, current_user
+from flask import Response, request
+from flask_login import current_user, login_required
 
+from api import settings
 from api.db import LLMType
-from api.db.services.dialog_service import DialogService, chat, ask
+from api.db.db_models import APIToken
+from api.db.services.conversation_service import ConversationService, structure_answer
+from api.db.services.dialog_service import DialogService, ask, chat
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import LLMBundle, TenantService
-from api import settings
-from api.utils.api_utils import get_json_result
-from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
+from api.db.services.user_service import UserTenantService
+from api.utils.api_utils import get_data_error_result, get_json_result, server_error_response, validate_request
 from graphrag.general.mind_map_extractor import MindMapExtractor
 from rag.app.tag import label_question
 
 
-@manager.route('/set', methods=['POST'])  # noqa: F821
+@manager.route("/set", methods=["POST"])  # noqa: F821
 @login_required
 def set_conversation():
     req = request.json
@@ -50,8 +49,7 @@ def set_conversation():
                 return get_data_error_result(message="Conversation not found!")
             e, conv = ConversationService.get_by_id(conv_id)
             if not e:
-                return get_data_error_result(
-                    message="Fail to update a conversation!")
+                return get_data_error_result(message="Fail to update a conversation!")
             conv = conv.to_dict()
             return get_json_result(data=conv)
         except Exception as e:
@@ -61,38 +59,30 @@ def set_conversation():
         e, dia = DialogService.get_by_id(req["dialog_id"])
         if not e:
             return get_data_error_result(message="Dialog not found")
-        conv = {
-            "id": conv_id,
-            "dialog_id": req["dialog_id"],
-            "name": req.get("name", "New conversation"),
-            "message": [{"role": "assistant", "content": dia.prompt_config["prologue"]}]
-        }
+        conv = {"id": conv_id, "dialog_id": req["dialog_id"], "name": req.get("name", "New conversation"), "message": [{"role": "assistant", "content": dia.prompt_config["prologue"]}]}
         ConversationService.save(**conv)
         return get_json_result(data=conv)
     except Exception as e:
         return server_error_response(e)
 
 
-@manager.route('/get', methods=['GET'])  # noqa: F821
+@manager.route("/get", methods=["GET"])  # noqa: F821
 @login_required
 def get():
     conv_id = request.args["conversation_id"]
     try:
-        
         e, conv = ConversationService.get_by_id(conv_id)
         if not e:
             return get_data_error_result(message="Conversation not found!")
         tenants = UserTenantService.query(user_id=current_user.id)
-        avatar =None
+        avatar = None
         for tenant in tenants:
             dialog = DialogService.query(tenant_id=tenant.tenant_id, id=conv.dialog_id)
-            if dialog and len(dialog)>0:
+            if dialog and len(dialog) > 0:
                 avatar = dialog[0].icon
                 break
         else:
-            return get_json_result(
-                data=False, message='Only owner of conversation authorized for this operation.',
-                code=settings.RetCode.OPERATING_ERROR)
+            return get_json_result(data=False, message="Only owner of conversation authorized for this operation.", code=settings.RetCode.OPERATING_ERROR)
 
         def get_value(d, k1, k2):
             return d.get(k1, d.get(k2))
@@ -100,26 +90,29 @@ def get():
         for ref in conv.reference:
             if isinstance(ref, list):
                 continue
-            ref["chunks"] = [{
-                "id": get_value(ck, "chunk_id", "id"),
-                "content": get_value(ck, "content", "content_with_weight"),
-                "document_id": get_value(ck, "doc_id", "document_id"),
-                "document_name": get_value(ck, "docnm_kwd", "document_name"),
-                "dataset_id": get_value(ck, "kb_id", "dataset_id"),
-                "image_id": get_value(ck, "image_id", "img_id"),
-                "positions": get_value(ck, "positions", "position_int"),
-            } for ck in ref.get("chunks", [])]
+            ref["chunks"] = [
+                {
+                    "id": get_value(ck, "chunk_id", "id"),
+                    "content": get_value(ck, "content", "content_with_weight"),
+                    "document_id": get_value(ck, "doc_id", "document_id"),
+                    "document_name": get_value(ck, "docnm_kwd", "document_name"),
+                    "dataset_id": get_value(ck, "kb_id", "dataset_id"),
+                    "image_id": get_value(ck, "image_id", "img_id"),
+                    "positions": get_value(ck, "positions", "position_int"),
+                }
+                for ck in ref.get("chunks", [])
+            ]
 
         conv = conv.to_dict()
-        conv["avatar"]=avatar
+        conv["avatar"] = avatar
         return get_json_result(data=conv)
     except Exception as e:
         return server_error_response(e)
 
-@manager.route('/getsse/<dialog_id>', methods=['GET'])  # type: ignore # noqa: F821
-def getsse(dialog_id):
 
-    token = request.headers.get('Authorization').split()
+@manager.route("/getsse/<dialog_id>", methods=["GET"])  # type: ignore # noqa: F821
+def getsse(dialog_id):
+    token = request.headers.get("Authorization").split()
     if len(token) != 2:
         return get_data_error_result(message='Authorization is not valid!"')
     token = token[1]
@@ -131,13 +124,14 @@ def getsse(dialog_id):
         if not e:
             return get_data_error_result(message="Dialog not found!")
         conv = conv.to_dict()
-        conv["avatar"]= conv["icon"]
+        conv["avatar"] = conv["icon"]
         del conv["icon"]
         return get_json_result(data=conv)
     except Exception as e:
         return server_error_response(e)
 
-@manager.route('/rm', methods=['POST'])  # noqa: F821
+
+@manager.route("/rm", methods=["POST"])  # noqa: F821
 @login_required
 def rm():
     conv_ids = request.json["conversation_ids"]
@@ -151,28 +145,21 @@ def rm():
                 if DialogService.query(tenant_id=tenant.tenant_id, id=conv.dialog_id):
                     break
             else:
-                return get_json_result(
-                    data=False, message='Only owner of conversation authorized for this operation.',
-                    code=settings.RetCode.OPERATING_ERROR)
+                return get_json_result(data=False, message="Only owner of conversation authorized for this operation.", code=settings.RetCode.OPERATING_ERROR)
             ConversationService.delete_by_id(cid)
         return get_json_result(data=True)
     except Exception as e:
         return server_error_response(e)
 
 
-@manager.route('/list', methods=['GET'])  # noqa: F821
+@manager.route("/list", methods=["GET"])  # noqa: F821
 @login_required
 def list_convsersation():
     dialog_id = request.args["dialog_id"]
     try:
         if not DialogService.query(tenant_id=current_user.id, id=dialog_id):
-            return get_json_result(
-                data=False, message='Only owner of dialog authorized for this operation.',
-                code=settings.RetCode.OPERATING_ERROR)
-        convs = ConversationService.query(
-            dialog_id=dialog_id,
-            order_by=ConversationService.model.create_time,
-            reverse=True)
+            return get_json_result(data=False, message="Only owner of dialog authorized for this operation.", code=settings.RetCode.OPERATING_ERROR)
+        convs = ConversationService.query(dialog_id=dialog_id, order_by=ConversationService.model.create_time, reverse=True)
 
         convs = [d.to_dict() for d in convs]
         return get_json_result(data=convs)
@@ -180,7 +167,7 @@ def list_convsersation():
         return server_error_response(e)
 
 
-@manager.route('/completion', methods=['POST'])  # noqa: F821
+@manager.route("/completion", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("conversation_id", "messages")
 def completion():
@@ -207,25 +194,30 @@ def completion():
         if not conv.reference:
             conv.reference = []
         else:
+
             def get_value(d, k1, k2):
                 return d.get(k1, d.get(k2))
 
             for ref in conv.reference:
                 if isinstance(ref, list):
                     continue
-                ref["chunks"] = [{
-                    "id": get_value(ck, "chunk_id", "id"),
-                    "content": get_value(ck, "content", "content_with_weight"),
-                    "document_id": get_value(ck, "doc_id", "document_id"),
-                    "document_name": get_value(ck, "docnm_kwd", "document_name"),
-                    "dataset_id": get_value(ck, "kb_id", "dataset_id"),
-                    "image_id": get_value(ck, "image_id", "img_id"),
-                    "positions": get_value(ck, "positions", "position_int"),
-                } for ck in ref.get("chunks", [])]
+                ref["chunks"] = [
+                    {
+                        "id": get_value(ck, "chunk_id", "id"),
+                        "content": get_value(ck, "content", "content_with_weight"),
+                        "document_id": get_value(ck, "doc_id", "document_id"),
+                        "document_name": get_value(ck, "docnm_kwd", "document_name"),
+                        "dataset_id": get_value(ck, "kb_id", "dataset_id"),
+                        "image_id": get_value(ck, "image_id", "img_id"),
+                        "positions": get_value(ck, "positions", "position_int"),
+                    }
+                    for ck in ref.get("chunks", [])
+                ]
 
         if not conv.reference:
             conv.reference = []
         conv.reference.append({"chunks": [], "doc_aggs": []})
+
         def stream():
             nonlocal dia, msg, req, conv
             try:
@@ -235,9 +227,7 @@ def completion():
                 ConversationService.update_by_id(conv.id, conv.to_dict())
             except Exception as e:
                 traceback.print_exc()
-                yield "data:" + json.dumps({"code": 500, "message": str(e),
-                                            "data": {"answer": "**ERROR**: " + str(e), "reference": []}},
-                                           ensure_ascii=False) + "\n\n"
+                yield "data:" + json.dumps({"code": 500, "message": str(e), "data": {"answer": "**ERROR**: " + str(e), "reference": []}}, ensure_ascii=False) + "\n\n"
             yield "data:" + json.dumps({"code": 0, "message": "", "data": True}, ensure_ascii=False) + "\n\n"
 
         if req.get("stream", True):
@@ -259,7 +249,7 @@ def completion():
         return server_error_response(e)
 
 
-@manager.route('/tts', methods=['POST'])  # noqa: F821
+@manager.route("/tts", methods=["POST"])  # noqa: F821
 @login_required
 def tts():
     req = request.json
@@ -281,9 +271,7 @@ def tts():
                 for chunk in tts_mdl.tts(txt):
                     yield chunk
         except Exception as e:
-            yield ("data:" + json.dumps({"code": 500, "message": str(e),
-                                         "data": {"answer": "**ERROR**: " + str(e)}},
-                                        ensure_ascii=False)).encode('utf-8')
+            yield ("data:" + json.dumps({"code": 500, "message": str(e), "data": {"answer": "**ERROR**: " + str(e)}}, ensure_ascii=False)).encode("utf-8")
 
     resp = Response(stream_audio(), mimetype="audio/mpeg")
     resp.headers.add_header("Cache-Control", "no-cache")
@@ -293,7 +281,7 @@ def tts():
     return resp
 
 
-@manager.route('/delete_msg', methods=['POST'])  # noqa: F821
+@manager.route("/delete_msg", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("conversation_id", "message_id")
 def delete_msg():
@@ -316,7 +304,7 @@ def delete_msg():
     return get_json_result(data=conv)
 
 
-@manager.route('/thumbup', methods=['POST'])  # noqa: F821
+@manager.route("/thumbup", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("conversation_id", "message_id")
 def thumbup():
@@ -324,7 +312,7 @@ def thumbup():
     e, conv = ConversationService.get_by_id(req["conversation_id"])
     if not e:
         return get_data_error_result(message="Conversation not found!")
-    up_down = req.get("set")
+    up_down = req.get("thumbup")
     feedback = req.get("feedback", "")
     conv = conv.to_dict()
     for i, msg in enumerate(conv["message"]):
@@ -343,7 +331,7 @@ def thumbup():
     return get_json_result(data=conv)
 
 
-@manager.route('/ask', methods=['POST'])  # noqa: F821
+@manager.route("/ask", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("question", "kb_ids")
 def ask_about():
@@ -356,9 +344,7 @@ def ask_about():
             for ans in ask(req["question"], req["kb_ids"], uid):
                 yield "data:" + json.dumps({"code": 0, "message": "", "data": ans}, ensure_ascii=False) + "\n\n"
         except Exception as e:
-            yield "data:" + json.dumps({"code": 500, "message": str(e),
-                                        "data": {"answer": "**ERROR**: " + str(e), "reference": []}},
-                                       ensure_ascii=False) + "\n\n"
+            yield "data:" + json.dumps({"code": 500, "message": str(e), "data": {"answer": "**ERROR**: " + str(e), "reference": []}}, ensure_ascii=False) + "\n\n"
         yield "data:" + json.dumps({"code": 0, "message": "", "data": True}, ensure_ascii=False) + "\n\n"
 
     resp = Response(stream(), mimetype="text/event-stream")
@@ -369,7 +355,7 @@ def ask_about():
     return resp
 
 
-@manager.route('/mindmap', methods=['POST'])  # noqa: F821
+@manager.route("/mindmap", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("question", "kb_ids")
 def mindmap():
@@ -382,10 +368,7 @@ def mindmap():
     embd_mdl = LLMBundle(kb.tenant_id, LLMType.EMBEDDING, llm_name=kb.embd_id)
     chat_mdl = LLMBundle(current_user.id, LLMType.CHAT)
     question = req["question"]
-    ranks = settings.retrievaler.retrieval(question, embd_mdl, kb.tenant_id, kb_ids, 1, 12,
-                                           0.3, 0.3, aggs=False,
-                                           rank_feature=label_question(question, [kb])
-                                           )
+    ranks = settings.retrievaler.retrieval(question, embd_mdl, kb.tenant_id, kb_ids, 1, 12, 0.3, 0.3, aggs=False, rank_feature=label_question(question, [kb]))
     mindmap = MindMapExtractor(chat_mdl)
     mind_map = trio.run(mindmap, [c["content_with_weight"] for c in ranks["chunks"]])
     mind_map = mind_map.output
@@ -394,7 +377,7 @@ def mindmap():
     return get_json_result(data=mind_map)
 
 
-@manager.route('/related_questions', methods=['POST'])  # noqa: F821
+@manager.route("/related_questions", methods=["POST"])  # noqa: F821
 @login_required
 @validate_request("question")
 def related_questions():
@@ -402,31 +385,49 @@ def related_questions():
     question = req["question"]
     chat_mdl = LLMBundle(current_user.id, LLMType.CHAT)
     prompt = """
-Objective: To generate search terms related to the user's search keywords, helping users find more valuable information.
-Instructions:
- - Based on the keywords provided by the user, generate 5-10 related search terms.
- - Each search term should be directly or indirectly related to the keyword, guiding the user to find more valuable information.
- - Use common, general terms as much as possible, avoiding obscure words or technical jargon.
- - Keep the term length between 2-4 words, concise and clear.
- - DO NOT translate, use the language of the original keywords.
+Role: You are an AI language model assistant tasked with generating 5-10 related questions based on a user’s original query. These questions should help expand the search query scope and improve search relevance.
 
-### Example:
-Keywords: Chinese football
-Related search terms:
-1. Current status of Chinese football
-2. Reform of Chinese football
-3. Youth training of Chinese football
-4. Chinese football in the Asian Cup
-5. Chinese football in the World Cup
+Instructions:
+	Input: You are provided with a user’s question.
+	Output: Generate 5-10 alternative questions that are related to the original user question. These alternatives should help retrieve a broader range of relevant documents from a vector database.
+	Context: Focus on rephrasing the original question in different ways, making sure the alternative questions are diverse but still connected to the topic of the original query. Do not create overly obscure, irrelevant, or unrelated questions.
+	Fallback: If you cannot generate any relevant alternatives, do not return any questions.
+	Guidance:
+	1. Each alternative should be unique but still relevant to the original query.
+	2. Keep the phrasing clear, concise, and easy to understand.
+	3. Avoid overly technical jargon or specialized terms unless directly relevant.
+	4. Ensure that each question contributes towards improving search results by broadening the search angle, not narrowing it.
+
+Example:
+Original Question: What are the benefits of electric vehicles?
+
+Alternative Questions:
+	1. How do electric vehicles impact the environment?
+	2. What are the advantages of owning an electric car?
+	3. What is the cost-effectiveness of electric vehicles?
+	4. How do electric vehicles compare to traditional cars in terms of fuel efficiency?
+	5. What are the environmental benefits of switching to electric cars?
+	6. How do electric vehicles help reduce carbon emissions?
+	7. Why are electric vehicles becoming more popular?
+	8. What are the long-term savings of using electric vehicles?
+	9. How do electric vehicles contribute to sustainability?
+	10. What are the key benefits of electric vehicles for consumers?
 
 Reason:
- - When searching, users often only use one or two keywords, making it difficult to fully express their information needs.
- - Generating related search terms can help users dig deeper into relevant information and improve search efficiency. 
- - At the same time, related terms can also help search engines better understand user needs and return more accurate search results.
- 
+	Rephrasing the original query into multiple alternative questions helps the user explore different aspects of their search topic, improving the quality of search results.
+	These questions guide the search engine to provide a more comprehensive set of relevant documents.
 """
-    ans = chat_mdl.chat(prompt, [{"role": "user", "content": f"""
+    ans = chat_mdl.chat(
+        prompt,
+        [
+            {
+                "role": "user",
+                "content": f"""
 Keywords: {question}
 Related search terms:
-    """}], {"temperature": 0.9})
+    """,
+            }
+        ],
+        {"temperature": 0.9},
+    )
     return get_json_result(data=[re.sub(r"^[0-9]\. ", "", a) for a in ans.split("\n") if re.match(r"^[0-9]\. ", a)])

api/apps/document_app.py

@@ -331,10 +331,10 @@ def rm():
                     message="Database error (Document removal)!")
 
             f2d = File2DocumentService.get_by_document_id(doc_id)
-            FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == f2d[0].file_id])
+            deleted_file_count = FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == f2d[0].file_id])
             File2DocumentService.delete_by_document_id(doc_id)
-
-            STORAGE_IMPL.rm(b, n)
+            if deleted_file_count > 0:
+                STORAGE_IMPL.rm(b, n)
         except Exception as e:
             errors += str(e)
 
@@ -380,7 +380,7 @@ def run():
                 doc = doc.to_dict()
                 doc["tenant_id"] = tenant_id
                 bucket, name = File2DocumentService.get_storage_address(doc_id=doc["id"])
-                queue_tasks(doc, bucket, name)
+                queue_tasks(doc, bucket, name, 0)
 
         return get_json_result(data=True)
     except Exception as e:

api/apps/file2document_app.py

@@ -38,8 +38,12 @@ def convert():
     file2documents = []
 
     try:
+        files = FileService.get_by_ids(file_ids)
+        files_set = dict({file.id: file for file in files})
         for file_id in file_ids:
-            e, file = FileService.get_by_id(file_id)
+            file = files_set[file_id]
+            if not file:
+                return get_data_error_result(message="File not found!")
             file_ids_list = [file_id]
             if file.type == FileType.FOLDER.value:
                 file_ids_list = FileService.get_all_innermost_file_ids(file_id, [])
@@ -86,6 +90,7 @@ def convert():
                         "file_id": id,
                         "document_id": doc.id,
                     })
+
                     file2documents.append(file2document.to_json())
         return get_json_result(data=file2documents)
     except Exception as e:

api/apps/file_app.py

@@ -55,20 +55,17 @@ def upload():
                 data=False, message='No file selected!', code=settings.RetCode.ARGUMENT_ERROR)
     file_res = []
     try:
+        e, pf_folder = FileService.get_by_id(pf_id)
+        if not e:
+            return get_data_error_result( message="Can't find this folder!")
         for file_obj in file_objs:
-            e, file = FileService.get_by_id(pf_id)
-            if not e:
-                return get_data_error_result(
-                    message="Can't find this folder!")
             MAX_FILE_NUM_PER_USER = int(os.environ.get('MAX_FILE_NUM_PER_USER', 0))
             if MAX_FILE_NUM_PER_USER > 0 and DocumentService.get_doc_count(current_user.id) >= MAX_FILE_NUM_PER_USER:
-                return get_data_error_result(
-                    message="Exceed the maximum file number of a free user!")
+                return get_data_error_result( message="Exceed the maximum file number of a free user!")
 
             # split file name path
             if not file_obj.filename:
-                e, file = FileService.get_by_id(pf_id)
-                file_obj_names = [file.name, file_obj.filename]
+                file_obj_names = [pf_folder.name, file_obj.filename]
             else:
                 full_path = '/' + file_obj.filename
                 file_obj_names = full_path.split('/')
@@ -184,7 +181,7 @@ def list_files():
             current_user.id, pf_id, page_number, items_per_page, orderby, desc, keywords)
 
         parent_folder = FileService.get_parent_folder(pf_id)
-        if not FileService.get_parent_folder(pf_id):
+        if not parent_folder:
             return get_json_result(message="File not found!")
 
         return get_json_result(data={"total": total, "files": files, "parent_folder": parent_folder.to_json()})
@@ -358,9 +355,14 @@ def move():
     try:
         file_ids = req["src_file_ids"]
         parent_id = req["dest_file_id"]
+        files = FileService.get_by_ids(file_ids)
+        files_dict = {}
+        for file in files:
+            files_dict[file.id] = file
+
         for file_id in file_ids:
-            e, file = FileService.get_by_id(file_id)
-            if not e:
+            file = files_dict[file_id]
+            if not file:
                 return get_data_error_result(message="File or Folder not found!")
             if not file.tenant_id:
                 return get_data_error_result(message="Tenant not found!")

api/apps/kb_app.py

@@ -73,7 +73,7 @@ def create():
 
 @manager.route('/update', methods=['post'])  # noqa: F821
 @login_required
-@validate_request("kb_id", "name", "description", "permission", "parser_id")
+@validate_request("kb_id", "name", "description", "parser_id")
 @not_allowed_parameters("id", "tenant_id", "created_by", "create_time", "update_time", "create_date", "update_date", "created_by")
 def update():
     req = request.json
@@ -157,25 +157,38 @@ def detail():
         return server_error_response(e)
 
 
-@manager.route('/list', methods=['GET'])  # noqa: F821
+@manager.route('/list', methods=['POST'])  # noqa: F821
 @login_required
 def list_kbs():
     keywords = request.args.get("keywords", "")
-    page_number = int(request.args.get("page", 1))
-    items_per_page = int(request.args.get("page_size", 150))
+    page_number = int(request.args.get("page", 0))
+    items_per_page = int(request.args.get("page_size", 0))
     parser_id = request.args.get("parser_id")
     orderby = request.args.get("orderby", "create_time")
     desc = request.args.get("desc", True)
+
+    req = request.get_json()
+    owner_ids = req.get("owner_ids", [])
     try:
-        tenants = TenantService.get_joined_tenants_by_user_id(current_user.id)
-        kbs, total = KnowledgebaseService.get_by_tenant_ids(
-            [m["tenant_id"] for m in tenants], current_user.id, page_number,
-            items_per_page, orderby, desc, keywords, parser_id)
+        if not owner_ids:
+            tenants = TenantService.get_joined_tenants_by_user_id(current_user.id)
+            tenants = [m["tenant_id"] for m in tenants]
+            kbs, total = KnowledgebaseService.get_by_tenant_ids(
+                tenants, current_user.id, page_number,
+                items_per_page, orderby, desc, keywords, parser_id)
+        else:
+            tenants = owner_ids
+            kbs, total = KnowledgebaseService.get_by_tenant_ids(
+                tenants, current_user.id, 0,
+                0, orderby, desc, keywords, parser_id)
+            kbs = [kb for kb in kbs if kb["tenant_id"] in tenants]
+            if page_number and items_per_page:
+                kbs = kbs[(page_number-1)*items_per_page:page_number*items_per_page]
+            total = len(kbs)
         return get_json_result(data={"kbs": kbs, "total": total})
     except Exception as e:
         return server_error_response(e)
 
-
 @manager.route('/rm', methods=['post'])  # noqa: F821
 @login_required
 @validate_request("kb_id")
@@ -323,3 +336,17 @@ def knowledge_graph(kb_id):
             filtered_edges = [o for o in obj["graph"]["edges"] if o["source"] != o["target"] and o["source"] in node_id_set and o["target"] in node_id_set]
             obj["graph"]["edges"] = sorted(filtered_edges, key=lambda x: x.get("weight", 0), reverse=True)[:128]
     return get_json_result(data=obj)
+
+@manager.route('/<kb_id>/knowledge_graph', methods=['DELETE'])  # noqa: F821
+@login_required
+def delete_knowledge_graph(kb_id):
+    if not KnowledgebaseService.accessible(kb_id, current_user.id):
+        return get_json_result(
+            data=False,
+            message='No authorization.',
+            code=settings.RetCode.AUTHENTICATION_ERROR
+        )
+    _, kb = KnowledgebaseService.get_by_id(kb_id)
+    settings.docStoreConn.delete({"knowledge_graph_kwd": ["graph", "subgraph", "entity", "relation"]}, search.index_name(kb.tenant_id), kb_id)
+
+    return get_json_result(data=True)

api/apps/langfuse_app.py

@@ -0,0 +1,97 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+
+from flask import request
+from flask_login import current_user, login_required
+from langfuse import Langfuse
+
+from api.db.db_models import DB
+from api.db.services.langfuse_service import TenantLangfuseService
+from api.utils.api_utils import get_error_data_result, get_json_result, server_error_response, validate_request
+
+
+@manager.route("/api_key", methods=["POST", "PUT"])  # noqa: F821
+@login_required
+@validate_request("secret_key", "public_key", "host")
+def set_api_key():
+    req = request.get_json()
+    secret_key = req.get("secret_key", "")
+    public_key = req.get("public_key", "")
+    host = req.get("host", "")
+    if not all([secret_key, public_key, host]):
+        return get_error_data_result(message="Missing required fields")
+
+    langfuse_keys = dict(
+        tenant_id=current_user.id,
+        secret_key=secret_key,
+        public_key=public_key,
+        host=host,
+    )
+
+    langfuse = Langfuse(public_key=langfuse_keys["public_key"], secret_key=langfuse_keys["secret_key"], host=langfuse_keys["host"])
+    if not langfuse.auth_check():
+        return get_error_data_result(message="Invalid Langfuse keys")
+
+    langfuse_entry = TenantLangfuseService.filter_by_tenant(tenant_id=current_user.id)
+    with DB.atomic():
+        try:
+            if not langfuse_entry:
+                TenantLangfuseService.save(**langfuse_keys)
+            else:
+                TenantLangfuseService.update_by_tenant(tenant_id=current_user.id, langfuse_keys=langfuse_keys)
+            return get_json_result(data=langfuse_keys)
+        except Exception as e:
+            server_error_response(e)
+
+
+@manager.route("/api_key", methods=["GET"])  # noqa: F821
+@login_required
+@validate_request()
+def get_api_key():
+    langfuse_entry = TenantLangfuseService.filter_by_tenant_with_info(tenant_id=current_user.id)
+    if not langfuse_entry:
+        return get_json_result(message="Have not record any Langfuse keys.")
+
+    langfuse = Langfuse(public_key=langfuse_entry["public_key"], secret_key=langfuse_entry["secret_key"], host=langfuse_entry["host"])
+    try:
+        if not langfuse.auth_check():
+            return get_error_data_result(message="Invalid Langfuse keys loaded")
+    except langfuse.api.core.api_error.ApiError as api_err:
+        return get_json_result(message=f"Error from Langfuse: {api_err}")
+    except Exception as e:
+        server_error_response(e)
+
+    langfuse_entry["project_id"] = langfuse.api.projects.get().dict()["data"][0]["id"]
+    langfuse_entry["project_name"] = langfuse.api.projects.get().dict()["data"][0]["name"]
+
+    return get_json_result(data=langfuse_entry)
+
+
+@manager.route("/api_key", methods=["DELETE"])  # noqa: F821
+@login_required
+@validate_request()
+def delete_api_key():
+    langfuse_entry = TenantLangfuseService.filter_by_tenant(tenant_id=current_user.id)
+    if not langfuse_entry:
+        return get_json_result(message="Have not record any Langfuse keys.")
+
+    with DB.atomic():
+        try:
+            TenantLangfuseService.delete_model(langfuse_entry)
+            return get_json_result(data=True)
+        except Exception as e:
+            server_error_response(e)

api/apps/llm_app.py

@@ -61,6 +61,7 @@ def set_api_key():
     msg = ""
     for llm in LLMService.query(fid=factory):
         if not embd_passed and llm.model_type == LLMType.EMBEDDING.value:
+            assert factory in EmbeddingModel, f"Embedding model from {factory} is not supported yet."
             mdl = EmbeddingModel[factory](
                 req["api_key"], llm.llm_name, base_url=req.get("base_url"))
             try:
@@ -71,6 +72,7 @@ def set_api_key():
             except Exception as e:
                 msg += f"\nFail to access embedding model({llm.llm_name}) using this api key." + str(e)
         elif not chat_passed and llm.model_type == LLMType.CHAT.value:
+            assert factory in ChatModel, f"Chat model from {factory} is not supported yet."
             mdl = ChatModel[factory](
                 req["api_key"], llm.llm_name, base_url=req.get("base_url"))
             try:
@@ -83,6 +85,7 @@ def set_api_key():
                 msg += f"\nFail to access model({llm.llm_name}) using this api key." + str(
                     e)
         elif not rerank_passed and llm.model_type == LLMType.RERANK:
+            assert factory in RerankModel, f"Re-rank model from {factory} is not supported yet."
             mdl = RerankModel[factory](
                 req["api_key"], llm.llm_name, base_url=req.get("base_url"))
             try:
@@ -136,7 +139,7 @@ def add_llm():
     req = request.json
     factory = req["llm_factory"]
     api_key = req.get("api_key", "x")
-    llm_name = req["llm_name"]
+    llm_name = req.get("llm_name")
 
     def apikey_json(keys):
         nonlocal req
@@ -203,6 +206,7 @@ def add_llm():
     msg = ""
     mdl_nm = llm["llm_name"].split("___")[0]
     if llm["model_type"] == LLMType.EMBEDDING.value:
+        assert factory in EmbeddingModel, f"Embedding model from {factory} is not supported yet."
         mdl = EmbeddingModel[factory](
             key=llm['api_key'],
             model_name=mdl_nm,
@@ -214,6 +218,7 @@ def add_llm():
         except Exception as e:
             msg += f"\nFail to access embedding model({mdl_nm})." + str(e)
     elif llm["model_type"] == LLMType.CHAT.value:
+        assert factory in ChatModel, f"Chat model from {factory} is not supported yet."
         mdl = ChatModel[factory](
             key=llm['api_key'],
             model_name=mdl_nm,
@@ -228,6 +233,7 @@ def add_llm():
             msg += f"\nFail to access model({mdl_nm})." + str(
                 e)
     elif llm["model_type"] == LLMType.RERANK:
+        assert factory in RerankModel, f"RE-rank model from {factory} is not supported yet."
         try:
             mdl = RerankModel[factory](
                 key=llm["api_key"],
@@ -243,6 +249,7 @@ def add_llm():
             msg += f"\nFail to access model({mdl_nm})." + str(
                 e)
     elif llm["model_type"] == LLMType.IMAGE2TEXT.value:
+        assert factory in CvModel, f"Image to text model from {factory} is not supported yet."
         mdl = CvModel[factory](
             key=llm["api_key"],
             model_name=mdl_nm,
@@ -256,6 +263,7 @@ def add_llm():
         except Exception as e:
             msg += f"\nFail to access model({mdl_nm})." + str(e)
     elif llm["model_type"] == LLMType.TTS:
+        assert factory in TTSModel, f"TTS model from {factory} is not supported yet."
         mdl = TTSModel[factory](
             key=llm["api_key"], model_name=mdl_nm, base_url=llm["api_base"]
         )

api/apps/sdk/chat.py

@@ -23,8 +23,8 @@ from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import TenantLLMService
 from api.db.services.user_service import TenantService
 from api.utils import get_uuid
-from api.utils.api_utils import get_error_data_result, token_required
-from api.utils.api_utils import get_result
+from api.utils.api_utils import get_error_data_result, token_required, get_result, check_duplicate_ids
+
 
 
 @manager.route('/chats', methods=['POST'])  # noqa: F821
@@ -41,11 +41,6 @@ def create(tenant_id):
         if kb.chunk_num == 0:
             return get_error_data_result(f"The dataset {kb_id} doesn't own parsed file")
     
-        # Check if all documents in the knowledge base have been parsed
-        is_done, error_msg = KnowledgebaseService.is_parsed_done(kb_id)
-        if not is_done:
-            return get_error_data_result(error_msg)
-    
     kbs = KnowledgebaseService.get_by_ids(ids) if ids else []
     embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs]  # remove vendor suffix for comparison
     embd_count = list(set(embd_ids))
@@ -183,11 +178,6 @@ def update(tenant_id, chat_id):
                 if kb.chunk_num == 0:
                     return get_error_data_result(f"The dataset {kb_id} doesn't own parsed file")
                 
-                # Check if all documents in the knowledge base have been parsed
-                is_done, error_msg = KnowledgebaseService.is_parsed_done(kb_id)
-                if not is_done:
-                    return get_error_data_result(error_msg)
-                
             kbs = KnowledgebaseService.get_by_ids(ids)
             embd_ids = [TenantLLMService.split_model_name_and_factory(kb.embd_id)[0] for kb in kbs]  # remove vendor suffix for comparison
             embd_count = list(set(embd_ids))
@@ -233,11 +223,11 @@ def update(tenant_id, chat_id):
             return get_error_data_result(f"`rerank_model` {req.get('rerank_id')} doesn't exist")
     if "name" in req:
         if not req.get("name"):
-            return get_error_data_result(message="`name` is not empty.")
+            return get_error_data_result(message="`name` cannot be empty.")
         if req["name"].lower() != res["name"].lower() \
                 and len(
             DialogService.query(name=req["name"], tenant_id=tenant_id, status=StatusEnum.VALID.value)) > 0:
-            return get_error_data_result(message="Duplicated chat name in updating dataset.")
+            return get_error_data_result(message="Duplicated chat name in updating chat.")
     if "prompt_config" in req:
         res["prompt_config"].update(req["prompt_config"])
         for p in res["prompt_config"]["parameters"]:
@@ -262,6 +252,8 @@ def update(tenant_id, chat_id):
 @manager.route('/chats', methods=['DELETE'])  # noqa: F821
 @token_required
 def delete(tenant_id):
+    errors = []
+    success_count = 0
     req = request.json
     if not req:
         ids = None
@@ -274,14 +266,39 @@ def delete(tenant_id):
             id_list.append(dia.id)
     else:
         id_list = ids
-    for id in id_list:
+
+    unique_id_list, duplicate_messages = check_duplicate_ids(id_list, "assistant")
+
+    for id in unique_id_list:
         if not DialogService.query(tenant_id=tenant_id, id=id, status=StatusEnum.VALID.value):
-            return get_error_data_result(message=f"You don't own the chat {id}")
+            errors.append(f"Assistant({id}) not found.")
+            continue
         temp_dict = {"status": StatusEnum.INVALID.value}
         DialogService.update_by_id(id, temp_dict)
+        success_count += 1
+        
+    if errors:
+        if success_count > 0:
+            return get_result(
+                data={"success_count": success_count, "errors": errors},
+                message=f"Partially deleted {success_count} chats with {len(errors)} errors"
+            )
+        else:
+            return get_error_data_result(message="; ".join(errors))
+    
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(
+                message=f"Partially deleted {success_count} chats with {len(duplicate_messages)} errors", 
+                data={"success_count": success_count, "errors": duplicate_messages}
+            )
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
+    
     return get_result()
 
 
+
 @manager.route('/chats', methods=['GET'])  # noqa: F821
 @token_required
 def list_chat(tenant_id):

api/apps/sdk/dataset.py

@@ -30,7 +30,7 @@ from api.utils.api_utils import (
     token_required,
     get_error_data_result,
     valid,
-    get_parser_config, valid_parser_config,
+    get_parser_config, valid_parser_config, dataset_readonly_fields,check_duplicate_ids
 )
 
 
@@ -69,7 +69,7 @@ def create(tenant_id):
             chunk_method:
               type: string
               enum: ["naive", "manual", "qa", "table", "paper", "book", "laws",
-                     "presentation", "picture", "one", "knowledge_graph", "email", "tag"
+                     "presentation", "picture", "one", "email", "tag"
                      ]
               description: Chunking method.
             parser_config:
@@ -85,6 +85,9 @@ def create(tenant_id):
               type: object
     """
     req = request.json
+    for k in req.keys():
+        if dataset_readonly_fields(k):
+            return get_result(code=settings.RetCode.ARGUMENT_ERROR, message=f"'{k}' is readonly.")
     e, t = TenantService.get_by_id(tenant_id)
     permission = req.get("permission")
     chunk_method = req.get("chunk_method")
@@ -102,7 +105,6 @@ def create(tenant_id):
         "presentation",
         "picture",
         "one",
-        "knowledge_graph",
         "email",
         "tag"
     ]
@@ -144,16 +146,6 @@ def create(tenant_id):
     else:
         valid_embedding_models = [
             "BAAI/bge-large-zh-v1.5",
-            "BAAI/bge-base-en-v1.5",
-            "BAAI/bge-large-en-v1.5",
-            "BAAI/bge-small-en-v1.5",
-            "BAAI/bge-small-zh-v1.5",
-            "jinaai/jina-embeddings-v2-base-en",
-            "jinaai/jina-embeddings-v2-small-en",
-            "nomic-ai/nomic-embed-text-v1.5",
-            "sentence-transformers/all-MiniLM-L6-v2",
-            "text-embedding-v2",
-            "text-embedding-v3",
             "maidalun1020/bce-embedding-base_v1",
         ]
         embd_model = LLMService.query(
@@ -242,6 +234,9 @@ def delete(tenant_id):
             id_list.append(kb.id)
     else:
         id_list = ids
+    unique_id_list, duplicate_messages = check_duplicate_ids(id_list, "dataset")
+    id_list = unique_id_list
+
     for id in id_list:
         kbs = KnowledgebaseService.query(id=id, tenant_id=tenant_id)
         if not kbs:
@@ -273,6 +268,11 @@ def delete(tenant_id):
             )
         else:
             return get_error_data_result(message="; ".join(errors))
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(message=f"Partially deleted {success_count} datasets with {len(duplicate_messages)} errors", data={"success_count": success_count, "errors": duplicate_messages},)
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
     return get_result(code=settings.RetCode.SUCCESS)
 
 
@@ -314,7 +314,7 @@ def update(tenant_id, dataset_id):
             chunk_method:
               type: string
               enum: ["naive", "manual", "qa", "table", "paper", "book", "laws",
-                     "presentation", "picture", "one", "knowledge_graph", "email", "tag"
+                     "presentation", "picture", "one", "email", "tag"
                      ]
               description: Updated chunking method.
             parser_config:
@@ -329,6 +329,9 @@ def update(tenant_id, dataset_id):
     if not KnowledgebaseService.query(id=dataset_id, tenant_id=tenant_id):
         return get_error_data_result(message="You don't own the dataset")
     req = request.json
+    for k in req.keys():
+        if dataset_readonly_fields(k):
+            return get_result(code=settings.RetCode.ARGUMENT_ERROR, message=f"'{k}' is readonly.")
     e, t = TenantService.get_by_id(tenant_id)
     invalid_keys = {"id", "embd_id", "chunk_num", "doc_num", "parser_id", "create_date", "create_time", "created_by", "status","token_num","update_date","update_time"}
     if any(key in req for key in invalid_keys):
@@ -349,7 +352,6 @@ def update(tenant_id, dataset_id):
         "presentation",
         "picture",
         "one",
-        "knowledge_graph",
         "email",
         "tag"
     ]

api/apps/sdk/doc.py

@@ -36,7 +36,7 @@ from api.db.services.document_service import DocumentService
 from api.db.services.file2document_service import File2DocumentService
 from api.db.services.file_service import FileService
 from api.db.services.knowledgebase_service import KnowledgebaseService
-from api.utils.api_utils import construct_json_result, get_parser_config
+from api.utils.api_utils import construct_json_result, get_parser_config, check_duplicate_ids
 from rag.nlp import search
 from rag.prompts import keyword_extraction
 from rag.app.tag import label_question
@@ -67,6 +67,7 @@ class Chunk(BaseModel):
                 raise ValueError("Each sublist in positions must have a length of 5")
         return value
 
+
 @manager.route("/datasets/<dataset_id>/documents", methods=["POST"])  # noqa: F821
 @token_required
 def upload(dataset_id, tenant_id):
@@ -136,6 +137,10 @@ def upload(dataset_id, tenant_id):
             return get_result(
                 message="No file selected!", code=settings.RetCode.ARGUMENT_ERROR
             )
+        if len(file_obj.filename.encode("utf-8")) >= 128:
+            return get_result(
+                message="File name should be less than 128 bytes.", code=settings.RetCode.ARGUMENT_ERROR
+            )
     '''
     # total size
     total_size = 0
@@ -246,6 +251,11 @@ def update_doc(tenant_id, dataset_id, document_id):
         DocumentService.update_meta_fields(document_id, req["meta_fields"])
 
     if "name" in req and req["name"] != doc.name:
+        if len(req["name"].encode("utf-8")) >= 128:
+            return get_result(
+                message="The name should be less than 128 bytes.",
+                code=settings.RetCode.ARGUMENT_ERROR,
+            )
         if (
                 pathlib.Path(req["name"].lower()).suffix
                 != pathlib.Path(doc.name.lower()).suffix
@@ -363,6 +373,10 @@ def download(tenant_id, dataset_id, document_id):
         schema:
           type: object
     """
+    if not document_id:
+        return get_error_data_result(
+            message="Specify document_id please."
+        )
     if not KnowledgebaseService.query(id=dataset_id, tenant_id=tenant_id):
         return get_error_data_result(message=f"You do not own the dataset {dataset_id}.")
     doc = DocumentService.query(kb_id=dataset_id, id=document_id)
@@ -578,15 +592,22 @@ def delete(tenant_id, dataset_id):
             doc_list.append(doc.id)
     else:
         doc_list = doc_ids
+
+    unique_doc_ids, duplicate_messages = check_duplicate_ids(doc_list, "document")
+    doc_list = unique_doc_ids
+
     root_folder = FileService.get_root_folder(tenant_id)
     pf_id = root_folder["id"]
     FileService.init_knowledgebase_docs(pf_id, tenant_id)
     errors = ""
+    not_found = []
+    success_count = 0
     for doc_id in doc_list:
         try:
             e, doc = DocumentService.get_by_id(doc_id)
             if not e:
-                return get_error_data_result(message="Document not found!")
+                not_found.append(doc_id)
+                continue
             tenant_id = DocumentService.get_tenant_id(doc_id)
             if not tenant_id:
                 return get_error_data_result(message="Tenant not found!")
@@ -608,12 +629,22 @@ def delete(tenant_id, dataset_id):
             File2DocumentService.delete_by_document_id(doc_id)
 
             STORAGE_IMPL.rm(b, n)
+            success_count += 1
         except Exception as e:
             errors += str(e)
 
+    if not_found:
+        return get_result(message=f"Documents not found: {not_found}", code=settings.RetCode.DATA_ERROR)
+
     if errors:
         return get_result(message=errors, code=settings.RetCode.SERVER_ERROR)
 
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(message=f"Partially deleted {success_count} datasets with {len(duplicate_messages)} errors", data={"success_count": success_count, "errors": duplicate_messages},)
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
+
     return get_result()
 
 
@@ -661,18 +692,24 @@ def parse(tenant_id, dataset_id):
     req = request.json
     if not req.get("document_ids"):
         return get_error_data_result("`document_ids` is required")
-    for id in req["document_ids"]:
+    doc_list = req.get("document_ids")
+    unique_doc_ids, duplicate_messages = check_duplicate_ids(doc_list, "document")
+    doc_list = unique_doc_ids
+
+    not_found = []
+    success_count = 0
+    for id in doc_list:
         doc = DocumentService.query(id=id, kb_id=dataset_id)
+        if not doc:
+            not_found.append(id)
+            continue
         if not doc:
             return get_error_data_result(message=f"You don't own the document {id}.")
-        if doc[0].progress != 0.0:
+        if 0.0 < doc[0].progress < 1.0:
             return get_error_data_result(
-                "Can't stop parsing document with progress at 0 or 100"
+                "Can't parse document that is currently being processed"
             )
-        info = {"run": "1", "progress": 0}
-        info["progress_msg"] = ""
-        info["chunk_num"] = 0
-        info["token_num"] = 0
+        info = {"run": "1", "progress": 0, "progress_msg": "", "chunk_num": 0, "token_num": 0}
         DocumentService.update_by_id(id, info)
         settings.docStoreConn.delete({"doc_id": id}, search.index_name(tenant_id), dataset_id)
         TaskService.filter_delete([Task.doc_id == id])
@@ -680,7 +717,16 @@ def parse(tenant_id, dataset_id):
         doc = doc.to_dict()
         doc["tenant_id"] = tenant_id
         bucket, name = File2DocumentService.get_storage_address(doc_id=doc["id"])
-        queue_tasks(doc, bucket, name)
+        queue_tasks(doc, bucket, name, 0)
+        success_count += 1
+    if not_found:
+        return get_result(message=f"Documents not found: {not_found}", code=settings.RetCode.DATA_ERROR)
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(message=f"Partially parsed {success_count} documents with {len(duplicate_messages)} errors", data={"success_count": success_count, "errors": duplicate_messages},)
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
+
     return get_result()
 
 
@@ -726,9 +772,15 @@ def stop_parsing(tenant_id, dataset_id):
     if not KnowledgebaseService.accessible(kb_id=dataset_id, user_id=tenant_id):
         return get_error_data_result(message=f"You don't own the dataset {dataset_id}.")
     req = request.json
+
     if not req.get("document_ids"):
         return get_error_data_result("`document_ids` is required")
-    for id in req["document_ids"]:
+    doc_list = req.get("document_ids")
+    unique_doc_ids, duplicate_messages = check_duplicate_ids(doc_list, "document")
+    doc_list = unique_doc_ids
+
+    success_count = 0
+    for id in doc_list:
         doc = DocumentService.query(id=id, kb_id=dataset_id)
         if not doc:
             return get_error_data_result(message=f"You don't own the document {id}.")
@@ -739,6 +791,12 @@ def stop_parsing(tenant_id, dataset_id):
         info = {"run": "2", "progress": 0, "chunk_num": 0}
         DocumentService.update_by_id(id, info)
         settings.docStoreConn.delete({"doc_id": doc[0].id}, search.index_name(tenant_id), dataset_id)
+        success_count += 1
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(message=f"Partially stopped {success_count} documents with {len(duplicate_messages)} errors", data={"success_count": success_count, "errors": duplicate_messages},)
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
     return get_result()
 
 
@@ -859,6 +917,8 @@ def list_chunks(tenant_id, dataset_id, document_id):
     res = {"total": 0, "chunks": [], "doc": renamed_doc}
     if req.get("id"):
         chunk = settings.docStoreConn.get(req.get("id"), search.index_name(tenant_id), [dataset_id])
+        if not chunk:
+            return get_result(message=f"Chunk not found: {dataset_id}/{req.get('id')}", code=settings.RetCode.NOT_FOUND)
         k = []
         for n in chunk.keys():
             if re.search(r"(_vec$|_sm_|_tks|_ltks)", n):
@@ -876,7 +936,7 @@ def list_chunks(tenant_id, dataset_id, document_id):
             "important_keywords":chunk.get("important_kwd",[]),
             "questions":chunk.get("question_kwd",[]),
             "dataset_id":chunk.get("kb_id",chunk.get("dataset_id")),
-            "image_id":chunk["img_id"],
+            "image_id":chunk.get("img_id", ""),
             "available":bool(chunk.get("available_int",1)),
             "positions":chunk.get("position_int",[]),
         }
@@ -901,7 +961,7 @@ def list_chunks(tenant_id, dataset_id, document_id):
                 "questions": sres.field[id].get("question_kwd", []),
                 "dataset_id": sres.field[id].get("kb_id", sres.field[id].get("dataset_id")),
                 "image_id": sres.field[id].get("img_id", ""),
-                "available": bool(sres.field[id].get("available_int", 1)),
+                "available": bool(int(sres.field[id].get("available_int", "1"))),
                 "positions": sres.field[id].get("position_int",[]),
             }
             res["chunks"].append(d)
@@ -986,7 +1046,7 @@ def add_chunk(tenant_id, dataset_id, document_id):
         )
     doc = doc[0]
     req = request.json
-    if not req.get("content"):
+    if not str(req.get("content", "")).strip():
         return get_error_data_result(message="`content` is required")
     if "important_keywords" in req:
         if not isinstance(req["important_keywords"], list):
@@ -1009,7 +1069,7 @@ def add_chunk(tenant_id, dataset_id, document_id):
     d["important_tks"] = rag_tokenizer.tokenize(
         " ".join(req.get("important_keywords", []))
     )
-    d["question_kwd"] = req.get("questions", [])
+    d["question_kwd"] = [str(q).strip() for q in req.get("questions", []) if str(q).strip()]
     d["question_tks"] = rag_tokenizer.tokenize(
         "\n".join(req.get("questions", []))
     )
@@ -1098,15 +1158,23 @@ def rm_chunk(tenant_id, dataset_id, document_id):
     """
     if not KnowledgebaseService.accessible(kb_id=dataset_id, user_id=tenant_id):
         return get_error_data_result(message=f"You don't own the dataset {dataset_id}.")
+    docs = DocumentService.get_by_ids([document_id])
+    if not docs:
+        raise LookupError(f"Can't find the document with ID {document_id}!")
     req = request.json
     condition = {"doc_id": document_id}
     if "chunk_ids" in req:
-        condition["id"] = req["chunk_ids"]
+        unique_chunk_ids, duplicate_messages = check_duplicate_ids(req["chunk_ids"], "chunk")
+        condition["id"] = unique_chunk_ids
     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)
-    if "chunk_ids" in req and chunk_number != len(req["chunk_ids"]):
-        return get_error_data_result(message=f"rm_chunk deleted chunks {chunk_number}, expect {len(req['chunk_ids'])}")
+    if "chunk_ids" in req and chunk_number != len(unique_chunk_ids):
+        if len(unique_chunk_ids) == 0:
+            return get_result(message=f"deleted {chunk_number} chunks")
+        return get_error_data_result(message=f"rm_chunk deleted chunks {chunk_number}, expect {len(unique_chunk_ids)}")
+    if duplicate_messages:
+        return get_result(message=f"Partially deleted {chunk_number} chunks with {len(duplicate_messages)} errors", data={"success_count": chunk_number, "errors": duplicate_messages},)
     return get_result(message=f"deleted {chunk_number} chunks")
 
 
@@ -1194,7 +1262,7 @@ def update_chunk(tenant_id, dataset_id, document_id, chunk_id):
     if "questions" in req:
         if not isinstance(req["questions"], list):
             return get_error_data_result("`questions` should be a list")
-        d["question_kwd"] = req.get("questions")
+        d["question_kwd"] = [str(q).strip() for q in req.get("questions", []) if str(q).strip()]
         d["question_tks"] = rag_tokenizer.tokenize("\n".join(req["questions"]))
     if "available" in req:
         d["available_int"] = int(req["available"])

api/apps/sdk/session.py

@@ -13,31 +13,30 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-import re
 import json
+import re
 import time
 
-from api.db import LLMType
+import tiktoken
+from flask import Response, jsonify, request
 from api.db.services.conversation_service import ConversationService, iframe_completion
 from api.db.services.conversation_service import completion as rag_completion
-from api.db.services.canvas_service import completion as agent_completion
-from api.db.services.dialog_service import ask, chat
+from api.db.services.canvas_service import completion as agent_completion, completionOpenAI
 from agent.canvas import Canvas
-from api.db import StatusEnum
+from api.db import LLMType, StatusEnum
 from api.db.db_models import APIToken
 from api.db.services.api_service import API4ConversationService
 from api.db.services.canvas_service import UserCanvasService
-from api.db.services.dialog_service import DialogService
+from api.db.services.dialog_service import DialogService, ask, chat
+from api.db.services.file_service import FileService
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.utils import get_uuid
-from api.utils.api_utils import get_error_data_result, validate_request
-from api.utils.api_utils import get_result, token_required
+from api.utils.api_utils import get_result, token_required, get_data_openai, get_error_data_result, validate_request, check_duplicate_ids
 from api.db.services.llm_service import LLMBundle
-from api.db.services.file_service import FileService
 
-from flask import jsonify, request, Response
 
-@manager.route('/chats/<chat_id>/sessions', methods=['POST'])  # noqa: F821
+
+@manager.route("/chats/<chat_id>/sessions", methods=["POST"])  # noqa: F821
 @token_required
 def create(tenant_id, chat_id):
     req = request.json
@@ -50,7 +49,7 @@ def create(tenant_id, chat_id):
         "dialog_id": req["dialog_id"],
         "name": req.get("name", "New session"),
         "message": [{"role": "assistant", "content": dia[0].prompt_config.get("prologue")}],
-        "user_id": req.get("user_id", "")
+        "user_id": req.get("user_id", ""),
     }
     if not conv.get("name"):
         return get_error_data_result(message="`name` can not be empty.")
@@ -59,28 +58,25 @@ def create(tenant_id, chat_id):
     if not e:
         return get_error_data_result(message="Fail to create a session!")
     conv = conv.to_dict()
-    conv['messages'] = conv.pop("message")
+    conv["messages"] = conv.pop("message")
     conv["chat_id"] = conv.pop("dialog_id")
     del conv["reference"]
     return get_result(data=conv)
 
 
-@manager.route('/agents/<agent_id>/sessions', methods=['POST'])  # noqa: F821
+@manager.route("/agents/<agent_id>/sessions", methods=["POST"])  # noqa: F821
 @token_required
 def create_agent_session(tenant_id, agent_id):
     req = request.json
     if not request.is_json:
         req = request.form
     files = request.files
-    user_id = request.args.get('user_id', '')
-
+    user_id = request.args.get("user_id", "")
     e, cvs = UserCanvasService.get_by_id(agent_id)
     if not e:
         return get_error_data_result("Agent not found.")
-
     if not UserCanvasService.query(user_id=tenant_id, id=agent_id):
         return get_error_data_result("You cannot access the agent.")
-
     if not isinstance(cvs.dsl, str):
         cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
 
@@ -113,28 +109,22 @@ def create_agent_session(tenant_id, agent_id):
                             ele.pop("value")
                 else:
                     if req is not None and req.get(ele["key"]):
-                        ele["value"] = req[ele['key']]
+                        ele["value"] = req[ele["key"]]
                     else:
                         if "value" in ele:
                             ele.pop("value")
-    else:
-        for ans in canvas.run(stream=False):
-            pass
+
+    for ans in canvas.run(stream=False):
+        pass
+
     cvs.dsl = json.loads(str(canvas))
-    conv = {
-        "id": get_uuid(),
-        "dialog_id": cvs.id,
-        "user_id": user_id,
-        "message": [{"role": "assistant", "content": canvas.get_prologue()}],
-        "source": "agent",
-        "dsl": cvs.dsl
-    }
+    conv = {"id": get_uuid(), "dialog_id": cvs.id, "user_id": user_id, "message": [{"role": "assistant", "content": canvas.get_prologue()}], "source": "agent", "dsl": cvs.dsl}
     API4ConversationService.save(**conv)
     conv["agent_id"] = conv.pop("dialog_id")
     return get_result(data=conv)
 
 
-@manager.route('/chats/<chat_id>/sessions/<session_id>', methods=['PUT'])  # noqa: F821
+@manager.route("/chats/<chat_id>/sessions/<session_id>", methods=["PUT"])  # noqa: F821
 @token_required
 def update(tenant_id, chat_id, session_id):
     req = request.json
@@ -156,14 +146,14 @@ def update(tenant_id, chat_id, session_id):
     return get_result()
 
 
-@manager.route('/chats/<chat_id>/completions', methods=['POST'])  # noqa: F821
+@manager.route("/chats/<chat_id>/completions", methods=["POST"])  # noqa: F821
 @token_required
 def chat_completion(tenant_id, chat_id):
     req = request.json
     if not req:
         req = {"question": ""}
     if not req.get("session_id"):
-        req["question"]=""
+        req["question"] = ""
     if not DialogService.query(tenant_id=tenant_id, id=chat_id, status=StatusEnum.VALID.value):
         return get_error_data_result(f"You don't own the chat {chat_id}")
     if req.get("session_id"):
@@ -185,7 +175,7 @@ def chat_completion(tenant_id, chat_id):
         return get_result(data=answer)
 
 
-@manager.route('chats_openai/<chat_id>/chat/completions', methods=['POST'])  # noqa: F821
+@manager.route("/chats_openai/<chat_id>/chat/completions", methods=["POST"])  # noqa: F821
 @validate_request("model", "messages")  # noqa: F821
 @token_required
 def chat_completion_openai_like(tenant_id, chat_id):
@@ -250,8 +240,18 @@ def chat_completion_openai_like(tenant_id, chat_id):
     dia = dia[0]
 
     # Filter system and non-sense assistant messages
-    msg = None
-    msg = [m for m in messages if m["role"] != "system" and (m["role"] != "assistant" or msg)]
+    msg = []
+    for m in messages:
+        if m["role"] == "system":
+            continue
+        if m["role"] == "assistant" and not msg:
+            continue
+        msg.append(m)
+
+    # tools = get_tools()
+    # toolcall_session = SimpleFunctionCallServer()
+    tools = None
+    toolcall_session = None
 
     if req.get("stream", True):
         # The value for the usage field on all chunks except for the last one will be null.
@@ -259,40 +259,61 @@ def chat_completion_openai_like(tenant_id, chat_id):
         # The choices field on the last chunk will always be an empty array [].
         def streamed_response_generator(chat_id, dia, msg):
             token_used = 0
-            should_split_index = 0
+            answer_cache = ""
+            reasoning_cache = ""
             response = {
                 "id": f"chatcmpl-{chat_id}",
-                "choices": [
-                    {
-                        "delta": {
-                            "content": "",
-                            "role": "assistant",
-                            "function_call": None,
-                            "tool_calls": None
-                        },
-                        "finish_reason": None,
-                        "index": 0,
-                        "logprobs": None
-                    }
-                ],
+                "choices": [{"delta": {"content": "", "role": "assistant", "function_call": None, "tool_calls": None, "reasoning_content": ""}, "finish_reason": None, "index": 0, "logprobs": None}],
                 "created": int(time.time()),
                 "model": "model",
                 "object": "chat.completion.chunk",
                 "system_fingerprint": "",
-                "usage": None
+                "usage": None,
             }
 
             try:
-                for ans in chat(dia, msg, True):
+                for ans in chat(dia, msg, True, toolcall_session=toolcall_session, tools=tools):
                     answer = ans["answer"]
-                    incremental = answer[should_split_index:]
-                    token_used += len(incremental)
-                    if incremental.endswith("</think>"):
-                        response_data_len = len(incremental.rstrip("</think>"))
+
+                    reasoning_match = re.search(r"<think>(.*?)</think>", answer, flags=re.DOTALL)
+                    if reasoning_match:
+                        reasoning_part = reasoning_match.group(1)
+                        content_part = answer[reasoning_match.end() :]
                     else:
-                        response_data_len = len(incremental)
-                    should_split_index += response_data_len
-                    response["choices"][0]["delta"]["content"] = incremental
+                        reasoning_part = ""
+                        content_part = answer
+
+                    reasoning_incremental = ""
+                    if reasoning_part:
+                        if reasoning_part.startswith(reasoning_cache):
+                            reasoning_incremental = reasoning_part.replace(reasoning_cache, "", 1)
+                        else:
+                            reasoning_incremental = reasoning_part
+                        reasoning_cache = reasoning_part
+
+                    content_incremental = ""
+                    if content_part:
+                        if content_part.startswith(answer_cache):
+                            content_incremental = content_part.replace(answer_cache, "", 1)
+                    else:
+                        content_incremental = content_part
+                    answer_cache = content_part
+
+                    token_used += len(reasoning_incremental) + len(content_incremental)
+
+                    if not any([reasoning_incremental, content_incremental]):
+                        continue
+
+                    if reasoning_incremental:
+                        response["choices"][0]["delta"]["reasoning_content"] = reasoning_incremental
+                    else:
+                        response["choices"][0]["delta"]["reasoning_content"] = None
+
+                    if content_incremental:
+                        response["choices"][0]["delta"]["content"] = content_incremental
+                    else:
+                        response["choices"][0]["delta"]["content"] = None
+
                     yield f"data:{json.dumps(response, ensure_ascii=False)}\n\n"
             except Exception as e:
                 response["choices"][0]["delta"]["content"] = "**ERROR**: " + str(e)
@@ -300,16 +321,12 @@ def chat_completion_openai_like(tenant_id, chat_id):
 
             # The last chunk
             response["choices"][0]["delta"]["content"] = None
+            response["choices"][0]["delta"]["reasoning_content"] = None
             response["choices"][0]["finish_reason"] = "stop"
-            response["usage"] = {
-                "prompt_tokens": len(prompt),
-                "completion_tokens": token_used,
-                "total_tokens": len(prompt) + token_used
-            }
+            response["usage"] = {"prompt_tokens": len(prompt), "completion_tokens": token_used, "total_tokens": len(prompt) + token_used}
             yield f"data:{json.dumps(response, ensure_ascii=False)}\n\n"
             yield "data:[DONE]\n\n"
 
-
         resp = Response(streamed_response_generator(chat_id, dia, msg), mimetype="text/event-stream")
         resp.headers.add_header("Cache-control", "no-cache")
         resp.headers.add_header("Connection", "keep-alive")
@@ -318,13 +335,13 @@ def chat_completion_openai_like(tenant_id, chat_id):
         return resp
     else:
         answer = None
-        for ans in chat(dia, msg, False):
+        for ans in chat(dia, msg, False, toolcall_session=toolcall_session, tools=tools):
             # focus answer content only
             answer = ans
             break
         content = answer["answer"]
 
-        response  = {
+        response = {
             "id": f"chatcmpl-{chat_id}",
             "object": "chat.completion",
             "created": int(time.time()),
@@ -336,25 +353,49 @@ def chat_completion_openai_like(tenant_id, chat_id):
                 "completion_tokens_details": {
                     "reasoning_tokens": context_token_used,
                     "accepted_prediction_tokens": len(content),
-                    "rejected_prediction_tokens": 0 # 0 for simplicity
-                }
+                    "rejected_prediction_tokens": 0,  # 0 for simplicity
+                },
             },
-            "choices": [
-                {
-                    "message": {
-                        "role": "assistant",
-                        "content": content
-                    },
-                    "logprobs": None,
-                    "finish_reason": "stop",
-                    "index": 0
-                }
-            ]
+            "choices": [{"message": {"role": "assistant", "content": content}, "logprobs": None, "finish_reason": "stop", "index": 0}],
         }
         return jsonify(response)
 
+@manager.route('/agents_openai/<agent_id>/chat/completions', methods=['POST'])  # noqa: F821
+@validate_request("model", "messages")  # noqa: F821
+@token_required
+def agents_completion_openai_compatibility (tenant_id, agent_id):
+    req = request.json
+    tiktokenenc = tiktoken.get_encoding("cl100k_base")
+    messages = req.get("messages", [])
+    if not messages:
+        return get_error_data_result("You must provide at least one message.")
+    if not UserCanvasService.query(user_id=tenant_id, id=agent_id):
+        return get_error_data_result(f"You don't own the agent {agent_id}")
   
-@manager.route('/agents/<agent_id>/completions', methods=['POST'])  # noqa: F821
+    filtered_messages = [m for m in messages if m["role"] in ["user", "assistant"]]
+    prompt_tokens = sum(len(tiktokenenc.encode(m["content"])) for m in filtered_messages)
+    if not filtered_messages:
+        return jsonify(get_data_openai( 
+            id=agent_id,
+            content="No valid messages found (user or assistant).",
+            finish_reason="stop",
+            model=req.get("model", ""), 
+            completion_tokens=len(tiktokenenc.encode("No valid messages found (user or assistant).")),
+            prompt_tokens=prompt_tokens, 
+        ))
+    
+    # Get the last user message as the question
+    question = next((m["content"] for m in reversed(messages) if m["role"] == "user"), "")
+    
+    if req.get("stream", True):
+        return Response(completionOpenAI(tenant_id, agent_id, question, session_id=req.get("id", ""), stream=True), mimetype="text/event-stream")
+    else:
+        # For non-streaming, just return the response directly
+        response = next(completionOpenAI(tenant_id, agent_id, question, session_id=req.get("id", ""), stream=False))
+        return jsonify(response)
+    
+
+@manager.route("/agents/<agent_id>/completions", methods=["POST"])  # noqa: F821
 @token_required
 def agent_completions(tenant_id, agent_id):
     req = request.json
@@ -365,9 +406,7 @@ def agent_completions(tenant_id, agent_id):
         dsl = cvs[0].dsl
         if not isinstance(dsl, str):
             dsl = json.dumps(dsl)
-        #canvas = Canvas(dsl, tenant_id)
-        #if canvas.get_preset_param():
-        #    req["question"] = ""
+
         conv = API4ConversationService.query(id=req["session_id"], dialog_id=agent_id)
         if not conv:
             return get_error_data_result(f"You don't own the session {req['session_id']}")
@@ -380,9 +419,7 @@ def agent_completions(tenant_id, agent_id):
             states = {field: current_dsl.get(field, []) for field in state_fields}
             current_dsl.update(new_dsl)
             current_dsl.update(states)
-            API4ConversationService.update_by_id(req["session_id"], {
-                "dsl": current_dsl
-            })
+            API4ConversationService.update_by_id(req["session_id"], {"dsl": current_dsl})
     else:
         req["question"] = ""
     if req.get("stream", True):
@@ -399,7 +436,7 @@ def agent_completions(tenant_id, agent_id):
         return get_error_data_result(str(e))
 
 
-@manager.route('/chats/<chat_id>/sessions', methods=['GET'])  # noqa: F821
+@manager.route("/chats/<chat_id>/sessions", methods=["GET"])  # noqa: F821
 @token_required
 def list_session(tenant_id, chat_id):
     if not DialogService.query(tenant_id=tenant_id, id=chat_id, status=StatusEnum.VALID.value):
@@ -418,7 +455,7 @@ def list_session(tenant_id, chat_id):
     if not convs:
         return get_result(data=[])
     for conv in convs:
-        conv['messages'] = conv.pop("message")
+        conv["messages"] = conv.pop("message")
         infos = conv["messages"]
         for info in infos:
             if "prompt" in info:
@@ -452,7 +489,7 @@ def list_session(tenant_id, chat_id):
     return get_result(data=convs)
 
 
-@manager.route('/agents/<agent_id>/sessions', methods=['GET'])  # noqa: F821
+@manager.route("/agents/<agent_id>/sessions", methods=["GET"])  # noqa: F821
 @token_required
 def list_agent_session(tenant_id, agent_id):
     if not UserCanvasService.query(user_id=tenant_id, id=agent_id):
@@ -468,12 +505,11 @@ def list_agent_session(tenant_id, agent_id):
         desc = True
     # dsl defaults to True in all cases except for False and false
     include_dsl = request.args.get("dsl") != "False" and request.args.get("dsl") != "false"
-    convs = API4ConversationService.get_list(agent_id, tenant_id, page_number, items_per_page, orderby, desc, id,
-                                             user_id, include_dsl)
+    convs = API4ConversationService.get_list(agent_id, tenant_id, page_number, items_per_page, orderby, desc, id, user_id, include_dsl)
     if not convs:
         return get_result(data=[])
     for conv in convs:
-        conv['messages'] = conv.pop("message")
+        conv["messages"] = conv.pop("message")
         infos = conv["messages"]
         for info in infos:
             if "prompt" in info:
@@ -506,11 +542,14 @@ def list_agent_session(tenant_id, agent_id):
     return get_result(data=convs)
 
 
-@manager.route('/chats/<chat_id>/sessions', methods=["DELETE"])  # noqa: F821
+@manager.route("/chats/<chat_id>/sessions", methods=["DELETE"])  # noqa: F821
 @token_required
 def delete(tenant_id, chat_id):
     if not DialogService.query(id=chat_id, tenant_id=tenant_id, status=StatusEnum.VALID.value):
         return get_error_data_result(message="You don't own the chat")
+    
+    errors = []
+    success_count = 0
     req = request.json
     convs = ConversationService.query(dialog_id=chat_id)
     if not req:
@@ -524,17 +563,44 @@ def delete(tenant_id, chat_id):
             conv_list.append(conv.id)
     else:
         conv_list = ids
+    
+    unique_conv_ids, duplicate_messages = check_duplicate_ids(conv_list, "session")
+    conv_list = unique_conv_ids
+    
     for id in conv_list:
         conv = ConversationService.query(id=id, dialog_id=chat_id)
         if not conv:
-            return get_error_data_result(message="The chat doesn't own the session")
+            errors.append(f"The chat doesn't own the session {id}")
+            continue
         ConversationService.delete_by_id(id)
+        success_count += 1
+    
+    if errors:
+        if success_count > 0:
+            return get_result(
+                data={"success_count": success_count, "errors": errors},
+                message=f"Partially deleted {success_count} sessions with {len(errors)} errors"
+            )
+        else:
+            return get_error_data_result(message="; ".join(errors))
+    
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(
+                message=f"Partially deleted {success_count} sessions with {len(duplicate_messages)} errors", 
+                data={"success_count": success_count, "errors": duplicate_messages}
+            )
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
+    
     return get_result()
 
 
-@manager.route('/agents/<agent_id>/sessions', methods=["DELETE"])  # noqa: F821
+@manager.route("/agents/<agent_id>/sessions", methods=["DELETE"])  # noqa: F821
 @token_required
 def delete_agent_session(tenant_id, agent_id):
+    errors = []
+    success_count = 0
     req = request.json
     cvs = UserCanvasService.query(user_id=tenant_id, id=agent_id)
     if not cvs:
@@ -556,15 +622,39 @@ def delete_agent_session(tenant_id, agent_id):
     else:
         conv_list = ids
 
+    unique_conv_ids, duplicate_messages = check_duplicate_ids(conv_list, "session")
+    conv_list = unique_conv_ids
+
     for session_id in conv_list:
         conv = API4ConversationService.query(id=session_id, dialog_id=agent_id)
         if not conv:
-            return get_error_data_result(f"The agent doesn't own the session ${session_id}")
+            errors.append(f"The agent doesn't own the session {session_id}")
+            continue
         API4ConversationService.delete_by_id(session_id)
+        success_count += 1
+    
+    if errors:
+        if success_count > 0:
+            return get_result(
+                data={"success_count": success_count, "errors": errors},
+                message=f"Partially deleted {success_count} sessions with {len(errors)} errors"
+            )
+        else:
+            return get_error_data_result(message="; ".join(errors))
+    
+    if duplicate_messages:
+        if success_count > 0:
+            return get_result(
+                message=f"Partially deleted {success_count} sessions with {len(duplicate_messages)} errors", 
+                data={"success_count": success_count, "errors": duplicate_messages}
+            )
+        else:
+            return get_error_data_result(message=";".join(duplicate_messages))
+    
     return get_result()
 
 
-@manager.route('/sessions/ask', methods=['POST'])  # noqa: F821
+@manager.route("/sessions/ask", methods=["POST"])  # noqa: F821
 @token_required
 def ask_about(tenant_id):
     req = request.json
@@ -590,9 +680,7 @@ def ask_about(tenant_id):
             for ans in ask(req["question"], req["kb_ids"], uid):
                 yield "data:" + json.dumps({"code": 0, "message": "", "data": ans}, ensure_ascii=False) + "\n\n"
         except Exception as e:
-            yield "data:" + json.dumps({"code": 500, "message": str(e),
-                                        "data": {"answer": "**ERROR**: " + str(e), "reference": []}},
-                                       ensure_ascii=False) + "\n\n"
+            yield "data:" + json.dumps({"code": 500, "message": str(e), "data": {"answer": "**ERROR**: " + str(e), "reference": []}}, ensure_ascii=False) + "\n\n"
         yield "data:" + json.dumps({"code": 0, "message": "", "data": True}, ensure_ascii=False) + "\n\n"
 
     resp = Response(stream(), mimetype="text/event-stream")
@@ -603,7 +691,7 @@ def ask_about(tenant_id):
     return resp
 
 
-@manager.route('/sessions/related_questions', methods=['POST'])  # noqa: F821
+@manager.route("/sessions/related_questions", methods=["POST"])  # noqa: F821
 @token_required
 def related_questions(tenant_id):
     req = request.json
@@ -635,18 +723,27 @@ Reason:
  - At the same time, related terms can also help search engines better understand user needs and return more accurate search results.
 
 """
-    ans = chat_mdl.chat(prompt, [{"role": "user", "content": f"""
+    ans = chat_mdl.chat(
+        prompt,
+        [
+            {
+                "role": "user",
+                "content": f"""
 Keywords: {question}
 Related search terms:
-    """}], {"temperature": 0.9})
+    """,
+            }
+        ],
+        {"temperature": 0.9},
+    )
     return get_result(data=[re.sub(r"^[0-9]\. ", "", a) for a in ans.split("\n") if re.match(r"^[0-9]\. ", a)])
 
 
-@manager.route('/chatbots/<dialog_id>/completions', methods=['POST'])  # noqa: F821
+@manager.route("/chatbots/<dialog_id>/completions", methods=["POST"])  # noqa: F821
 def chatbot_completions(dialog_id):
     req = request.json
 
-    token = request.headers.get('Authorization').split()
+    token = request.headers.get("Authorization").split()
     if len(token) != 2:
         return get_error_data_result(message='Authorization is not valid!"')
     token = token[1]
@@ -669,11 +766,11 @@ def chatbot_completions(dialog_id):
         return get_result(data=answer)
 
 
-@manager.route('/agentbots/<agent_id>/completions', methods=['POST'])  # noqa: F821
+@manager.route("/agentbots/<agent_id>/completions", methods=["POST"])  # noqa: F821
 def agent_bot_completions(agent_id):
     req = request.json
 
-    token = request.headers.get('Authorization').split()
+    token = request.headers.get("Authorization").split()
     if len(token) != 2:
         return get_error_data_result(message='Authorization is not valid!"')
     token = token[1]

api/apps/system_app.py

@@ -37,7 +37,6 @@ from timeit import default_timer as timer
 
 from rag.utils.redis_conn import REDIS_CONN
 
-
 @manager.route("/version", methods=["GET"])  # noqa: F821
 @login_required
 def version():
@@ -298,3 +297,25 @@ def rm(token):
         [APIToken.tenant_id == current_user.id, APIToken.token == token]
     )
     return get_json_result(data=True)
+
+
+@manager.route('/config', methods=['GET'])  # noqa: F821
+def get_config():
+    """
+    Get system configuration.
+    ---
+    tags:
+        - System
+    responses:
+        200:
+            description: Return system configuration
+            schema:
+                type: object
+                properties:
+                    registerEnable:
+                        type: integer 0 means disabled, 1 means enabled
+                        description: Whether user registration is enabled
+    """
+    return get_json_result(data={
+        "registerEnabled": settings.REGISTER_ENABLED
+    })

api/apps/user_app.py

@@ -562,6 +562,14 @@ def user_add():
         schema:
           type: object
     """
+
+    if not settings.REGISTER_ENABLED:
+        return get_json_result(
+            data=False,
+            message="User registration is disabled!",
+            code=settings.RetCode.OPERATING_ERROR,
+        )
+
     req = request.json
     email_address = req["email"]
 

api/db/init_data.py

@@ -103,16 +103,12 @@ def init_llm_factory():
     except Exception:
         pass
 
-    factory_llm_infos = json.load(
-        open(
-            os.path.join(get_project_base_directory(), "conf", "llm_factories.json"),
-            "r",
-        )
-    )
-    for factory_llm_info in factory_llm_infos["factory_llm_infos"]:
-        llm_infos = factory_llm_info.pop("llm")
+    factory_llm_infos = settings.FACTORY_LLM_INFOS    
+    for factory_llm_info in factory_llm_infos:
+        info = deepcopy(factory_llm_info)
+        llm_infos = info.pop("llm")
         try:
-            LLMFactoriesService.save(**factory_llm_info)
+            LLMFactoriesService.save(**info)
         except Exception:
             pass
         LLMService.filter_delete([LLM.fid == factory_llm_info["name"]])
@@ -152,7 +148,7 @@ def init_llm_factory():
                 pass
             break
     for kb_id in KnowledgebaseService.get_all_ids():
-        KnowledgebaseService.update_by_id(kb_id, {"doc_num": DocumentService.get_kb_doc_count(kb_id)})
+        KnowledgebaseService.update_document_number_in_init(kb_id=kb_id, doc_num=DocumentService.get_kb_doc_count(kb_id))
 
 
 

api/db/services/canvas_service.py

@@ -18,13 +18,15 @@ import time
 import traceback
 from uuid import uuid4
 from agent.canvas import Canvas
-from api.db.db_models import DB, CanvasTemplate, UserCanvas, API4Conversation
+from api.db import TenantPermission
+from api.db.db_models import DB, CanvasTemplate, User, UserCanvas, API4Conversation
 from api.db.services.api_service import API4ConversationService
 from api.db.services.common_service import CommonService
 from api.db.services.conversation_service import structure_answer
 from api.utils import get_uuid
-
-
+from api.utils.api_utils import get_data_openai
+import tiktoken
+from peewee import fn
 class CanvasTemplateService(CommonService):
     model = CanvasTemplate
 
@@ -51,6 +53,73 @@ class UserCanvasService(CommonService):
 
         return list(agents.dicts())
    
+    @classmethod
+    @DB.connection_context()
+    def get_by_tenant_id(cls, pid):
+        try:
+            
+            fields = [
+                cls.model.id,
+                cls.model.avatar,
+                cls.model.title,
+                cls.model.dsl,
+                cls.model.description,
+                cls.model.permission,
+                cls.model.update_time,
+                cls.model.user_id,
+                cls.model.create_time,
+                cls.model.create_date,
+                cls.model.update_date,
+                User.nickname,
+                User.avatar.alias('tenant_avatar'),
+            ]
+            angents = cls.model.select(*fields) \
+            .join(User, on=(cls.model.user_id == User.id)) \
+            .where(cls.model.id == pid)
+            # obj = cls.model.query(id=pid)[0]
+            return True, angents.dicts()[0]
+        except Exception as e:
+            print(e)
+            return False, None
+     
+    @classmethod
+    @DB.connection_context()
+    def get_by_tenant_ids(cls, joined_tenant_ids, user_id,
+                          page_number, items_per_page,
+                          orderby, desc, keywords,
+                          ):
+        fields = [
+            cls.model.id,
+            cls.model.avatar,
+            cls.model.title,
+            cls.model.dsl,
+            cls.model.description,
+            cls.model.permission,
+            User.nickname,
+            User.avatar.alias('tenant_avatar'),
+            cls.model.update_time
+        ]
+        if keywords:
+            angents = cls.model.select(*fields).join(User, on=(cls.model.user_id == User.id)).where(
+                ((cls.model.user_id.in_(joined_tenant_ids) & (cls.model.permission == 
+                                                                TenantPermission.TEAM.value)) | (
+                    cls.model.user_id == user_id)),
+                (fn.LOWER(cls.model.title).contains(keywords.lower()))
+            )
+        else:
+            angents = cls.model.select(*fields).join(User, on=(cls.model.user_id == User.id)).where(
+                ((cls.model.user_id.in_(joined_tenant_ids) & (cls.model.permission == 
+                                                                TenantPermission.TEAM.value)) | (
+                    cls.model.user_id == user_id))
+            )
+        if desc:
+            angents = angents.order_by(cls.model.getter_by(orderby).desc())
+        else:
+            angents = angents.order_by(cls.model.getter_by(orderby).asc())
+        count = angents.count()
+        angents = angents.paginate(page_number, items_per_page)
+        return list(angents.dicts()), count
+   
 
 def completion(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
     e, cvs = UserCanvasService.get_by_id(agent_id)
@@ -86,8 +155,6 @@ def completion(tenant_id, agent_id, question, session_id=None, stream=True, **kw
             "dsl": cvs.dsl
         }
         API4ConversationService.save(**conv)
-
-        
         conv = API4Conversation(**conv)
     else:
         e, conv = API4ConversationService.get_by_id(session_id)
@@ -153,3 +220,206 @@ def completion(tenant_id, agent_id, question, session_id=None, stream=True, **kw
             API4ConversationService.append_message(conv.id, conv.to_dict())
             yield result
             break
+def completionOpenAI(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
+    """Main function for OpenAI-compatible completions, structured similarly to the completion function."""
+    tiktokenenc = tiktoken.get_encoding("cl100k_base")
+    e, cvs = UserCanvasService.get_by_id(agent_id)
+    
+    if not e:
+        yield get_data_openai(
+            id=session_id,
+            model=agent_id,
+            content="**ERROR**: Agent not found."
+        )
+        return
+    
+    if cvs.user_id != tenant_id:
+        yield get_data_openai(
+            id=session_id,
+            model=agent_id,
+            content="**ERROR**: You do not own the agent"
+        )
+        return
+    
+    if not isinstance(cvs.dsl, str):
+        cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
+    
+    canvas = Canvas(cvs.dsl, tenant_id)
+    canvas.reset()
+    message_id = str(uuid4())
+    
+    # Handle new session creation
+    if not session_id:
+        query = canvas.get_preset_param()
+        if query:
+            for ele in query:
+                if not ele["optional"]:
+                    if not kwargs.get(ele["key"]):
+                        yield get_data_openai(
+                            id=None,
+                            model=agent_id,
+                            content=f"`{ele['key']}` is required",
+                            completion_tokens=len(tiktokenenc.encode(f"`{ele['key']}` is required")),
+                            prompt_tokens=len(tiktokenenc.encode(question if question else ""))
+                        )
+                        return
+                    ele["value"] = kwargs[ele["key"]]
+                if ele["optional"]:
+                    if kwargs.get(ele["key"]):
+                        ele["value"] = kwargs[ele['key']]
+                    else:
+                        if "value" in ele:
+                            ele.pop("value")
+        
+        cvs.dsl = json.loads(str(canvas))
+        session_id = get_uuid()
+        conv = {
+            "id": session_id,
+            "dialog_id": cvs.id,
+            "user_id": kwargs.get("user_id", "") if isinstance(kwargs, dict) else "",
+            "message": [{"role": "assistant", "content": canvas.get_prologue(), "created_at": time.time()}],
+            "source": "agent",
+            "dsl": cvs.dsl
+        }
+        API4ConversationService.save(**conv)
+        conv = API4Conversation(**conv)
+            
+    # Handle existing session
+    else:
+        e, conv = API4ConversationService.get_by_id(session_id)
+        if not e:
+            yield get_data_openai(
+                id=session_id,
+                model=agent_id,
+                content="**ERROR**: Session not found!"
+            )
+            return
+        
+        canvas = Canvas(json.dumps(conv.dsl), tenant_id)
+        canvas.messages.append({"role": "user", "content": question, "id": message_id})
+        canvas.add_user_input(question)
+        
+        if not conv.message:
+            conv.message = []
+        conv.message.append({
+            "role": "user",
+            "content": question,
+            "id": message_id
+        })
+        
+        if not conv.reference:
+            conv.reference = []
+        conv.reference.append({"chunks": [], "doc_aggs": []})
+    
+    # Process request based on stream mode
+    final_ans = {"reference": [], "content": ""}
+    prompt_tokens = len(tiktokenenc.encode(str(question)))
+    
+    if stream:
+        try:
+            completion_tokens = 0
+            for ans in canvas.run(stream=True):
+                if ans.get("running_status"):
+                    completion_tokens += len(tiktokenenc.encode(ans.get("content", "")))
+                    yield "data: " + json.dumps(
+                        get_data_openai(
+                            id=session_id,
+                            model=agent_id,
+                            content=ans["content"],
+                            object="chat.completion.chunk",
+                            completion_tokens=completion_tokens,
+                            prompt_tokens=prompt_tokens
+                        ),
+                        ensure_ascii=False
+                    ) + "\n\n"
+                    continue
+                
+                for k in ans.keys():
+                    final_ans[k] = ans[k]
+                
+                completion_tokens += len(tiktokenenc.encode(final_ans.get("content", "")))
+                yield "data: " + json.dumps(
+                    get_data_openai(
+                        id=session_id,
+                        model=agent_id,
+                        content=final_ans["content"],
+                        object="chat.completion.chunk",
+                        finish_reason="stop",
+                        completion_tokens=completion_tokens,
+                        prompt_tokens=prompt_tokens
+                    ),
+                    ensure_ascii=False
+                ) + "\n\n"
+            
+            # Update conversation
+            canvas.messages.append({"role": "assistant", "content": final_ans["content"], "created_at": time.time(), "id": message_id})
+            canvas.history.append(("assistant", final_ans["content"]))
+            if final_ans.get("reference"):
+                canvas.reference.append(final_ans["reference"])
+            conv.dsl = json.loads(str(canvas))
+            API4ConversationService.append_message(conv.id, conv.to_dict())
+            
+            yield "data: [DONE]\n\n"
+            
+        except Exception as e:
+            traceback.print_exc()
+            conv.dsl = json.loads(str(canvas))
+            API4ConversationService.append_message(conv.id, conv.to_dict())
+            yield "data: " + json.dumps(
+                get_data_openai(
+                    id=session_id,
+                    model=agent_id,
+                    content="**ERROR**: " + str(e),
+                    finish_reason="stop",
+                    completion_tokens=len(tiktokenenc.encode("**ERROR**: " + str(e))),
+                    prompt_tokens=prompt_tokens
+                ),
+                ensure_ascii=False
+            ) + "\n\n"
+            yield "data: [DONE]\n\n"
+    
+    else:  # Non-streaming mode
+        try:
+            all_answer_content = ""
+            for answer in canvas.run(stream=False):
+                if answer.get("running_status"):
+                    continue
+                
+                final_ans["content"] = "\n".join(answer["content"]) if "content" in answer else ""
+                final_ans["reference"] = answer.get("reference", [])
+                all_answer_content += final_ans["content"]
+            
+            final_ans["content"] = all_answer_content
+            
+            # Update conversation
+            canvas.messages.append({"role": "assistant", "content": final_ans["content"], "created_at": time.time(), "id": message_id})
+            canvas.history.append(("assistant", final_ans["content"]))
+            if final_ans.get("reference"):
+                canvas.reference.append(final_ans["reference"])
+            conv.dsl = json.loads(str(canvas))
+            API4ConversationService.append_message(conv.id, conv.to_dict())
+            
+            # Return the response in OpenAI format
+            yield get_data_openai(
+                id=session_id,
+                model=agent_id,
+                content=final_ans["content"],
+                finish_reason="stop",
+                completion_tokens=len(tiktokenenc.encode(final_ans["content"])),
+                prompt_tokens=prompt_tokens,
+                param=canvas.get_preset_param()  # Added param info like in completion
+            )
+            
+        except Exception as e:
+            traceback.print_exc()
+            conv.dsl = json.loads(str(canvas))
+            API4ConversationService.append_message(conv.id, conv.to_dict())
+            yield get_data_openai(
+                id=session_id,
+                model=agent_id,
+                content="**ERROR**: " + str(e),
+                finish_reason="stop",
+                completion_tokens=len(tiktokenenc.encode("**ERROR**: " + str(e))),
+                prompt_tokens=prompt_tokens
+            )
+

api/db/services/common_service.py

@@ -22,17 +22,56 @@ from api.utils import datetime_format, current_timestamp, get_uuid
 
 
 class CommonService:
+    """Base service class that provides common database operations.
+    
+    This class serves as a foundation for all service classes in the application,
+    implementing standard CRUD operations and common database query patterns.
+    It uses the Peewee ORM for database interactions and provides a consistent
+    interface for database operations across all derived service classes.
+    
+    Attributes:
+        model: The Peewee model class that this service operates on. Must be set by subclasses.
+    """
     model = None
 
     @classmethod
     @DB.connection_context()
     def query(cls, cols=None, reverse=None, order_by=None, **kwargs):
+        """Execute a database query with optional column selection and ordering.
+    
+        This method provides a flexible way to query the database with various filters
+        and sorting options. It supports column selection, sort order control, and
+        additional filter conditions.
+    
+        Args:
+            cols (list, optional): List of column names to select. If None, selects all columns.
+            reverse (bool, optional): If True, sorts in descending order. If False, sorts in ascending order.
+            order_by (str, optional): Column name to sort results by.
+            **kwargs: Additional filter conditions passed as keyword arguments.
+    
+        Returns:
+            peewee.ModelSelect: A query result containing matching records.
+        """
         return cls.model.query(cols=cols, reverse=reverse,
                                order_by=order_by, **kwargs)
 
     @classmethod
     @DB.connection_context()
     def get_all(cls, cols=None, reverse=None, order_by=None):
+        """Retrieve all records from the database with optional column selection and ordering.
+    
+        This method fetches all records from the model's table with support for
+        column selection and result ordering. If no order_by is specified and reverse
+        is True, it defaults to ordering by create_time.
+    
+        Args:
+            cols (list, optional): List of column names to select. If None, selects all columns.
+            reverse (bool, optional): If True, sorts in descending order. If False, sorts in ascending order.
+            order_by (str, optional): Column name to sort results by. Defaults to 'create_time' if reverse is specified.
+    
+        Returns:
+            peewee.ModelSelect: A query containing all matching records.
+        """
         if cols:
             query_records = cls.model.select(*cols)
         else:
@@ -51,11 +90,36 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def get(cls, **kwargs):
+        """Get a single record matching the given criteria.
+    
+        This method retrieves a single record from the database that matches
+        the specified filter conditions.
+    
+        Args:
+            **kwargs: Filter conditions as keyword arguments.
+    
+        Returns:
+            Model instance: Single matching record.
+    
+        Raises:
+            peewee.DoesNotExist: If no matching record is found.
+        """
         return cls.model.get(**kwargs)
 
     @classmethod
     @DB.connection_context()
     def get_or_none(cls, **kwargs):
+        """Get a single record or None if not found.
+    
+        This method attempts to retrieve a single record matching the given criteria,
+        returning None if no match is found instead of raising an exception.
+    
+        Args:
+            **kwargs: Filter conditions as keyword arguments.
+    
+        Returns:
+            Model instance or None: Matching record if found, None otherwise.
+        """
         try:
             return cls.model.get(**kwargs)
         except peewee.DoesNotExist:
@@ -64,14 +128,34 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def save(cls, **kwargs):
-        # if "id" not in kwargs:
-        #    kwargs["id"] = get_uuid()
+        """Save a new record to database.
+    
+        This method creates a new record in the database with the provided field values,
+        forcing an insert operation rather than an update.
+    
+        Args:
+            **kwargs: Record field values as keyword arguments.
+    
+        Returns:
+            Model instance: The created record object.
+        """
         sample_obj = cls.model(**kwargs).save(force_insert=True)
         return sample_obj
 
     @classmethod
     @DB.connection_context()
     def insert(cls, **kwargs):
+        """Insert a new record with automatic ID and timestamps.
+    
+        This method creates a new record with automatically generated ID and timestamp fields.
+        It handles the creation of create_time, create_date, update_time, and update_date fields.
+    
+        Args:
+            **kwargs: Record field values as keyword arguments.
+    
+        Returns:
+            Model instance: The newly created record object.
+        """
         if "id" not in kwargs:
             kwargs["id"] = get_uuid()
         kwargs["create_time"] = current_timestamp()
@@ -84,6 +168,15 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def insert_many(cls, data_list, batch_size=100):
+        """Insert multiple records in batches.
+    
+        This method efficiently inserts multiple records into the database using batch processing.
+        It automatically sets creation timestamps for all records.
+    
+        Args:
+            data_list (list): List of dictionaries containing record data to insert.
+            batch_size (int, optional): Number of records to insert in each batch. Defaults to 100.
+        """
         with DB.atomic():
             for d in data_list:
                 d["create_time"] = current_timestamp()
@@ -94,6 +187,15 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def update_many_by_id(cls, data_list):
+        """Update multiple records by their IDs.
+    
+        This method updates multiple records in the database, identified by their IDs.
+        It automatically updates the update_time and update_date fields for each record.
+    
+        Args:
+            data_list (list): List of dictionaries containing record data to update.
+                             Each dictionary must include an 'id' field.
+        """
         with DB.atomic():
             for data in data_list:
                 data["update_time"] = current_timestamp()
@@ -104,6 +206,12 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def update_by_id(cls, pid, data):
+        # Update a single record by ID
+        # Args:
+        #     pid: Record ID
+        #     data: Updated field values
+        # Returns:
+        #     Number of records updated
         data["update_time"] = current_timestamp()
         data["update_date"] = datetime_format(datetime.now())
         num = cls.model.update(data).where(cls.model.id == pid).execute()
@@ -112,15 +220,28 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def get_by_id(cls, pid):
+        # Get a record by ID
+        # Args:
+        #     pid: Record ID
+        # Returns:
+        #     Tuple of (success, record)
         try:
-            obj = cls.model.query(id=pid)[0]
-            return True, obj
+            obj = cls.model.get_or_none(cls.model.id == pid)
+            if obj:
+                return True, obj
         except Exception:
-            return False, None
+            pass
+        return False, None
 
     @classmethod
     @DB.connection_context()
     def get_by_ids(cls, pids, cols=None):
+        # Get multiple records by their IDs
+        # Args:
+        #     pids: List of record IDs
+        #     cols: List of columns to select
+        # Returns:
+        #     Query of matching records
         if cols:
             objs = cls.model.select(*cols)
         else:
@@ -130,11 +251,21 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def delete_by_id(cls, pid):
+        # Delete a record by ID
+        # Args:
+        #     pid: Record ID
+        # Returns:
+        #     Number of records deleted
         return cls.model.delete().where(cls.model.id == pid).execute()
 
     @classmethod
     @DB.connection_context()
     def filter_delete(cls, filters):
+        # Delete records matching given filters
+        # Args:
+        #     filters: List of filter conditions
+        # Returns:
+        #     Number of records deleted
         with DB.atomic():
             num = cls.model.delete().where(*filters).execute()
             return num
@@ -142,11 +273,23 @@ class CommonService:
     @classmethod
     @DB.connection_context()
     def filter_update(cls, filters, update_data):
+        # Update records matching given filters
+        # Args:
+        #     filters: List of filter conditions
+        #     update_data: Updated field values
+        # Returns:
+        #     Number of records updated
         with DB.atomic():
             return cls.model.update(update_data).where(*filters).execute()
 
     @staticmethod
     def cut_list(tar_list, n):
+        # Split a list into chunks of size n
+        # Args:
+        #     tar_list: List to split
+        #     n: Chunk size
+        # Returns:
+        #     List of tuples containing chunks
         length = len(tar_list)
         arr = range(length)
         result = [tuple(tar_list[x:(x + n)]) for x in arr[::n]]
@@ -156,6 +299,14 @@ class CommonService:
     @DB.connection_context()
     def filter_scope_list(cls, in_key, in_filters_list,
                           filters=None, cols=None):
+        # Get records matching IN clause filters with optional column selection
+        # Args:
+        #     in_key: Field name for IN clause
+        #     in_filters_list: List of values for IN clause
+        #     filters: Additional filter conditions
+        #     cols: List of columns to select
+        # Returns:
+        #     List of matching records
         in_filters_tuple_list = cls.cut_list(in_filters_list, 20)
         if not filters:
             filters = []

api/db/services/dialog_service.py

@@ -13,45 +13,79 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-import logging
 import binascii
-import time
-from functools import partial
+from datetime import datetime
+import logging
 import re
+import time
 from copy import deepcopy
+from functools import partial
 from timeit import default_timer as timer
+
+from langfuse import Langfuse
+
 from agentic_reasoning import DeepResearcher
+from api import settings
 from api.db import LLMType, ParserType, StatusEnum
-from api.db.db_models import Dialog, DB
+from api.db.db_models import DB, Dialog
 from api.db.services.common_service import CommonService
 from api.db.services.knowledgebase_service import KnowledgebaseService
-from api.db.services.llm_service import TenantLLMService, LLMBundle
-from api import settings
+from api.db.services.langfuse_service import TenantLangfuseService
+from api.db.services.llm_service import LLMBundle, TenantLLMService
+from api.utils import current_timestamp, datetime_format
 from rag.app.resume import forbidden_select_fields4resume
 from rag.app.tag import label_question
 from rag.nlp.search import index_name
-from rag.prompts import kb_prompt, message_fit_in, llm_id2llm_type, keyword_extraction, full_question, chunks_format, \
-    citation_prompt
-from rag.utils import rmSpace, num_tokens_from_string
+from rag.prompts import chunks_format, citation_prompt, full_question, kb_prompt, keyword_extraction, llm_id2llm_type, message_fit_in
+from rag.utils import num_tokens_from_string, rmSpace
 from rag.utils.tavily_conn import Tavily
 
 
 class DialogService(CommonService):
     model = Dialog
 
+    @classmethod
+    def save(cls, **kwargs):
+        """Save a new record to database.
+
+        This method creates a new record in the database with the provided field values,
+        forcing an insert operation rather than an update.
+
+        Args:
+            **kwargs: Record field values as keyword arguments.
+
+        Returns:
+            Model instance: The created record object.
+        """
+        sample_obj = cls.model(**kwargs).save(force_insert=True)
+        return sample_obj
+
+    @classmethod
+    def update_many_by_id(cls, data_list):
+        """Update multiple records by their IDs.
+
+        This method updates multiple records in the database, identified by their IDs.
+        It automatically updates the update_time and update_date fields for each record.
+
+        Args:
+            data_list (list): List of dictionaries containing record data to update.
+                             Each dictionary must include an 'id' field.
+        """
+        with DB.atomic():
+            for data in data_list:
+                data["update_time"] = current_timestamp()
+                data["update_date"] = datetime_format(datetime.now())
+                cls.model.update(data).where(cls.model.id == data["id"]).execute()
+
     @classmethod
     @DB.connection_context()
-    def get_list(cls, tenant_id,
-                 page_number, items_per_page, orderby, desc, id, name):
+    def get_list(cls, tenant_id, page_number, items_per_page, orderby, desc, id, name):
         chats = cls.model.select()
         if id:
             chats = chats.where(cls.model.id == id)
         if name:
             chats = chats.where(cls.model.name == name)
-        chats = chats.where(
-            (cls.model.tenant_id == tenant_id)
-            & (cls.model.status == StatusEnum.VALID.value)
-        )
+        chats = chats.where((cls.model.tenant_id == tenant_id) & (cls.model.status == StatusEnum.VALID.value))
         if desc:
             chats = chats.order_by(cls.model.getter_by(orderby).desc())
         else:
@@ -72,13 +106,12 @@ def chat_solo(dialog, messages, stream=True):
     tts_mdl = None
     if prompt_config.get("tts"):
         tts_mdl = LLMBundle(dialog.tenant_id, LLMType.TTS)
-    msg = [{"role": m["role"], "content": re.sub(r"##\d+\$\$", "", m["content"])}
-           for m in messages if m["role"] != "system"]
+    msg = [{"role": m["role"], "content": re.sub(r"##\d+\$\$", "", m["content"])} for m in messages if m["role"] != "system"]
     if stream:
         last_ans = ""
         for ans in chat_mdl.chat_streamly(prompt_config.get("system", ""), msg, dialog.llm_setting):
             answer = ans
-            delta_ans = ans[len(last_ans):]
+            delta_ans = ans[len(last_ans) :]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
@@ -110,6 +143,16 @@ def chat(dialog, messages, stream=True, **kwargs):
 
     check_llm_ts = timer()
 
+    langfuse_tracer = None
+    langfuse_keys = TenantLangfuseService.filter_by_tenant(tenant_id=dialog.tenant_id)
+    if langfuse_keys:
+        langfuse = Langfuse(public_key=langfuse_keys.public_key, secret_key=langfuse_keys.secret_key, host=langfuse_keys.host)
+        if langfuse.auth_check():
+            langfuse_tracer = langfuse
+            langfuse.trace = langfuse_tracer.trace(name=f"{dialog.name}-{llm_model_config['llm_name']}")
+
+    check_langfuse_tracer_ts = timer()
+
     kbs = KnowledgebaseService.get_by_ids(dialog.kb_ids)
     embedding_list = list(set([kb.embd_id for kb in kbs]))
     if len(embedding_list) != 1:
@@ -137,6 +180,9 @@ def chat(dialog, messages, stream=True, **kwargs):
         chat_mdl = LLMBundle(dialog.tenant_id, LLMType.IMAGE2TEXT, dialog.llm_id)
     else:
         chat_mdl = LLMBundle(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
+        toolcall_session, tools = kwargs.get("toolcall_session"), kwargs.get("tools")
+        if toolcall_session and tools:
+            chat_mdl.bind_tools(toolcall_session, tools)
 
     bind_llm_ts = timer()
 
@@ -159,8 +205,7 @@ def chat(dialog, messages, stream=True, **kwargs):
         if p["key"] not in kwargs and not p["optional"]:
             raise KeyError("Miss parameter: " + p["key"])
         if p["key"] not in kwargs:
-            prompt_config["system"] = prompt_config["system"].replace(
-                "{%s}" % p["key"], " ")
+            prompt_config["system"] = prompt_config["system"].replace("{%s}" % p["key"], " ")
 
     if len(questions) > 1 and prompt_config.get("refine_multiturn"):
         questions = [full_question(dialog.tenant_id, dialog.llm_id, messages)]
@@ -189,9 +234,11 @@ def chat(dialog, messages, stream=True, **kwargs):
 
         knowledges = []
         if prompt_config.get("reasoning", False):
-            reasoner = DeepResearcher(chat_mdl,
-                                      prompt_config,
-                                      partial(retriever.retrieval, embd_mdl=embd_mdl, tenant_ids=tenant_ids, kb_ids=dialog.kb_ids, page=1, page_size=dialog.top_n, similarity_threshold=0.2, vector_similarity_weight=0.3))
+            reasoner = DeepResearcher(
+                chat_mdl,
+                prompt_config,
+                partial(retriever.retrieval, embd_mdl=embd_mdl, tenant_ids=tenant_ids, kb_ids=dialog.kb_ids, page=1, page_size=dialog.top_n, similarity_threshold=0.2, vector_similarity_weight=0.3),
+            )
 
             for think in reasoner.thinking(kbinfos, " ".join(questions)):
                 if isinstance(think, str):
@@ -200,31 +247,34 @@ def chat(dialog, messages, stream=True, **kwargs):
                 elif stream:
                     yield think
         else:
-            kbinfos = retriever.retrieval(" ".join(questions), embd_mdl, tenant_ids, dialog.kb_ids, 1, dialog.top_n,
-                                          dialog.similarity_threshold,
-                                          dialog.vector_similarity_weight,
-                                          doc_ids=attachments,
-                                          top=dialog.top_k, aggs=False, rerank_mdl=rerank_mdl,
-                                          rank_feature=label_question(" ".join(questions), kbs)
-                                          )
+            kbinfos = retriever.retrieval(
+                " ".join(questions),
+                embd_mdl,
+                tenant_ids,
+                dialog.kb_ids,
+                1,
+                dialog.top_n,
+                dialog.similarity_threshold,
+                dialog.vector_similarity_weight,
+                doc_ids=attachments,
+                top=dialog.top_k,
+                aggs=False,
+                rerank_mdl=rerank_mdl,
+                rank_feature=label_question(" ".join(questions), kbs),
+            )
             if prompt_config.get("tavily_api_key"):
                 tav = Tavily(prompt_config["tavily_api_key"])
                 tav_res = tav.retrieve_chunks(" ".join(questions))
                 kbinfos["chunks"].extend(tav_res["chunks"])
                 kbinfos["doc_aggs"].extend(tav_res["doc_aggs"])
             if prompt_config.get("use_kg"):
-                ck = settings.kg_retrievaler.retrieval(" ".join(questions),
-                                                       tenant_ids,
-                                                       dialog.kb_ids,
-                                                       embd_mdl,
-                                                       LLMBundle(dialog.tenant_id, LLMType.CHAT))
+                ck = settings.kg_retrievaler.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)
 
             knowledges = kb_prompt(kbinfos, max_tokens)
 
-    logging.debug(
-        "{}->{}".format(" ".join(questions), "\n->".join(knowledges)))
+    logging.debug("{}->{}".format(" ".join(questions), "\n->".join(knowledges)))
 
     retrieval_ts = timer()
     if not knowledges and prompt_config.get("empty_response"):
@@ -239,19 +289,16 @@ def chat(dialog, messages, stream=True, **kwargs):
     prompt4citation = ""
     if knowledges and (prompt_config.get("quote", True) and kwargs.get("quote", True)):
         prompt4citation = citation_prompt()
-    msg.extend([{"role": m["role"], "content": re.sub(r"##\d+\$\$", "", m["content"])}
-                for m in messages if m["role"] != "system"])
+    msg.extend([{"role": m["role"], "content": re.sub(r"##\d+\$\$", "", m["content"])} for m in messages if m["role"] != "system"])
     used_token_count, msg = message_fit_in(msg, int(max_tokens * 0.95))
     assert len(msg) >= 2, f"message_fit_in has bug: {msg}"
     prompt = msg[0]["content"]
 
     if "max_tokens" in gen_conf:
-        gen_conf["max_tokens"] = min(
-            gen_conf["max_tokens"],
-            max_tokens - used_token_count)
+        gen_conf["max_tokens"] = min(gen_conf["max_tokens"], max_tokens - used_token_count)
 
     def decorate_answer(answer):
-        nonlocal prompt_config, knowledges, kwargs, kbinfos, prompt, retrieval_ts, questions
+        nonlocal prompt_config, knowledges, kwargs, kbinfos, prompt, retrieval_ts, questions, langfuse_tracer
 
         refs = []
         ans = answer.split("</think>")
@@ -259,27 +306,37 @@ def chat(dialog, messages, stream=True, **kwargs):
         if len(ans) == 2:
             think = ans[0] + "</think>"
             answer = ans[1]
+
         if knowledges and (prompt_config.get("quote", True) and kwargs.get("quote", True)):
             answer = re.sub(r"##[ij]\$\$", "", answer, flags=re.DOTALL)
+            idx = set([])
             if not re.search(r"##[0-9]+\$\$", answer):
-                answer, idx = retriever.insert_citations(answer,
-                                                         [ck["content_ltks"]
-                                                          for ck in kbinfos["chunks"]],
-                                                         [ck["vector"]
-                                                          for ck in kbinfos["chunks"]],
-                                                         embd_mdl,
-                                                         tkweight=1 - dialog.vector_similarity_weight,
-                                                         vtweight=dialog.vector_similarity_weight)
+                answer, idx = retriever.insert_citations(
+                    answer,
+                    [ck["content_ltks"] for ck in kbinfos["chunks"]],
+                    [ck["vector"] for ck in kbinfos["chunks"]],
+                    embd_mdl,
+                    tkweight=1 - dialog.vector_similarity_weight,
+                    vtweight=dialog.vector_similarity_weight,
+                )
             else:
-                idx = set([])
-                for r in re.finditer(r"##([0-9]+)\$\$", answer):
-                    i = int(r.group(1))
+                for match in re.finditer(r"##([0-9]+)\$\$", answer):
+                    i = int(match.group(1))
                     if i < len(kbinfos["chunks"]):
                         idx.add(i)
 
+            # handle (ID: 1), ID: 2 etc.
+            for match in re.finditer(r"\(\s*ID:\s*(\d+)\s*\)|ID[: ]+\s*(\d+)", answer):
+                full_match = match.group(0)
+                id = match.group(1) or match.group(2)
+                if id:
+                    i = int(id)
+                    if i < len(kbinfos["chunks"]):
+                        idx.add(i)
+                    answer = answer.replace(full_match, f"##{i}$$")
+
             idx = set([kbinfos["chunks"][int(i)]["doc_id"] for i in idx])
-            recall_docs = [
-                d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
+            recall_docs = [d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
             if not recall_docs:
                 recall_docs = kbinfos["doc_aggs"]
             kbinfos["doc_aggs"] = recall_docs
@@ -295,7 +352,8 @@ def chat(dialog, messages, stream=True, **kwargs):
 
         total_time_cost = (finish_chat_ts - chat_start_ts) * 1000
         check_llm_time_cost = (check_llm_ts - chat_start_ts) * 1000
-        create_retriever_time_cost = (create_retriever_ts - check_llm_ts) * 1000
+        check_langfuse_tracer_cost = (check_langfuse_tracer_ts - check_llm_ts) * 1000
+        create_retriever_time_cost = (create_retriever_ts - check_langfuse_tracer_ts) * 1000
         bind_embedding_time_cost = (bind_embedding_ts - create_retriever_ts) * 1000
         bind_llm_time_cost = (bind_llm_ts - bind_embedding_ts) * 1000
         refine_question_time_cost = (refine_question_ts - bind_llm_ts) * 1000
@@ -304,28 +362,57 @@ def chat(dialog, messages, stream=True, **kwargs):
         retrieval_time_cost = (retrieval_ts - generate_keyword_ts) * 1000
         generate_result_time_cost = (finish_chat_ts - retrieval_ts) * 1000
 
+        tk_num = num_tokens_from_string(think + answer)
         prompt += "\n\n### Query:\n%s" % " ".join(questions)
-        prompt = f"{prompt}\n\n - Total: {total_time_cost:.1f}ms\n  - Check LLM: {check_llm_time_cost:.1f}ms\n  - Create retriever: {create_retriever_time_cost:.1f}ms\n  - Bind embedding: {bind_embedding_time_cost:.1f}ms\n  - Bind LLM: {bind_llm_time_cost:.1f}ms\n  - Tune question: {refine_question_time_cost:.1f}ms\n  - Bind reranker: {bind_reranker_time_cost:.1f}ms\n  - Generate keyword: {generate_keyword_time_cost:.1f}ms\n  - Retrieval: {retrieval_time_cost:.1f}ms\n  - Generate answer: {generate_result_time_cost:.1f}ms"
-        return {"answer": think+answer, "reference": refs, "prompt": re.sub(r"\n", "  \n", prompt), "created_at": time.time()}
+        prompt = (
+            f"{prompt}\n\n"
+            "## Time elapsed:\n"
+            f"  - Total: {total_time_cost:.1f}ms\n"
+            f"  - Check LLM: {check_llm_time_cost:.1f}ms\n"
+            f"  - Check Langfuse tracer: {check_langfuse_tracer_cost:.1f}ms\n"
+            f"  - Create retriever: {create_retriever_time_cost:.1f}ms\n"
+            f"  - Bind embedding: {bind_embedding_time_cost:.1f}ms\n"
+            f"  - Bind LLM: {bind_llm_time_cost:.1f}ms\n"
+            f"  - Multi-turn optimization: {refine_question_time_cost:.1f}ms\n"
+            f"  - Bind reranker: {bind_reranker_time_cost:.1f}ms\n"
+            f"  - Generate keyword: {generate_keyword_time_cost:.1f}ms\n"
+            f"  - Retrieval: {retrieval_time_cost:.1f}ms\n"
+            f"  - Generate answer: {generate_result_time_cost:.1f}ms\n\n"
+            "## Token usage:\n"
+            f"  - Generated tokens(approximately): {tk_num}\n"
+            f"  - Token speed: {int(tk_num / (generate_result_time_cost / 1000.0))}/s"
+        )
+
+        langfuse_output = "\n" + re.sub(r"^.*?(### Query:.*)", r"\1", prompt, flags=re.DOTALL)
+        langfuse_output = {"time_elapsed:": re.sub(r"\n", "  \n", langfuse_output), "created_at": time.time()}
+
+        # Add a condition check to call the end method only if langfuse_tracer exists
+        if langfuse_tracer and "langfuse_generation" in locals():
+            langfuse_generation.end(output=langfuse_output)
+
+        return {"answer": think + answer, "reference": refs, "prompt": re.sub(r"\n", "  \n", prompt), "created_at": time.time()}
+
+    if langfuse_tracer:
+        langfuse_generation = langfuse_tracer.trace.generation(name="chat", model=llm_model_config["llm_name"], input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg})
 
     if stream:
         last_ans = ""
         answer = ""
-        for ans in chat_mdl.chat_streamly(prompt+prompt4citation, msg[1:], gen_conf):
+        for ans in chat_mdl.chat_streamly(prompt + prompt4citation, msg[1:], gen_conf):
             if thought:
                 ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
             answer = ans
-            delta_ans = ans[len(last_ans):]
+            delta_ans = ans[len(last_ans) :]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
-            yield {"answer": thought+answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
-        delta_ans = answer[len(last_ans):]
+            yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
+        delta_ans = answer[len(last_ans) :]
         if delta_ans:
-            yield {"answer": thought+answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
-        yield decorate_answer(thought+answer)
+            yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
+        yield decorate_answer(thought + answer)
     else:
-        answer = chat_mdl.chat(prompt+prompt4citation, msg[1:], gen_conf)
+        answer = chat_mdl.chat(prompt + prompt4citation, msg[1:], gen_conf)
         user_content = msg[-1].get("content", "[content not available]")
         logging.debug("User: {}|Assistant: {}".format(user_content, answer))
         res = decorate_answer(answer)
@@ -343,27 +430,22 @@ Table of database fields are as follows:
 Question are as follows:
 {}
 Please write the SQL, only SQL, without any other explanations or text.
-""".format(
-        index_name(tenant_id),
-        "\n".join([f"{k}: {v}" for k, v in field_map.items()]),
-        question
-    )
+""".format(index_name(tenant_id), "\n".join([f"{k}: {v}" for k, v in field_map.items()]), question)
     tried_times = 0
 
     def get_table():
         nonlocal sys_prompt, user_prompt, question, tried_times
-        sql = chat_mdl.chat(sys_prompt, [{"role": "user", "content": user_prompt}], {
-            "temperature": 0.06})
+        sql = chat_mdl.chat(sys_prompt, [{"role": "user", "content": user_prompt}], {"temperature": 0.06})
         sql = re.sub(r"<think>.*</think>", "", sql, flags=re.DOTALL)
         logging.debug(f"{question} ==> {user_prompt} get SQL: {sql}")
         sql = re.sub(r"[\r\n]+", " ", sql.lower())
         sql = re.sub(r".*select ", "select ", sql.lower())
         sql = re.sub(r" +", " ", sql)
         sql = re.sub(r"([;；]|```).*", "", sql)
-        if sql[:len("select ")] != "select ":
+        if sql[: len("select ")] != "select ":
             return None, None
         if not re.search(r"((sum|avg|max|min)\(|group by )", sql.lower()):
-            if sql[:len("select *")] != "select *":
+            if sql[: len("select *")] != "select *":
                 sql = "select doc_id,docnm_kwd," + sql[6:]
             else:
                 flds = []
@@ -400,11 +482,7 @@ Please write the SQL, only SQL, without any other explanations or text.
         {}
 
         Please correct the error and write SQL again, only SQL, without any other explanations or text.
-        """.format(
-            index_name(tenant_id),
-            "\n".join([f"{k}: {v}" for k, v in field_map.items()]),
-            question, sql, tbl["error"]
-        )
+        """.format(index_name(tenant_id), "\n".join([f"{k}: {v}" for k, v in field_map.items()]), question, sql, tbl["error"])
         tbl, sql = get_table()
         logging.debug("TRY it again: {}".format(sql))
 
@@ -412,24 +490,18 @@ Please write the SQL, only SQL, without any other explanations or text.
     if tbl.get("error") or len(tbl["rows"]) == 0:
         return None
 
-    docid_idx = set([ii for ii, c in enumerate(
-        tbl["columns"]) if c["name"] == "doc_id"])
-    doc_name_idx = set([ii for ii, c in enumerate(
-        tbl["columns"]) if c["name"] == "docnm_kwd"])
-    column_idx = [ii for ii in range(
-        len(tbl["columns"])) if ii not in (docid_idx | doc_name_idx)]
+    docid_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "doc_id"])
+    doc_name_idx = set([ii for ii, c in enumerate(tbl["columns"]) if c["name"] == "docnm_kwd"])
+    column_idx = [ii for ii in range(len(tbl["columns"])) if ii not in (docid_idx | doc_name_idx)]
 
     # compose Markdown table
-    columns = "|" + "|".join([re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"],
-                                                                          tbl["columns"][i]["name"])) for i in
-                              column_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
+    columns = (
+            "|" + "|".join([re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in column_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
+    )
 
-    line = "|" + "|".join(["------" for _ in range(len(column_idx))]) + \
-           ("|------|" if docid_idx and docid_idx else "")
+    line = "|" + "|".join(["------" for _ in range(len(column_idx))]) + ("|------|" if docid_idx and docid_idx else "")
 
-    rows = ["|" +
-            "|".join([rmSpace(str(r[i])) for i in column_idx]).replace("None", " ") +
-            "|" for r in tbl["rows"]]
+    rows = ["|" + "|".join([rmSpace(str(r[i])) for i in column_idx]).replace("None", " ") + "|" for r in tbl["rows"]]
     rows = [r for r in rows if re.sub(r"[ |]+", "", r)]
     if quota:
         rows = "\n".join([r + f" ##{ii}$$ |" for ii, r in enumerate(rows)])
@@ -439,11 +511,7 @@ Please write the SQL, only SQL, without any other explanations or text.
 
     if not docid_idx or not doc_name_idx:
         logging.warning("SQL missing field: " + sql)
-        return {
-            "answer": "\n".join([columns, line, rows]),
-            "reference": {"chunks": [], "doc_aggs": []},
-            "prompt": sys_prompt
-        }
+        return {"answer": "\n".join([columns, line, rows]), "reference": {"chunks": [], "doc_aggs": []}, "prompt": sys_prompt}
 
     docid_idx = list(docid_idx)[0]
     doc_name_idx = list(doc_name_idx)[0]
@@ -454,10 +522,11 @@ Please write the SQL, only SQL, without any other explanations or text.
         doc_aggs[r[docid_idx]]["count"] += 1
     return {
         "answer": "\n".join([columns, line, rows]),
-        "reference": {"chunks": [{"doc_id": r[docid_idx], "docnm_kwd": r[doc_name_idx]} for r in tbl["rows"]],
-                      "doc_aggs": [{"doc_id": did, "doc_name": d["doc_name"], "count": d["count"]} for did, d in
-                                   doc_aggs.items()]},
-        "prompt": sys_prompt
+        "reference": {
+            "chunks": [{"doc_id": r[docid_idx], "docnm_kwd": r[doc_name_idx]} for r in tbl["rows"]],
+            "doc_aggs": [{"doc_id": did, "doc_name": d["doc_name"], "count": d["count"]} for did, d in doc_aggs.items()],
+        },
+        "prompt": sys_prompt,
     }
 
 
@@ -481,10 +550,7 @@ def ask(question, kb_ids, tenant_id):
     chat_mdl = LLMBundle(tenant_id, LLMType.CHAT)
     max_tokens = chat_mdl.max_length
     tenant_ids = list(set([kb.tenant_id for kb in kbs]))
-    kbinfos = retriever.retrieval(question, embd_mdl, tenant_ids, kb_ids,
-                                  1, 12, 0.1, 0.3, aggs=False,
-                                  rank_feature=label_question(question, kbs)
-                                  )
+    kbinfos = retriever.retrieval(question, embd_mdl, tenant_ids, kb_ids, 1, 12, 0.1, 0.3, aggs=False, rank_feature=label_question(question, kbs))
     knowledges = kb_prompt(kbinfos, max_tokens)
     prompt = """
     Role: You're a smart assistant. Your name is Miss R.
@@ -506,17 +572,9 @@ def ask(question, kb_ids, tenant_id):
 
     def decorate_answer(answer):
         nonlocal knowledges, kbinfos, prompt
-        answer, idx = retriever.insert_citations(answer,
-                                                 [ck["content_ltks"]
-                                                  for ck in kbinfos["chunks"]],
-                                                 [ck["vector"]
-                                                  for ck in kbinfos["chunks"]],
-                                                 embd_mdl,
-                                                 tkweight=0.7,
-                                                 vtweight=0.3)
+        answer, idx = retriever.insert_citations(answer, [ck["content_ltks"] for ck in kbinfos["chunks"]], [ck["vector"] for ck in kbinfos["chunks"]], embd_mdl, tkweight=0.7, vtweight=0.3)
         idx = set([kbinfos["chunks"][int(i)]["doc_id"] for i in idx])
-        recall_docs = [
-            d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
+        recall_docs = [d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
         if not recall_docs:
             recall_docs = kbinfos["doc_aggs"]
         kbinfos["doc_aggs"] = recall_docs

api/db/services/document_service.py

@@ -13,33 +13,30 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-import logging
-import xxhash
 import json
+import logging
 import random
 import re
 from concurrent.futures import ThreadPoolExecutor
 from copy import deepcopy
 from datetime import datetime
 from io import BytesIO
-import trio
 
+import trio
+import xxhash
 from peewee import fn
 
-from api.db.db_utils import bulk_insert_into_db
 from api import settings
-from api.utils import current_timestamp, get_format_time, get_uuid
-from rag.settings import SVR_QUEUE_NAME
-from rag.utils.storage_factory import STORAGE_IMPL
-from rag.nlp import search, rag_tokenizer
-
-from api.db import FileType, TaskStatus, ParserType, LLMType
-from api.db.db_models import DB, Knowledgebase, Tenant, Task, UserTenant
-from api.db.db_models import Document
+from api.db import FileType, LLMType, ParserType, StatusEnum, TaskStatus, UserTenantRole
+from api.db.db_models import DB, Document, Knowledgebase, Task, Tenant, UserTenant
+from api.db.db_utils import bulk_insert_into_db
 from api.db.services.common_service import CommonService
 from api.db.services.knowledgebase_service import KnowledgebaseService
-from api.db import StatusEnum
+from api.utils import current_timestamp, get_format_time, get_uuid
+from rag.nlp import rag_tokenizer, search
+from rag.settings import get_svr_queue_name
 from rag.utils.redis_conn import REDIS_CONN
+from rag.utils.storage_factory import STORAGE_IMPL
 
 
 class DocumentService(CommonService):
@@ -96,9 +93,7 @@ class DocumentService(CommonService):
     def insert(cls, doc):
         if not cls.save(**doc):
             raise RuntimeError("Database error (Document)!")
-        e, kb = KnowledgebaseService.get_by_id(doc["kb_id"])
-        if not KnowledgebaseService.update_by_id(
-                kb.id, {"doc_num": kb.doc_num + 1}):
+        if not KnowledgebaseService.atomic_increase_doc_num_by_id(doc["kb_id"]):
             raise RuntimeError("Database error (Knowledgebase)!")
         return Document(**doc)
 
@@ -108,13 +103,13 @@ class DocumentService(CommonService):
         cls.clear_chunk_num(doc.id)
         try:
             settings.docStoreConn.delete({"doc_id": doc.id}, search.index_name(tenant_id), doc.kb_id)
-            settings.docStoreConn.update({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "community_report"], "source_id": doc.id},
+            settings.docStoreConn.update({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "subgraph", "community_report"], "source_id": doc.id},
                                          {"remove": {"source_id": doc.id}},
                                          search.index_name(tenant_id), doc.kb_id)
             settings.docStoreConn.update({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["graph"]},
                                          {"removed_kwd": "Y"},
                                          search.index_name(tenant_id), doc.kb_id)
-            settings.docStoreConn.delete({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "community_report"], "must_not": {"exists": "source_id"}},
+            settings.docStoreConn.delete({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "subgraph", "community_report"], "must_not": {"exists": "source_id"}},
                                          search.index_name(tenant_id), doc.kb_id)
         except Exception:
             pass
@@ -174,9 +169,9 @@ class DocumentService(CommonService):
                 "Document not found which is supposed to be there")
         num = Knowledgebase.update(
             token_num=Knowledgebase.token_num +
-                      token_num,
+            token_num,
             chunk_num=Knowledgebase.chunk_num +
-                      chunk_num).where(
+            chunk_num).where(
             Knowledgebase.id == kb_id).execute()
         return num
 
@@ -192,9 +187,9 @@ class DocumentService(CommonService):
                 "Document not found which is supposed to be there")
         num = Knowledgebase.update(
             token_num=Knowledgebase.token_num -
-                      token_num,
+            token_num,
             chunk_num=Knowledgebase.chunk_num -
-                      chunk_num
+            chunk_num
         ).where(
             Knowledgebase.id == kb_id).execute()
         return num
@@ -207,9 +202,9 @@ class DocumentService(CommonService):
 
         num = Knowledgebase.update(
             token_num=Knowledgebase.token_num -
-                      doc.token_num,
+            doc.token_num,
             chunk_num=Knowledgebase.chunk_num -
-                      doc.chunk_num,
+            doc.chunk_num,
             doc_num=Knowledgebase.doc_num - 1
         ).where(
             Knowledgebase.id == doc.kb_id).execute()
@@ -221,7 +216,7 @@ class DocumentService(CommonService):
         docs = cls.model.select(
             Knowledgebase.tenant_id).join(
             Knowledgebase, on=(
-                    Knowledgebase.id == cls.model.kb_id)).where(
+                Knowledgebase.id == cls.model.kb_id)).where(
             cls.model.id == doc_id, Knowledgebase.status == StatusEnum.VALID.value)
         docs = docs.dicts()
         if not docs:
@@ -243,7 +238,7 @@ class DocumentService(CommonService):
         docs = cls.model.select(
             Knowledgebase.tenant_id).join(
             Knowledgebase, on=(
-                    Knowledgebase.id == cls.model.kb_id)).where(
+                Knowledgebase.id == cls.model.kb_id)).where(
             cls.model.name == name, Knowledgebase.status == StatusEnum.VALID.value)
         docs = docs.dicts()
         if not docs:
@@ -256,7 +251,7 @@ class DocumentService(CommonService):
         docs = cls.model.select(
             cls.model.id).join(
             Knowledgebase, on=(
-                    Knowledgebase.id == cls.model.kb_id)
+                Knowledgebase.id == cls.model.kb_id)
         ).join(UserTenant, on=(UserTenant.tenant_id == Knowledgebase.tenant_id)
                ).where(cls.model.id == doc_id, UserTenant.user_id == user_id).paginate(0, 1)
         docs = docs.dicts()
@@ -267,11 +262,18 @@ class DocumentService(CommonService):
     @classmethod
     @DB.connection_context()
     def accessible4deletion(cls, doc_id, user_id):
-        docs = cls.model.select(
-            cls.model.id).join(
+        docs = cls.model.select(cls.model.id
+        ).join(
             Knowledgebase, on=(
-                    Knowledgebase.id == cls.model.kb_id)
-        ).where(cls.model.id == doc_id, Knowledgebase.created_by == user_id).paginate(0, 1)
+                Knowledgebase.id == cls.model.kb_id)
+        ).join(
+            UserTenant, on=(
+                (UserTenant.tenant_id == Knowledgebase.created_by) & (UserTenant.user_id == user_id))
+        ).where(
+            cls.model.id == doc_id,
+            UserTenant.status == StatusEnum.VALID.value,
+            ((UserTenant.role == UserTenantRole.NORMAL) | (UserTenant.role == UserTenantRole.OWNER))
+        ).paginate(0, 1)
         docs = docs.dicts()
         if not docs:
             return False
@@ -283,7 +285,7 @@ class DocumentService(CommonService):
         docs = cls.model.select(
             Knowledgebase.embd_id).join(
             Knowledgebase, on=(
-                    Knowledgebase.id == cls.model.kb_id)).where(
+                Knowledgebase.id == cls.model.kb_id)).where(
             cls.model.id == doc_id, Knowledgebase.status == StatusEnum.VALID.value)
         docs = docs.dicts()
         if not docs:
@@ -306,9 +308,9 @@ class DocumentService(CommonService):
                 Tenant.asr_id,
                 Tenant.llm_id,
             )
-                .join(Knowledgebase, on=(cls.model.kb_id == Knowledgebase.id))
-                .join(Tenant, on=(Knowledgebase.tenant_id == Tenant.id))
-                .where(cls.model.id == doc_id)
+            .join(Knowledgebase, on=(cls.model.kb_id == Knowledgebase.id))
+            .join(Tenant, on=(Knowledgebase.tenant_id == Tenant.id))
+            .where(cls.model.id == doc_id)
         )
         configs = configs.dicts()
         if not configs:
@@ -336,6 +338,8 @@ class DocumentService(CommonService):
     @classmethod
     @DB.connection_context()
     def update_parser_config(cls, id, config):
+        if not config:
+            return
         e, d = cls.get_by_id(id)
         if not e:
             raise LookupError(f"Document({id}) not found.")
@@ -372,6 +376,7 @@ class DocumentService(CommonService):
                     "progress_msg": "Task is queued...",
                     "process_begin_at": get_format_time()
                     })
+
     @classmethod
     @DB.connection_context()
     def update_meta_fields(cls, doc_id, meta_fields):
@@ -394,6 +399,7 @@ class DocumentService(CommonService):
                 has_graphrag = False
                 e, doc = DocumentService.get_by_id(d["id"])
                 status = doc.run  # TaskStatus.RUNNING.value
+                priority = 0
                 for t in tsks:
                     if 0 <= t.progress < 1:
                         finished = False
@@ -405,16 +411,17 @@ class DocumentService(CommonService):
                         has_raptor = True
                     elif t.task_type == "graphrag":
                         has_graphrag = True
+                    priority = max(priority, t.priority)
                 prg /= len(tsks)
                 if finished and bad:
                     prg = -1
                     status = TaskStatus.FAIL.value
                 elif finished:
                     if d["parser_config"].get("raptor", {}).get("use_raptor") and not has_raptor:
-                        queue_raptor_o_graphrag_tasks(d, "raptor")
+                        queue_raptor_o_graphrag_tasks(d, "raptor", priority)
                         prg = 0.98 * len(tsks) / (len(tsks) + 1)
                     elif d["parser_config"].get("graphrag", {}).get("use_graphrag") and not has_graphrag:
-                        queue_raptor_o_graphrag_tasks(d, "graphrag")
+                        queue_raptor_o_graphrag_tasks(d, "graphrag", priority)
                         prg = 0.98 * len(tsks) / (len(tsks) + 1)
                     else:
                         status = TaskStatus.DONE.value
@@ -423,7 +430,7 @@ class DocumentService(CommonService):
                 info = {
                     "process_duation": datetime.timestamp(
                         datetime.now()) -
-                                       d["process_begin_at"].timestamp(),
+                    d["process_begin_at"].timestamp(),
                     "run": status}
                 if prg != 0:
                     info["progress"] = prg
@@ -451,7 +458,7 @@ class DocumentService(CommonService):
         return False
 
 
-def queue_raptor_o_graphrag_tasks(doc, ty):
+def queue_raptor_o_graphrag_tasks(doc, ty, priority):
     chunking_config = DocumentService.get_chunking_config(doc["id"])
     hasher = xxhash.xxh64()
     for field in sorted(chunking_config.keys()):
@@ -474,17 +481,17 @@ def queue_raptor_o_graphrag_tasks(doc, ty):
     hasher.update(ty.encode("utf-8"))
     task["digest"] = hasher.hexdigest()
     bulk_insert_into_db(Task, [task], True)
-    assert REDIS_CONN.queue_product(SVR_QUEUE_NAME, message=task), "Can't access Redis. Please check the Redis' status."
+    assert REDIS_CONN.queue_product(get_svr_queue_name(priority), message=task), "Can't access Redis. Please check the Redis' status."
 
 
 def doc_upload_and_parse(conversation_id, file_objs, user_id):
-    from rag.app import presentation, picture, naive, audio, email
+    from api.db.services.api_service import API4ConversationService
+    from api.db.services.conversation_service import ConversationService
     from api.db.services.dialog_service import DialogService
     from api.db.services.file_service import FileService
     from api.db.services.llm_service import LLMBundle
     from api.db.services.user_service import TenantService
-    from api.db.services.api_service import API4ConversationService
-    from api.db.services.conversation_service import ConversationService
+    from rag.app import audio, email, naive, picture, presentation
 
     e, conv = ConversationService.get_by_id(conversation_id)
     if not e:

api/db/services/file_service.py

@@ -34,12 +34,24 @@ from rag.utils.storage_factory import STORAGE_IMPL
 
 
 class FileService(CommonService):
+    # Service class for managing file operations and storage
     model = File
 
     @classmethod
     @DB.connection_context()
     def get_by_pf_id(cls, tenant_id, pf_id, page_number, items_per_page,
                      orderby, desc, keywords):
+        # Get files by parent folder ID with pagination and filtering
+        # Args:
+        #     tenant_id: ID of the tenant
+        #     pf_id: Parent folder ID
+        #     page_number: Page number for pagination
+        #     items_per_page: Number of items per page
+        #     orderby: Field to order by
+        #     desc: Boolean indicating descending order
+        #     keywords: Search keywords
+        # Returns:
+        #     Tuple of (file_list, total_count)
         if keywords:
             files = cls.model.select().where(
                 (cls.model.tenant_id == tenant_id),
@@ -80,6 +92,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_kb_id_by_file_id(cls, file_id):
+        # Get knowledge base IDs associated with a file
+        # Args:
+        #     file_id: File ID
+        # Returns:
+        #     List of dictionaries containing knowledge base IDs and names
         kbs = (cls.model.select(*[Knowledgebase.id, Knowledgebase.name])
                .join(File2Document, on=(File2Document.file_id == file_id))
                .join(Document, on=(File2Document.document_id == Document.id))
@@ -95,6 +112,12 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_by_pf_id_name(cls, id, name):
+        # Get file by parent folder ID and name
+        # Args:
+        #     id: Parent folder ID
+        #     name: File name
+        # Returns:
+        #     File object or None if not found
         file = cls.model.select().where((cls.model.parent_id == id) & (cls.model.name == name))
         if file.count():
             e, file = cls.get_by_id(file[0].id)
@@ -106,6 +129,14 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_id_list_by_id(cls, id, name, count, res):
+        # Recursively get list of file IDs by traversing folder structure
+        # Args:
+        #     id: Starting folder ID
+        #     name: List of folder names to traverse
+        #     count: Current depth in traversal
+        #     res: List to store results
+        # Returns:
+        #     List of file IDs
         if count < len(name):
             file = cls.get_by_pf_id_name(id, name[count])
             if file:
@@ -119,6 +150,12 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_all_innermost_file_ids(cls, folder_id, result_ids):
+        # Get IDs of all files in the deepest level of folders
+        # Args:
+        #     folder_id: Starting folder ID
+        #     result_ids: List to store results
+        # Returns:
+        #     List of file IDs
         subfolders = cls.model.select().where(cls.model.parent_id == folder_id)
         if subfolders.exists():
             for subfolder in subfolders:
@@ -130,6 +167,14 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def create_folder(cls, file, parent_id, name, count):
+        # Recursively create folder structure
+        # Args:
+        #     file: Current file object
+        #     parent_id: Parent folder ID
+        #     name: List of folder names to create
+        #     count: Current depth in creation
+        # Returns:
+        #     Created file object
         if count > len(name) - 2:
             return file
         else:
@@ -148,6 +193,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def is_parent_folder_exist(cls, parent_id):
+        # Check if parent folder exists
+        # Args:
+        #     parent_id: Parent folder ID
+        # Returns:
+        #     Boolean indicating if folder exists
         parent_files = cls.model.select().where(cls.model.id == parent_id)
         if parent_files.count():
             return True
@@ -157,6 +207,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_root_folder(cls, tenant_id):
+        # Get or create root folder for tenant
+        # Args:
+        #     tenant_id: Tenant ID
+        # Returns:
+        #     Root folder dictionary
         for file in cls.model.select().where((cls.model.tenant_id == tenant_id),
                                         (cls.model.parent_id == cls.model.id)
                                         ):
@@ -179,6 +234,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_kb_folder(cls, tenant_id):
+        # Get knowledge base folder for tenant
+        # Args:
+        #     tenant_id: Tenant ID
+        # Returns:
+        #     Knowledge base folder dictionary
         for root in cls.model.select().where(
                 (cls.model.tenant_id == tenant_id), (cls.model.parent_id == cls.model.id)):
             for folder in cls.model.select().where(
@@ -190,6 +250,16 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def new_a_file_from_kb(cls, tenant_id, name, parent_id, ty=FileType.FOLDER.value, size=0, location=""):
+        # Create a new file from knowledge base
+        # Args:
+        #     tenant_id: Tenant ID
+        #     name: File name
+        #     parent_id: Parent folder ID
+        #     ty: File type
+        #     size: File size
+        #     location: File location
+        # Returns:
+        #     Created file dictionary
         for file in cls.query(tenant_id=tenant_id, parent_id=parent_id, name=name):
             return file.to_dict()
         file = {
@@ -209,6 +279,10 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def init_knowledgebase_docs(cls, root_id, tenant_id):
+        # Initialize knowledge base documents
+        # Args:
+        #     root_id: Root folder ID
+        #     tenant_id: Tenant ID
         for _ in cls.model.select().where((cls.model.name == KNOWLEDGEBASE_FOLDER_NAME)\
                                           & (cls.model.parent_id == root_id)):
             return
@@ -222,6 +296,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_parent_folder(cls, file_id):
+        # Get parent folder of a file
+        # Args:
+        #     file_id: File ID
+        # Returns:
+        #     Parent folder object
         file = cls.model.select().where(cls.model.id == file_id)
         if file.count():
             e, file = cls.get_by_id(file[0].parent_id)
@@ -234,6 +313,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_all_parent_folders(cls, start_id):
+        # Get all parent folders in path
+        # Args:
+        #     start_id: Starting file ID
+        # Returns:
+        #     List of parent folder objects
         parent_folders = []
         current_id = start_id
         while current_id:
@@ -249,6 +333,11 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def insert(cls, file):
+        # Insert a new file record
+        # Args:
+        #     file: File data dictionary
+        # Returns:
+        #     Created file object
         if not cls.save(**file):
             raise RuntimeError("Database error (File)!")
         return File(**file)
@@ -256,6 +345,7 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def delete(cls, file):
+        #
         return cls.delete_by_id(file.id)
 
     @classmethod

api/db/services/knowledgebase_service.py

@@ -13,28 +13,80 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-from api.db import StatusEnum, TenantPermission
-from api.db.db_models import Knowledgebase, DB, Tenant, User, UserTenant,Document
-from api.db.services.common_service import CommonService
+from datetime import datetime
+
 from peewee import fn
 
+from api.db import StatusEnum, TenantPermission
+from api.db.db_models import DB, Document, Knowledgebase, Tenant, User, UserTenant
+from api.db.services.common_service import CommonService
+from api.utils import current_timestamp, datetime_format
+
 
 class KnowledgebaseService(CommonService):
+    """Service class for managing knowledge base operations.
+
+    This class extends CommonService to provide specialized functionality for knowledge base
+    management, including document parsing status tracking, access control, and configuration
+    management. It handles operations such as listing, creating, updating, and deleting
+    knowledge bases, as well as managing their associated documents and permissions.
+
+    The class implements a comprehensive set of methods for:
+    - Document parsing status verification
+    - Knowledge base access control
+    - Parser configuration management
+    - Tenant-based knowledge base organization
+
+    Attributes:
+        model: The Knowledgebase model class for database operations.
+    """
     model = Knowledgebase
 
     @classmethod
     @DB.connection_context()
-    def is_parsed_done(cls, kb_id):
-        """
-        Check if all documents in the knowledge base have completed parsing
+    def accessible4deletion(cls, kb_id, user_id):
+        """Check if a knowledge base can be deleted by a specific user.
+
+        This method verifies whether a user has permission to delete a knowledge base
+        by checking if they are the creator of that knowledge base.
 
         Args:
-            kb_id: Knowledge base ID
+            kb_id (str): The unique identifier of the knowledge base to check.
+            user_id (str): The unique identifier of the user attempting the deletion.
 
         Returns:
-            If all documents are parsed successfully, returns (True, None)
-            If any document is not fully parsed, returns (False, error_message)
+            bool: True if the user has permission to delete the knowledge base,
+                  False if the user doesn't have permission or the knowledge base doesn't exist.
+
+        Example:
+            >>> KnowledgebaseService.accessible4deletion("kb123", "user456")
+            True
+
+        Note:
+            - This method only checks creator permissions
+            - A return value of False can mean either:
+                1. The knowledge base doesn't exist
+                2. The user is not the creator of the knowledge base
         """
+        # Check if a knowledge base can be deleted by a user
+        docs = cls.model.select(
+            cls.model.id).where(cls.model.id == kb_id, cls.model.created_by == user_id).paginate(0, 1)
+        docs = docs.dicts()
+        if not docs:
+            return False
+        return True
+
+    @classmethod
+    @DB.connection_context()
+    def is_parsed_done(cls, kb_id):
+        # Check if all documents in the knowledge base have completed parsing
+        #
+        # Args:
+        #     kb_id: Knowledge base ID
+        #
+        # Returns:
+        #     If all documents are parsed successfully, returns (True, None)
+        #     If any document is not fully parsed, returns (False, error_message)
         from api.db import TaskStatus
         from api.db.services.document_service import DocumentService
 
@@ -60,11 +112,16 @@ class KnowledgebaseService(CommonService):
 
     @classmethod
     @DB.connection_context()
-    def list_documents_by_ids(cls,kb_ids):
-        doc_ids=cls.model.select(Document.id.alias("document_id")).join(Document,on=(cls.model.id == Document.kb_id)).where(
+    def list_documents_by_ids(cls, kb_ids):
+        # Get document IDs associated with given knowledge base IDs
+        # Args:
+        #     kb_ids: List of knowledge base IDs
+        # Returns:
+        #     List of document IDs
+        doc_ids = cls.model.select(Document.id.alias("document_id")).join(Document, on=(cls.model.id == Document.kb_id)).where(
             cls.model.id.in_(kb_ids)
         )
-        doc_ids =list(doc_ids.dicts())
+        doc_ids = list(doc_ids.dicts())
         doc_ids = [doc["document_id"] for doc in doc_ids]
         return doc_ids
 
@@ -75,12 +132,25 @@ class KnowledgebaseService(CommonService):
                           orderby, desc, keywords,
                           parser_id=None
                           ):
+        # Get knowledge bases by tenant IDs with pagination and filtering
+        # Args:
+        #     joined_tenant_ids: List of tenant IDs
+        #     user_id: Current user ID
+        #     page_number: Page number for pagination
+        #     items_per_page: Number of items per page
+        #     orderby: Field to order by
+        #     desc: Boolean indicating descending order
+        #     keywords: Search keywords
+        #     parser_id: Optional parser ID filter
+        # Returns:
+        #     Tuple of (knowledge_base_list, total_count)
         fields = [
             cls.model.id,
             cls.model.avatar,
             cls.model.name,
             cls.model.language,
             cls.model.description,
+            cls.model.tenant_id,
             cls.model.permission,
             cls.model.doc_num,
             cls.model.token_num,
@@ -115,13 +185,19 @@ class KnowledgebaseService(CommonService):
 
         count = kbs.count()
 
-        kbs = kbs.paginate(page_number, items_per_page)
+        if page_number and items_per_page:
+            kbs = kbs.paginate(page_number, items_per_page)
 
         return list(kbs.dicts()), count
 
     @classmethod
     @DB.connection_context()
     def get_kb_ids(cls, tenant_id):
+        # Get all knowledge base IDs for a tenant
+        # Args:
+        #     tenant_id: Tenant ID
+        # Returns:
+        #     List of knowledge base IDs
         fields = [
             cls.model.id,
         ]
@@ -132,9 +208,13 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_detail(cls, kb_id):
+        # Get detailed information about a knowledge base
+        # Args:
+        #     kb_id: Knowledge base ID
+        # Returns:
+        #     Dictionary containing knowledge base details
         fields = [
             cls.model.id,
-            # Tenant.embd_id,
             cls.model.embd_id,
             cls.model.avatar,
             cls.model.name,
@@ -148,24 +228,28 @@ class KnowledgebaseService(CommonService):
             cls.model.parser_config,
             cls.model.pagerank]
         kbs = cls.model.select(*fields).join(Tenant, on=(
-                (Tenant.id == cls.model.tenant_id) & (Tenant.status == StatusEnum.VALID.value))).where(
+            (Tenant.id == cls.model.tenant_id) & (Tenant.status == StatusEnum.VALID.value))).where(
             (cls.model.id == kb_id),
             (cls.model.status == StatusEnum.VALID.value)
         )
         if not kbs:
             return
         d = kbs[0].to_dict()
-        # d["embd_id"] = kbs[0].tenant.embd_id
         return d
 
     @classmethod
     @DB.connection_context()
     def update_parser_config(cls, id, config):
+        # Update parser configuration for a knowledge base
+        # Args:
+        #     id: Knowledge base ID
+        #     config: New parser configuration
         e, m = cls.get_by_id(id)
         if not e:
             raise LookupError(f"knowledgebase({id}) not found.")
 
         def dfs_update(old, new):
+            # Deep update of nested configuration
             for k, v in new.items():
                 if k not in old:
                     old[k] = v
@@ -185,6 +269,11 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_field_map(cls, ids):
+        # Get field mappings for knowledge bases
+        # Args:
+        #     ids: List of knowledge base IDs
+        # Returns:
+        #     Dictionary of field mappings
         conf = {}
         for k in cls.get_by_ids(ids):
             if k.parser_config and "field_map" in k.parser_config:
@@ -194,6 +283,12 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_by_name(cls, kb_name, tenant_id):
+        # Get knowledge base by name and tenant ID
+        # Args:
+        #     kb_name: Knowledge base name
+        #     tenant_id: Tenant ID
+        # Returns:
+        #     Tuple of (exists, knowledge_base)
         kb = cls.model.select().where(
             (cls.model.name == kb_name)
             & (cls.model.tenant_id == tenant_id)
@@ -206,12 +301,27 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_all_ids(cls):
+        # Get all knowledge base IDs
+        # Returns:
+        #     List of all knowledge base IDs
         return [m["id"] for m in cls.model.select(cls.model.id).dicts()]
 
     @classmethod
     @DB.connection_context()
     def get_list(cls, joined_tenant_ids, user_id,
                  page_number, items_per_page, orderby, desc, id, name):
+        # Get list of knowledge bases with filtering and pagination
+        # Args:
+        #     joined_tenant_ids: List of tenant IDs
+        #     user_id: Current user ID
+        #     page_number: Page number for pagination
+        #     items_per_page: Number of items per page
+        #     orderby: Field to order by
+        #     desc: Boolean indicating descending order
+        #     id: Optional ID filter
+        #     name: Optional name filter
+        # Returns:
+        #     List of knowledge bases
         kbs = cls.model.select()
         if id:
             kbs = kbs.where(cls.model.id == id)
@@ -220,7 +330,7 @@ class KnowledgebaseService(CommonService):
         kbs = kbs.where(
             ((cls.model.tenant_id.in_(joined_tenant_ids) & (cls.model.permission ==
                                                             TenantPermission.TEAM.value)) | (
-                     cls.model.tenant_id == user_id))
+                cls.model.tenant_id == user_id))
             & (cls.model.status == StatusEnum.VALID.value)
         )
         if desc:
@@ -235,9 +345,15 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def accessible(cls, kb_id, user_id):
+        # Check if a knowledge base is accessible by a user
+        # Args:
+        #     kb_id: Knowledge base ID
+        #     user_id: User ID
+        # Returns:
+        #     Boolean indicating accessibility
         docs = cls.model.select(
             cls.model.id).join(UserTenant, on=(UserTenant.tenant_id == Knowledgebase.tenant_id)
-            ).where(cls.model.id == kb_id, UserTenant.user_id == user_id).paginate(0, 1)
+                               ).where(cls.model.id == kb_id, UserTenant.user_id == user_id).paginate(0, 1)
         docs = docs.dicts()
         if not docs:
             return False
@@ -246,26 +362,64 @@ class KnowledgebaseService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_kb_by_id(cls, kb_id, user_id):
+        # Get knowledge base by ID and user ID
+        # Args:
+        #     kb_id: Knowledge base ID
+        #     user_id: User ID
+        # Returns:
+        #     List containing knowledge base information
         kbs = cls.model.select().join(UserTenant, on=(UserTenant.tenant_id == Knowledgebase.tenant_id)
-            ).where(cls.model.id == kb_id, UserTenant.user_id == user_id).paginate(0, 1)
+                                      ).where(cls.model.id == kb_id, UserTenant.user_id == user_id).paginate(0, 1)
         kbs = kbs.dicts()
         return list(kbs)
 
     @classmethod
     @DB.connection_context()
     def get_kb_by_name(cls, kb_name, user_id):
+        # Get knowledge base by name and user ID
+        # Args:
+        #     kb_name: Knowledge base name
+        #     user_id: User ID
+        # Returns:
+        #     List containing knowledge base information
         kbs = cls.model.select().join(UserTenant, on=(UserTenant.tenant_id == Knowledgebase.tenant_id)
-            ).where(cls.model.name == kb_name, UserTenant.user_id == user_id).paginate(0, 1)
+                                      ).where(cls.model.name == kb_name, UserTenant.user_id == user_id).paginate(0, 1)
         kbs = kbs.dicts()
         return list(kbs)
 
     @classmethod
     @DB.connection_context()
-    def accessible4deletion(cls, kb_id, user_id):
-        docs = cls.model.select(
-            cls.model.id).where(cls.model.id == kb_id, cls.model.created_by == user_id).paginate(0, 1)
-        docs = docs.dicts()
-        if not docs:
-            return False
-        return True
+    def atomic_increase_doc_num_by_id(cls, kb_id):
+        data = {}
+        data["update_time"] = current_timestamp()
+        data["update_date"] = datetime_format(datetime.now())
+        data["doc_num"] = cls.model.doc_num + 1
+        num = cls.model.update(data).where(cls.model.id == kb_id).execute()
+        return num
+
+    @classmethod
+    @DB.connection_context()
+    def update_document_number_in_init(cls, kb_id, doc_num):
+        """
+        Only use this function when init system
+        """
+        ok, kb = cls.get_by_id(kb_id)
+        if not ok:
+            return
+        kb.doc_num = doc_num
+
+        dirty_fields = kb.dirty_fields
+        if cls.model._meta.combined.get("update_time") in dirty_fields:
+            dirty_fields.remove(cls.model._meta.combined["update_time"])
+
+        if cls.model._meta.combined.get("update_date") in dirty_fields:
+            dirty_fields.remove(cls.model._meta.combined["update_date"])
+
+        try:
+            kb.save(only=dirty_fields)
+        except ValueError as e:
+            if str(e) == "no data to save!":
+                pass # that's OK
+            else:
+                raise e
 

api/db/services/langfuse_service.py

@@ -0,0 +1,71 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+from datetime import datetime
+
+import peewee
+
+from api.db.db_models import DB, TenantLangfuse
+from api.db.services.common_service import CommonService
+from api.utils import current_timestamp, datetime_format
+
+
+class TenantLangfuseService(CommonService):
+    """
+    All methods that modify the status should be enclosed within a DB.atomic() context to ensure atomicity
+    and maintain data integrity in case of errors during execution.
+    """
+
+    model = TenantLangfuse
+
+    @classmethod
+    @DB.connection_context()
+    def filter_by_tenant(cls, tenant_id):
+        fields = [cls.model.tenant_id, cls.model.host, cls.model.secret_key, cls.model.public_key]
+        try:
+            keys = cls.model.select(*fields).where(cls.model.tenant_id == tenant_id).first()
+            return keys
+        except peewee.DoesNotExist:
+            return None
+
+    @classmethod
+    @DB.connection_context()
+    def filter_by_tenant_with_info(cls, tenant_id):
+        fields = [cls.model.tenant_id, cls.model.host, cls.model.secret_key, cls.model.public_key]
+        try:
+            keys = cls.model.select(*fields).where(cls.model.tenant_id == tenant_id).dicts().first()
+            return keys
+        except peewee.DoesNotExist:
+            return None
+
+    @classmethod
+    def update_by_tenant(cls, tenant_id, langfuse_keys):
+        langfuse_keys["update_time"] = current_timestamp()
+        langfuse_keys["update_date"] = datetime_format(datetime.now())
+        return cls.model.update(**langfuse_keys).where(cls.model.tenant_id == tenant_id).execute()
+
+    @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())
+        obj = cls.model.create(**kwargs)
+        return obj
+
+    @classmethod
+    def delete_model(cls, langfuse_model):
+        langfuse_model.delete_instance()

api/db/services/llm_service.py

@@ -13,17 +13,17 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-import json
 import logging
-import os
 
-from api.db.services.user_service import TenantService
-from api.utils.file_utils import get_project_base_directory
-from rag.llm import EmbeddingModel, CvModel, ChatModel, RerankModel, Seq2txtModel, TTSModel
+from langfuse import Langfuse
+
+from api import settings
 from api.db import LLMType
-from api.db.db_models import DB
-from api.db.db_models import LLMFactories, LLM, TenantLLM
+from api.db.db_models import DB, LLM, LLMFactories, TenantLLM
 from api.db.services.common_service import CommonService
+from api.db.services.langfuse_service import TenantLangfuseService
+from api.db.services.user_service import TenantService
+from rag.llm import ChatModel, CvModel, EmbeddingModel, RerankModel, Seq2txtModel, TTSModel
 
 
 class LLMFactoriesService(CommonService):
@@ -52,16 +52,8 @@ class TenantLLMService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_my_llms(cls, tenant_id):
-        fields = [
-            cls.model.llm_factory,
-            LLMFactories.logo,
-            LLMFactories.tags,
-            cls.model.model_type,
-            cls.model.llm_name,
-            cls.model.used_tokens
-        ]
-        objs = cls.model.select(*fields).join(LLMFactories, on=(cls.model.llm_factory == LLMFactories.name)).where(
-            cls.model.tenant_id == tenant_id, ~cls.model.api_key.is_null()).dicts()
+        fields = [cls.model.llm_factory, LLMFactories.logo, LLMFactories.tags, cls.model.model_type, cls.model.llm_name, cls.model.used_tokens]
+        objs = cls.model.select(*fields).join(LLMFactories, on=(cls.model.llm_factory == LLMFactories.name)).where(cls.model.tenant_id == tenant_id, ~cls.model.api_key.is_null()).dicts()
 
         return list(objs)
 
@@ -75,7 +67,7 @@ class TenantLLMService(CommonService):
 
         # model name must be xxx@yyy
         try:
-            model_factories = json.load(open(os.path.join(get_project_base_directory(), "conf/llm_factories.json"), "r"))["factory_llm_infos"]
+            model_factories = settings.FACTORY_LLM_INFOS
             model_providers = set([f["name"] for f in model_factories])
             if arr[-1] not in model_providers:
                 return model_name, None
@@ -110,6 +102,9 @@ class TenantLLMService(CommonService):
         mdlnm, fid = TenantLLMService.split_model_name_and_factory(mdlnm)
         if model_config:
             model_config = model_config.to_dict()
+            llm = LLMService.query(llm_name=mdlnm) if not fid else LLMService.query(llm_name=mdlnm, fid=fid)
+            if llm:
+                model_config["is_tools"] = llm[0].is_tools
         if not model_config:
             if llm_type in [LLMType.EMBEDDING, LLMType.RERANK]:
                 llm = LLMService.query(llm_name=mdlnm) if not fid else LLMService.query(llm_name=mdlnm, fid=fid)
@@ -117,8 +112,7 @@ class TenantLLMService(CommonService):
                     model_config = {"llm_factory": llm[0].fid, "api_key": "", "llm_name": mdlnm, "api_base": ""}
             if not model_config:
                 if mdlnm == "flag-embedding":
-                    model_config = {"llm_factory": "Tongyi-Qianwen", "api_key": "",
-                                    "llm_name": llm_name, "api_base": ""}
+                    model_config = {"llm_factory": "Tongyi-Qianwen", "api_key": "", "llm_name": llm_name, "api_base": ""}
                 else:
                     if not mdlnm:
                         raise LookupError(f"Type of {llm_type} model is not set.")
@@ -127,43 +121,32 @@ class TenantLLMService(CommonService):
 
     @classmethod
     @DB.connection_context()
-    def model_instance(cls, tenant_id, llm_type,
-                       llm_name=None, lang="Chinese"):
+    def model_instance(cls, tenant_id, llm_type, llm_name=None, lang="Chinese"):
         model_config = TenantLLMService.get_model_config(tenant_id, llm_type, llm_name)
         if llm_type == LLMType.EMBEDDING.value:
             if model_config["llm_factory"] not in EmbeddingModel:
                 return
-            return EmbeddingModel[model_config["llm_factory"]](
-                model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
+            return EmbeddingModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
 
         if llm_type == LLMType.RERANK:
             if model_config["llm_factory"] not in RerankModel:
                 return
-            return RerankModel[model_config["llm_factory"]](
-                model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
+            return RerankModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
 
         if llm_type == LLMType.IMAGE2TEXT.value:
             if model_config["llm_factory"] not in CvModel:
                 return
-            return CvModel[model_config["llm_factory"]](
-                model_config["api_key"], model_config["llm_name"], lang,
-                base_url=model_config["api_base"]
-            )
+            return CvModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], lang, base_url=model_config["api_base"])
 
         if llm_type == LLMType.CHAT.value:
             if model_config["llm_factory"] not in ChatModel:
                 return
-            return ChatModel[model_config["llm_factory"]](
-                model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
+            return ChatModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
 
         if llm_type == LLMType.SPEECH2TEXT:
             if model_config["llm_factory"] not in Seq2txtModel:
                 return
-            return Seq2txtModel[model_config["llm_factory"]](
-                key=model_config["api_key"], model_name=model_config["llm_name"],
-                lang=lang,
-                base_url=model_config["api_base"]
-            )
+            return Seq2txtModel[model_config["llm_factory"]](key=model_config["api_key"], model_name=model_config["llm_name"], lang=lang, base_url=model_config["api_base"])
         if llm_type == LLMType.TTS:
             if model_config["llm_factory"] not in TTSModel:
                 return
@@ -176,6 +159,12 @@ class TenantLLMService(CommonService):
     @classmethod
     @DB.connection_context()
     def increase_usage(cls, tenant_id, llm_type, used_tokens, llm_name=None):
+        try:
+            if not DB.is_connection_usable():
+                DB.connect()
+        except Exception:
+            DB.close()
+            DB.connect()
         e, tenant = TenantService.get_by_id(tenant_id)
         if not e:
             logging.error(f"Tenant not found: {tenant_id}")
@@ -187,7 +176,7 @@ class TenantLLMService(CommonService):
             LLMType.IMAGE2TEXT.value: tenant.img2txt_id,
             LLMType.CHAT.value: tenant.llm_id if not llm_name else llm_name,
             LLMType.RERANK.value: tenant.rerank_id if not llm_name else llm_name,
-            LLMType.TTS.value: tenant.tts_id if not llm_name else llm_name
+            LLMType.TTS.value: tenant.tts_id if not llm_name else llm_name,
         }
 
         mdlnm = llm_map.get(llm_type)
@@ -198,17 +187,13 @@ class TenantLLMService(CommonService):
         llm_name, llm_factory = TenantLLMService.split_model_name_and_factory(mdlnm)
 
         try:
-            num = cls.model.update(
-                used_tokens=cls.model.used_tokens + used_tokens
-            ).where(
-                cls.model.tenant_id == tenant_id,
-                cls.model.llm_name == llm_name,
-                cls.model.llm_factory == llm_factory if llm_factory else True
-            ).execute()
+            num = (
+                cls.model.update(used_tokens=cls.model.used_tokens + used_tokens)
+                .where(cls.model.tenant_id == tenant_id, cls.model.llm_name == llm_name, cls.model.llm_factory == llm_factory if llm_factory else True)
+                .execute()
+            )
         except Exception:
-            logging.exception(
-                "TenantLLMService.increase_usage got exception,Failed to update used_tokens for tenant_id=%s, llm_name=%s",
-                tenant_id, llm_name)
+            logging.exception("TenantLLMService.increase_usage got exception,Failed to update used_tokens for tenant_id=%s, llm_name=%s", tenant_id, llm_name)
             return 0
 
         return num
@@ -216,11 +201,7 @@ class TenantLLMService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_openai_models(cls):
-        objs = cls.model.select().where(
-            (cls.model.llm_factory == "OpenAI"),
-            ~(cls.model.llm_name == "text-embedding-3-small"),
-            ~(cls.model.llm_name == "text-embedding-3-large")
-        ).dicts()
+        objs = cls.model.select().where((cls.model.llm_factory == "OpenAI"), ~(cls.model.llm_name == "text-embedding-3-small"), ~(cls.model.llm_name == "text-embedding-3-large")).dicts()
         return list(objs)
 
 
@@ -229,79 +210,174 @@ class LLMBundle:
         self.tenant_id = tenant_id
         self.llm_type = llm_type
         self.llm_name = llm_name
-        self.mdl = TenantLLMService.model_instance(
-            tenant_id, llm_type, llm_name, lang=lang)
-        assert self.mdl, "Can't find model for {}/{}/{}".format(
-            tenant_id, llm_type, llm_name)
+        self.mdl = TenantLLMService.model_instance(tenant_id, llm_type, llm_name, lang=lang)
+        assert self.mdl, "Can't find model for {}/{}/{}".format(tenant_id, llm_type, llm_name)
         model_config = TenantLLMService.get_model_config(tenant_id, llm_type, llm_name)
         self.max_length = model_config.get("max_tokens", 8192)
 
+        self.is_tools = model_config.get("is_tools", False)
+
+        langfuse_keys = TenantLangfuseService.filter_by_tenant(tenant_id=tenant_id)
+        if langfuse_keys:
+            langfuse = Langfuse(public_key=langfuse_keys.public_key, secret_key=langfuse_keys.secret_key, host=langfuse_keys.host)
+            if langfuse.auth_check():
+                self.langfuse = langfuse
+                self.trace = self.langfuse.trace(name=f"{self.llm_type}-{self.llm_name}")
+        else:
+            self.langfuse = None
+
+    def bind_tools(self, toolcall_session, tools):
+        if not self.is_tools:
+            return
+        self.mdl.bind_tools(toolcall_session, tools)
+
     def encode(self, texts: list):
+        if self.langfuse:
+            generation = self.trace.generation(name="encode", model=self.llm_name, input={"texts": texts})
+
         embeddings, used_tokens = self.mdl.encode(texts)
-        if not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens):
-            logging.error(
-                "LLMBundle.encode can't update token usage for {}/EMBEDDING used_tokens: {}".format(self.tenant_id, used_tokens))
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.encode can't update token usage for {}/EMBEDDING used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(usage_details={"total_tokens": used_tokens})
+
         return embeddings, used_tokens
 
     def encode_queries(self, query: str):
+        if self.langfuse:
+            generation = self.trace.generation(name="encode_queries", model=self.llm_name, input={"query": query})
+
         emd, used_tokens = self.mdl.encode_queries(query)
-        if not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens):
-            logging.error(
-                "LLMBundle.encode_queries can't update token usage for {}/EMBEDDING used_tokens: {}".format(self.tenant_id, used_tokens))
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.encode_queries can't update token usage for {}/EMBEDDING used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(usage_details={"total_tokens": used_tokens})
+
         return emd, used_tokens
 
     def similarity(self, query: str, texts: list):
+        if self.langfuse:
+            generation = self.trace.generation(name="similarity", model=self.llm_name, input={"query": query, "texts": texts})
+
         sim, used_tokens = self.mdl.similarity(query, texts)
-        if not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens):
-            logging.error(
-                "LLMBundle.similarity can't update token usage for {}/RERANK used_tokens: {}".format(self.tenant_id, used_tokens))
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.similarity can't update token usage for {}/RERANK used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(usage_details={"total_tokens": used_tokens})
+
         return sim, used_tokens
 
     def describe(self, image, max_tokens=300):
-        txt, used_tokens = self.mdl.describe(image, max_tokens)
-        if not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens):
-            logging.error(
-                "LLMBundle.describe can't update token usage for {}/IMAGE2TEXT used_tokens: {}".format(self.tenant_id, used_tokens))
+        if self.langfuse:
+            generation = self.trace.generation(name="describe", metadata={"model": self.llm_name})
+
+        txt, used_tokens = self.mdl.describe(image)
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.describe can't update token usage for {}/IMAGE2TEXT used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(output={"output": txt}, usage_details={"total_tokens": used_tokens})
+
+        return txt
+
+    def describe_with_prompt(self, image, prompt):
+        if self.langfuse:
+            generation = self.trace.generation(name="describe_with_prompt", metadata={"model": self.llm_name, "prompt": prompt})
+
+        txt, used_tokens = self.mdl.describe_with_prompt(image, prompt)
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.describe can't update token usage for {}/IMAGE2TEXT used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(output={"output": txt}, usage_details={"total_tokens": used_tokens})
+
         return txt
 
     def transcription(self, audio):
+        if self.langfuse:
+            generation = self.trace.generation(name="transcription", metadata={"model": self.llm_name})
+
         txt, used_tokens = self.mdl.transcription(audio)
-        if not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens):
-            logging.error(
-                "LLMBundle.transcription can't update token usage for {}/SEQUENCE2TXT used_tokens: {}".format(self.tenant_id, used_tokens))
+        if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens):
+            logging.error("LLMBundle.transcription can't update token usage for {}/SEQUENCE2TXT used_tokens: {}".format(self.tenant_id, used_tokens))
+
+        if self.langfuse:
+            generation.end(output={"output": txt}, usage_details={"total_tokens": used_tokens})
+
         return txt
 
     def tts(self, text):
+        if self.langfuse:
+            span = self.trace.span(name="tts", input={"text": text})
+
         for chunk in self.mdl.tts(text):
             if isinstance(chunk, int):
-                if not TenantLLMService.increase_usage(
-                        self.tenant_id, self.llm_type, chunk, self.llm_name):
-                    logging.error(
-                        "LLMBundle.tts can't update token usage for {}/TTS".format(self.tenant_id))
+                if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, chunk, self.llm_name):
+                    logging.error("LLMBundle.tts can't update token usage for {}/TTS".format(self.tenant_id))
                 return
             yield chunk
 
+        if self.langfuse:
+            span.end()
+
+    def _remove_reasoning_content(self, txt: str) -> str:
+        first_think_start = txt.find("<think>")
+        if first_think_start == -1:
+            return txt
+
+        last_think_end = txt.rfind("</think>")
+        if last_think_end == -1:
+            return txt
+
+        if last_think_end < first_think_start:
+            return txt
+
+        return txt[last_think_end + len("</think>") :]
+
     def chat(self, system, history, gen_conf):
-        txt, used_tokens = self.mdl.chat(system, history, gen_conf)
-        if isinstance(txt, int) and not TenantLLMService.increase_usage(
-                self.tenant_id, self.llm_type, used_tokens, self.llm_name):
-            logging.error(
-                "LLMBundle.chat can't update token usage for {}/CHAT llm_name: {}, used_tokens: {}".format(self.tenant_id, self.llm_name,
-                                                                                                           used_tokens))
+        if self.langfuse:
+            generation = self.trace.generation(name="chat", model=self.llm_name, input={"system": system, "history": history})
+
+        chat = self.mdl.chat
+        if self.is_tools and self.mdl.is_tools:
+            chat = self.mdl.chat_with_tools
+
+        txt, used_tokens = chat(system, history, gen_conf)
+        txt = self._remove_reasoning_content(txt)
+
+        if isinstance(txt, int) and not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, used_tokens, self.llm_name):
+            logging.error("LLMBundle.chat can't update token usage for {}/CHAT llm_name: {}, used_tokens: {}".format(self.tenant_id, self.llm_name, used_tokens))
+
+        if self.langfuse:
+            generation.end(output={"output": txt}, usage_details={"total_tokens": used_tokens})
+
         return txt
 
     def chat_streamly(self, system, history, gen_conf):
-        for txt in self.mdl.chat_streamly(system, history, gen_conf):
+        if self.langfuse:
+            generation = self.trace.generation(name="chat_streamly", model=self.llm_name, input={"system": system, "history": history})
+
+        ans = ""
+        chat_streamly = self.mdl.chat_streamly
+        total_tokens = 0 
+        if self.is_tools and self.mdl.is_tools:
+            chat_streamly = self.mdl.chat_streamly_with_tools
+
+        for txt in chat_streamly(system, history, gen_conf):
             if isinstance(txt, int):
-                if not TenantLLMService.increase_usage(
-                        self.tenant_id, self.llm_type, txt, self.llm_name):
-                    logging.error(
-                        "LLMBundle.chat_streamly can't update token usage for {}/CHAT llm_name: {}, content: {}".format(self.tenant_id, self.llm_name,
-                                                                                                                        txt))
-                return
-            yield txt
+                total_tokens = txt
+                if self.langfuse:
+                    generation.end(output={"output": ans})
+                break
+
+            if txt.endswith("</think>"):
+                ans = ans.rstrip("</think>")
+
+            ans += txt
+            yield ans
+        if total_tokens > 0:
+            if not TenantLLMService.increase_usage(self.tenant_id, self.llm_type, txt, self.llm_name):
+                logging.error("LLMBundle.chat_streamly can't update token usage for {}/CHAT llm_name: {}, content: {}".format(self.tenant_id, self.llm_name, txt))

api/db/services/task_service.py

@@ -28,7 +28,7 @@ from api.db.services.common_service import CommonService
 from api.db.services.document_service import DocumentService
 from api.utils import current_timestamp, get_uuid
 from deepdoc.parser.excel_parser import RAGFlowExcelParser
-from rag.settings import SVR_QUEUE_NAME
+from rag.settings import get_svr_queue_name
 from rag.utils.storage_factory import STORAGE_IMPL
 from rag.utils.redis_conn import REDIS_CONN
 from api import settings
@@ -36,6 +36,12 @@ from rag.nlp import search
 
 
 def trim_header_by_lines(text: str, max_length) -> str:
+    # Trim header text to maximum length while preserving line breaks
+    # Args:
+    #     text: Input text to trim
+    #     max_length: Maximum allowed length
+    # Returns:
+    #     Trimmed text
     len_text = len(text)
     if len_text <= max_length:
         return text
@@ -46,11 +52,37 @@ def trim_header_by_lines(text: str, max_length) -> str:
 
 
 class TaskService(CommonService):
+    """Service class for managing document processing tasks.
+    
+    This class extends CommonService to provide specialized functionality for document
+    processing task management, including task creation, progress tracking, and chunk
+    management. It handles various document types (PDF, Excel, etc.) and manages their
+    processing lifecycle.
+    
+    The class implements a robust task queue system with retry mechanisms and progress
+    tracking, supporting both synchronous and asynchronous task execution.
+    
+    Attributes:
+        model: The Task model class for database operations.
+    """
     model = Task
 
     @classmethod
     @DB.connection_context()
     def get_task(cls, task_id):
+        """Retrieve detailed task information by task ID.
+    
+        This method fetches comprehensive task details including associated document,
+        knowledge base, and tenant information. It also handles task retry logic and
+        progress updates.
+    
+        Args:
+            task_id (str): The unique identifier of the task to retrieve.
+    
+        Returns:
+            dict: Task details dictionary containing all task information and related metadata.
+                 Returns None if task is not found or has exceeded retry limit.
+        """
         fields = [
             cls.model.id,
             cls.model.doc_id,
@@ -105,6 +137,18 @@ class TaskService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_tasks(cls, doc_id: str):
+        """Retrieve all tasks associated with a document.
+    
+        This method fetches all processing tasks for a given document, ordered by page
+        number and creation time. It includes task progress and chunk information.
+    
+        Args:
+            doc_id (str): The unique identifier of the document.
+    
+        Returns:
+            list[dict]: List of task dictionaries containing task details.
+                       Returns None if no tasks are found.
+        """
         fields = [
             cls.model.id,
             cls.model.from_page,
@@ -124,11 +168,31 @@ class TaskService(CommonService):
     @classmethod
     @DB.connection_context()
     def update_chunk_ids(cls, id: str, chunk_ids: str):
+        """Update the chunk IDs associated with a task.
+    
+        This method updates the chunk_ids field of a task, which stores the IDs of
+        processed document chunks in a space-separated string format.
+    
+        Args:
+            id (str): The unique identifier of the task.
+            chunk_ids (str): Space-separated string of chunk identifiers.
+        """
         cls.model.update(chunk_ids=chunk_ids).where(cls.model.id == id).execute()
 
     @classmethod
     @DB.connection_context()
     def get_ongoing_doc_name(cls):
+        """Get names of documents that are currently being processed.
+    
+        This method retrieves information about documents that are in the processing state,
+        including their locations and associated IDs. It uses database locking to ensure
+        thread safety when accessing the task information.
+    
+        Returns:
+            list[tuple]: A list of tuples, each containing (parent_id/kb_id, location)
+                        for documents currently being processed. Returns empty list if
+                        no documents are being processed.
+        """
         with DB.lock("get_task", -1):
             docs = (
                 cls.model.select(
@@ -172,6 +236,18 @@ class TaskService(CommonService):
     @classmethod
     @DB.connection_context()
     def do_cancel(cls, id):
+        """Check if a task should be cancelled based on its document status.
+    
+        This method determines whether a task should be cancelled by checking the
+        associated document's run status and progress. A task should be cancelled
+        if its document is marked for cancellation or has negative progress.
+    
+        Args:
+            id (str): The unique identifier of the task to check.
+    
+        Returns:
+            bool: True if the task should be cancelled, False otherwise.
+        """
         task = cls.model.get_by_id(id)
         _, doc = DocumentService.get_by_id(task.doc_id)
         return doc.run == TaskStatus.CANCEL.value or doc.progress < 0
@@ -179,6 +255,18 @@ class TaskService(CommonService):
     @classmethod
     @DB.connection_context()
     def update_progress(cls, id, info):
+        """Update the progress information for a task.
+    
+        This method updates both the progress message and completion percentage of a task.
+        It handles platform-specific behavior (macOS vs others) and uses database locking
+        when necessary to ensure thread safety.
+    
+        Args:
+            id (str): The unique identifier of the task to update.
+            info (dict): Dictionary containing progress information with keys:
+                        - progress_msg (str, optional): Progress message to append
+                        - progress (float, optional): Progress percentage (0.0 to 1.0)
+        """
         if os.environ.get("MACOS"):
             if info["progress_msg"]:
                 task = cls.model.get_by_id(id)
@@ -201,7 +289,26 @@ class TaskService(CommonService):
                 ).execute()
 
 
-def queue_tasks(doc: dict, bucket: str, name: str):
+def queue_tasks(doc: dict, bucket: str, name: str, priority: int):
+    """Create and queue document processing tasks.
+    
+    This function creates processing tasks for a document based on its type and configuration.
+    It handles different document types (PDF, Excel, etc.) differently and manages task
+    chunking and configuration. It also implements task reuse optimization by checking
+    for previously completed tasks.
+    
+    Args:
+        doc (dict): Document dictionary containing metadata and configuration.
+        bucket (str): Storage bucket name where the document is stored.
+        name (str): File name of the document.
+        priority (int, optional): Priority level for task queueing (default is 0).
+    
+    Note:
+        - For PDF documents, tasks are created per page range based on configuration
+        - For Excel documents, tasks are created per row range
+        - Task digests are calculated for optimization and reuse
+        - Previous task chunks may be reused if available
+    """
     def new_task():
         return {"id": get_uuid(), "doc_id": doc["id"], "progress": 0.0, "from_page": 0, "to_page": 100000000}
 
@@ -252,6 +359,7 @@ def queue_tasks(doc: dict, bucket: str, name: str):
         task_digest = hasher.hexdigest()
         task["digest"] = task_digest
         task["progress"] = 0.0
+        task["priority"] = priority
 
     prev_tasks = TaskService.get_tasks(doc["id"])
     ck_num = 0
@@ -274,11 +382,31 @@ def queue_tasks(doc: dict, bucket: str, name: str):
     unfinished_task_array = [task for task in parse_task_array if task["progress"] < 1.0]
     for unfinished_task in unfinished_task_array:
         assert REDIS_CONN.queue_product(
-            SVR_QUEUE_NAME, message=unfinished_task
+            get_svr_queue_name(priority), message=unfinished_task
         ), "Can't access Redis. Please check the Redis' status."
 
 
 def reuse_prev_task_chunks(task: dict, prev_tasks: list[dict], chunking_config: dict):
+    """Attempt to reuse chunks from previous tasks for optimization.
+    
+    This function checks if chunks from previously completed tasks can be reused for
+    the current task, which can significantly improve processing efficiency. It matches
+    tasks based on page ranges and configuration digests.
+    
+    Args:
+        task (dict): Current task dictionary to potentially reuse chunks for.
+        prev_tasks (list[dict]): List of previous task dictionaries to check for reuse.
+        chunking_config (dict): Configuration dictionary for chunk processing.
+    
+    Returns:
+        int: Number of chunks successfully reused. Returns 0 if no chunks could be reused.
+    
+    Note:
+        Chunks can only be reused if:
+        - A previous task exists with matching page range and configuration digest
+        - The previous task was completed successfully (progress = 1.0)
+        - The previous task has valid chunk IDs
+    """
     idx = 0
     while idx < len(prev_tasks):
         prev_task = prev_tasks[idx]

api/db/services/user_canvas_version.py

@@ -0,0 +1,43 @@
+from api.db.db_models import UserCanvasVersion, DB
+from api.db.services.common_service import CommonService
+from peewee import DoesNotExist
+
+class UserCanvasVersionService(CommonService):
+    model = UserCanvasVersion
+    
+    
+    @classmethod
+    @DB.connection_context()
+    def list_by_canvas_id(cls, user_canvas_id):
+        try:
+            user_canvas_version = cls.model.select(
+                *[cls.model.id, 
+                cls.model.create_time,
+                cls.model.title,
+                cls.model.create_date, 
+                cls.model.update_date,
+                cls.model.user_canvas_id, 
+                cls.model.update_time]
+            ).where(cls.model.user_canvas_id == user_canvas_id)
+            return user_canvas_version
+        except DoesNotExist:
+            return None
+        except Exception:
+            return None
+    
+    @classmethod
+    @DB.connection_context()
+    def delete_all_versions(cls, user_canvas_id):
+        try:
+            user_canvas_version = cls.model.select().where(cls.model.user_canvas_id == user_canvas_id).order_by(cls.model.create_time.desc())
+            if user_canvas_version.count() > 20:
+                for i in range(20, user_canvas_version.count()):
+                    cls.delete(user_canvas_version[i].id)
+            return True
+        except DoesNotExist:
+            return None
+        except Exception:
+            return None
+
+
+

api/db/services/user_service.py

@@ -29,11 +29,27 @@ from rag.settings import MINIO
 
 
 class UserService(CommonService):
+    """Service class for managing user-related database operations.
+
+    This class extends CommonService to provide specialized functionality for user management,
+    including authentication, user creation, updates, and deletions.
+
+    Attributes:
+        model: The User model class for database operations.
+    """
     model = User
 
     @classmethod
     @DB.connection_context()
     def filter_by_id(cls, user_id):
+        """Retrieve a user by their ID.
+
+        Args:
+            user_id: The unique identifier of the user.
+
+        Returns:
+            User object if found, None otherwise.
+        """
         try:
             user = cls.model.select().where(cls.model.id == user_id).get()
             return user
@@ -43,6 +59,15 @@ class UserService(CommonService):
     @classmethod
     @DB.connection_context()
     def query_user(cls, email, password):
+        """Authenticate a user with email and password.
+
+        Args:
+            email: User's email address.
+            password: User's password in plain text.
+
+        Returns:
+            User object if authentication successful, None otherwise.
+        """
         user = cls.model.select().where((cls.model.email == email),
                                         (cls.model.status == StatusEnum.VALID.value)).first()
         if user and check_password_hash(str(user.password), password):
@@ -85,6 +110,14 @@ class UserService(CommonService):
 
 
 class TenantService(CommonService):
+    """Service class for managing tenant-related database operations.
+
+    This class extends CommonService to provide functionality for tenant management,
+    including tenant information retrieval and credit management.
+
+    Attributes:
+        model: The Tenant model class for database operations.
+    """
     model = Tenant
 
     @classmethod
@@ -136,8 +169,25 @@ class TenantService(CommonService):
 
 
 class UserTenantService(CommonService):
+    """Service class for managing user-tenant relationship operations.
+
+    This class extends CommonService to handle the many-to-many relationship
+    between users and tenants, managing user roles and tenant memberships.
+
+    Attributes:
+        model: The UserTenant model class for database operations.
+    """
     model = UserTenant
 
+    @classmethod
+    @DB.connection_context()
+    def filter_by_id(cls, user_tenant_id):
+        try:
+            user_tenant = cls.model.select().where((cls.model.id == user_tenant_id) & (cls.model.status == StatusEnum.VALID.value)).get()
+            return user_tenant
+        except peewee.DoesNotExist:
+            return None
+
     @classmethod
     @DB.connection_context()
     def save(cls, **kwargs):
@@ -150,6 +200,7 @@ class UserTenantService(CommonService):
     @DB.connection_context()
     def get_by_tenant_id(cls, tenant_id):
         fields = [
+            cls.model.id,
             cls.model.user_id,
             cls.model.status,
             cls.model.role,
@@ -181,3 +232,21 @@ class UserTenantService(CommonService):
         return list(cls.model.select(*fields)
                     .join(User, on=((cls.model.tenant_id == User.id) & (UserTenant.user_id == user_id) & (UserTenant.status == StatusEnum.VALID.value)))
                     .where(cls.model.status == StatusEnum.VALID.value).dicts())
+
+    @classmethod
+    @DB.connection_context()
+    def get_num_members(cls, user_id: str):
+        cnt_members = cls.model.select(peewee.fn.COUNT(cls.model.id)).where(cls.model.tenant_id == user_id).scalar()
+        return cnt_members
+
+    @classmethod
+    @DB.connection_context()
+    def filter_by_tenant_and_user_id(cls, tenant_id, user_id):
+        try:
+            user_tenant = cls.model.select().where(
+                (cls.model.tenant_id == tenant_id) & (cls.model.status == StatusEnum.VALID.value) &
+                (cls.model.user_id == user_id)
+            ).first()
+            return user_tenant
+        except peewee.DoesNotExist:
+            return None

api/ragflow_server.py

@@ -29,6 +29,7 @@ import time
 import traceback
 from concurrent.futures import ThreadPoolExecutor
 import threading
+import uuid
 
 from werkzeug.serving import run_simple
 from api import settings
@@ -46,13 +47,17 @@ from rag.utils.redis_conn import RedisDistributedLock
 
 stop_event = threading.Event()
 
+RAGFLOW_DEBUGPY_LISTEN = int(os.environ.get('RAGFLOW_DEBUGPY_LISTEN', "0"))
+
 def update_progress():
-    redis_lock = RedisDistributedLock("update_progress", timeout=60)
+    lock_value = str(uuid.uuid4())
+    redis_lock = RedisDistributedLock("update_progress", lock_value=lock_value, timeout=60)
+    logging.info(f"update_progress lock_value: {lock_value}")
     while not stop_event.is_set():
         try:
-            if not redis_lock.acquire():
-                continue
-            DocumentService.update_progress()
+            if redis_lock.acquire():
+                DocumentService.update_progress()
+                redis_lock.release()
             stop_event.wait(6)
         except Exception:
             logging.exception("update_progress exception")
@@ -84,6 +89,11 @@ if __name__ == '__main__':
     settings.init_settings()
     print_rag_settings()
 
+    if RAGFLOW_DEBUGPY_LISTEN > 0:
+        logging.info(f"debugpy listen on {RAGFLOW_DEBUGPY_LISTEN}")
+        import debugpy
+        debugpy.listen(("0.0.0.0", RAGFLOW_DEBUGPY_LISTEN))
+
     # init db
     init_web_db()
     init_web_data()

api/settings.py

@@ -16,6 +16,7 @@
 import os
 from datetime import date
 from enum import IntEnum, Enum
+import json
 import rag.utils.es_conn
 import rag.utils.infinity_conn
 
@@ -24,6 +25,7 @@ from rag.nlp import search
 from graphrag import search as kg_search
 from api.utils import get_base_config, decrypt_database_config
 from api.constants import RAG_FLOW_SERVICE_NAME
+from api.utils.file_utils import get_project_base_directory
 
 LIGHTEN = int(os.environ.get('LIGHTEN', "0"))
 
@@ -40,6 +42,7 @@ PARSERS = None
 HOST_IP = None
 HOST_PORT = None
 SECRET_KEY = None
+FACTORY_LLM_INFOS = None
 
 DATABASE_TYPE = os.getenv("DB_TYPE", 'mysql')
 DATABASE = decrypt_database_config(name=DATABASE_TYPE)
@@ -59,9 +62,12 @@ docStoreConn = None
 retrievaler = None
 kg_retrievaler = None
 
+# user registration switch
+REGISTER_ENABLED = 1
+
 
 def init_settings():
-    global LLM, LLM_FACTORY, LLM_BASE_URL, LIGHTEN, DATABASE_TYPE, DATABASE
+    global LLM, LLM_FACTORY, LLM_BASE_URL, LIGHTEN, DATABASE_TYPE, DATABASE, FACTORY_LLM_INFOS, REGISTER_ENABLED
     LIGHTEN = int(os.environ.get('LIGHTEN', "0"))
     DATABASE_TYPE = os.getenv("DB_TYPE", 'mysql')
     DATABASE = decrypt_database_config(name=DATABASE_TYPE)
@@ -69,6 +75,16 @@ def init_settings():
     LLM_DEFAULT_MODELS = LLM.get("default_models", {})
     LLM_FACTORY = LLM.get("factory", "Tongyi-Qianwen")
     LLM_BASE_URL = LLM.get("base_url")
+    try:
+        REGISTER_ENABLED = int(os.environ.get("REGISTER_ENABLED", "1"))
+    except Exception:
+        pass  
+    
+    try:
+        with open(os.path.join(get_project_base_directory(), "conf", "llm_factories.json"), "r") as f:
+            FACTORY_LLM_INFOS = json.load(f)["factory_llm_infos"]
+    except Exception:
+        FACTORY_LLM_INFOS = []
 
     global CHAT_MDL, EMBEDDING_MDL, RERANK_MDL, ASR_MDL, IMAGE2TEXT_MDL
     if not LIGHTEN:
@@ -93,7 +109,7 @@ def init_settings():
     API_KEY = LLM.get("api_key", "")
     PARSERS = LLM.get(
         "parsers",
-        "naive:General,qa:Q&A,resume:Resume,manual:Manual,table:Table,paper:Paper,book:Book,laws:Laws,presentation:Presentation,picture:Picture,one:One,audio:Audio,knowledge_graph:Knowledge Graph,email:Email,tag:Tag")
+        "naive:General,qa:Q&A,resume:Resume,manual:Manual,table:Table,paper:Paper,book:Book,laws:Laws,presentation:Presentation,picture:Picture,one:One,audio:Audio,email:Email,tag:Tag")
 
     HOST_IP = get_base_config(RAG_FLOW_SERVICE_NAME, {}).get("host", "127.0.0.1")
     HOST_PORT = get_base_config(RAG_FLOW_SERVICE_NAME, {}).get("http_port")

api/utils/api_utils.py

@@ -13,9 +13,9 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-import logging
 import functools
 import json
+import logging
 import random
 import time
 from base64 import b64encode
@@ -27,59 +27,60 @@ from uuid import uuid1
 
 import requests
 from flask import (
-    Response, jsonify, send_file, make_response,
+    Response,
+    jsonify,
+    make_response,
+    send_file,
+)
+from flask import (
     request as flask_request,
 )
 from itsdangerous import URLSafeTimedSerializer
 from werkzeug.http import HTTP_STATUS_CODES
 
-from api.db.db_models import APIToken
 from api import settings
+from api.constants import REQUEST_MAX_WAIT_SEC, REQUEST_WAIT_SEC
+from api.db.db_models import APIToken
+from api.utils import CustomJSONEncoder, get_uuid, json_dumps
 
-from api.utils import CustomJSONEncoder, get_uuid
-from api.utils import json_dumps
-from api.constants import REQUEST_WAIT_SEC, REQUEST_MAX_WAIT_SEC
-
-requests.models.complexjson.dumps = functools.partial(
-    json.dumps, cls=CustomJSONEncoder)
+requests.models.complexjson.dumps = functools.partial(json.dumps, cls=CustomJSONEncoder)
 
 
 def request(**kwargs):
     sess = requests.Session()
-    stream = kwargs.pop('stream', sess.stream)
-    timeout = kwargs.pop('timeout', None)
-    kwargs['headers'] = {
-        k.replace(
-            '_',
-            '-').upper(): v for k,
-                                v in kwargs.get(
-            'headers',
-            {}).items()}
+    stream = kwargs.pop("stream", sess.stream)
+    timeout = kwargs.pop("timeout", None)
+    kwargs["headers"] = {k.replace("_", "-").upper(): v for k, v in kwargs.get("headers", {}).items()}
     prepped = requests.Request(**kwargs).prepare()
 
     if settings.CLIENT_AUTHENTICATION and settings.HTTP_APP_KEY and settings.SECRET_KEY:
         timestamp = str(round(time() * 1000))
         nonce = str(uuid1())
-        signature = b64encode(HMAC(settings.SECRET_KEY.encode('ascii'), b'\n'.join([
-            timestamp.encode('ascii'),
-            nonce.encode('ascii'),
-            settings.HTTP_APP_KEY.encode('ascii'),
-            prepped.path_url.encode('ascii'),
-            prepped.body if kwargs.get('json') else b'',
-            urlencode(
-                sorted(
-                    kwargs['data'].items()),
-                quote_via=quote,
-                safe='-._~').encode('ascii')
-            if kwargs.get('data') and isinstance(kwargs['data'], dict) else b'',
-        ]), 'sha1').digest()).decode('ascii')
+        signature = b64encode(
+            HMAC(
+                settings.SECRET_KEY.encode("ascii"),
+                b"\n".join(
+                    [
+                        timestamp.encode("ascii"),
+                        nonce.encode("ascii"),
+                        settings.HTTP_APP_KEY.encode("ascii"),
+                        prepped.path_url.encode("ascii"),
+                        prepped.body if kwargs.get("json") else b"",
+                        urlencode(sorted(kwargs["data"].items()), quote_via=quote, safe="-._~").encode("ascii") if kwargs.get("data") and isinstance(kwargs["data"], dict) else b"",
+                    ]
+                ),
+                "sha1",
+            ).digest()
+        ).decode("ascii")
 
-        prepped.headers.update({
-            'TIMESTAMP': timestamp,
-            'NONCE': nonce,
-            'APP-KEY': settings.HTTP_APP_KEY,
-            'SIGNATURE': signature,
-        })
+        prepped.headers.update(
+            {
+                "TIMESTAMP": timestamp,
+                "NONCE": nonce,
+                "APP-KEY": settings.HTTP_APP_KEY,
+                "SIGNATURE": signature,
+            }
+        )
 
     return sess.send(prepped, stream=stream, timeout=timeout)
 
@@ -87,7 +88,7 @@ def request(**kwargs):
 def get_exponential_backoff_interval(retries, full_jitter=False):
     """Calculate the exponential backoff wait time."""
     # Will be zero if factor equals 0
-    countdown = min(REQUEST_MAX_WAIT_SEC, REQUEST_WAIT_SEC * (2 ** retries))
+    countdown = min(REQUEST_MAX_WAIT_SEC, REQUEST_WAIT_SEC * (2**retries))
     # Full jitter according to
     # https://aws.amazon.com/blogs/architecture/exponential-backoff-and-jitter/
     if full_jitter:
@@ -96,12 +97,9 @@ def get_exponential_backoff_interval(retries, full_jitter=False):
     return max(0, countdown)
 
 
-def get_data_error_result(code=settings.RetCode.DATA_ERROR,
-                          message='Sorry! Data missing!'):
+def get_data_error_result(code=settings.RetCode.DATA_ERROR, message="Sorry! Data missing!"):
     logging.exception(Exception(message))
-    result_dict = {
-        "code": code,
-        "message": message}
+    result_dict = {"code": code, "message": message}
     response = {}
     for key, value in result_dict.items():
         if value is None and key != "code":
@@ -119,23 +117,27 @@ def server_error_response(e):
     except BaseException:
         pass
     if len(e.args) > 1:
-        return get_json_result(
-            code=settings.RetCode.EXCEPTION_ERROR, message=repr(e.args[0]), data=e.args[1])
+        return get_json_result(code=settings.RetCode.EXCEPTION_ERROR, message=repr(e.args[0]), data=e.args[1])
     if repr(e).find("index_not_found_exception") >= 0:
-        return get_json_result(code=settings.RetCode.EXCEPTION_ERROR,
-                               message="No chunk found, please upload file and parse it.")
+        return get_json_result(code=settings.RetCode.EXCEPTION_ERROR, message="No chunk found, please upload file and parse it.")
 
     return get_json_result(code=settings.RetCode.EXCEPTION_ERROR, message=repr(e))
 
 
 def error_response(response_code, message=None):
     if message is None:
-        message = HTTP_STATUS_CODES.get(response_code, 'Unknown Error')
+        message = HTTP_STATUS_CODES.get(response_code, "Unknown Error")
 
-    return Response(json.dumps({
-        'message': message,
-        'code': response_code,
-    }), status=response_code, mimetype='application/json')
+    return Response(
+        json.dumps(
+            {
+                "message": message,
+                "code": response_code,
+            }
+        ),
+        status=response_code,
+        mimetype="application/json",
+    )
 
 
 def validate_request(*args, **kwargs):
@@ -160,13 +162,10 @@ def validate_request(*args, **kwargs):
             if no_arguments or error_arguments:
                 error_string = ""
                 if no_arguments:
-                    error_string += "required argument are missing: {}; ".format(
-                        ",".join(no_arguments))
+                    error_string += "required argument are missing: {}; ".format(",".join(no_arguments))
                 if error_arguments:
-                    error_string += "required argument values: {}".format(
-                        ",".join(["{}={}".format(a[0], a[1]) for a in error_arguments]))
-                return get_json_result(
-                    code=settings.RetCode.ARGUMENT_ERROR, message=error_string)
+                    error_string += "required argument values: {}".format(",".join(["{}={}".format(a[0], a[1]) for a in error_arguments]))
+                return get_json_result(code=settings.RetCode.ARGUMENT_ERROR, message=error_string)
             return func(*_args, **_kwargs)
 
         return decorated_function
@@ -180,8 +179,7 @@ def not_allowed_parameters(*params):
             input_arguments = flask_request.json or flask_request.form.to_dict()
             for param in params:
                 if param in input_arguments:
-                    return get_json_result(
-                        code=settings.RetCode.ARGUMENT_ERROR, message=f"Parameter {param} isn't allowed")
+                    return get_json_result(code=settings.RetCode.ARGUMENT_ERROR, message=f"Parameter {param} isn't allowed")
             return f(*args, **kwargs)
 
         return wrapper
@@ -190,14 +188,14 @@ def not_allowed_parameters(*params):
 
 
 def is_localhost(ip):
-    return ip in {'127.0.0.1', '::1', '[::1]', 'localhost'}
+    return ip in {"127.0.0.1", "::1", "[::1]", "localhost"}
 
 
 def send_file_in_mem(data, filename):
     if not isinstance(data, (str, bytes)):
         data = json_dumps(data)
     if isinstance(data, str):
-        data = data.encode('utf-8')
+        data = data.encode("utf-8")
 
     f = BytesIO()
     f.write(data)
@@ -206,7 +204,7 @@ def send_file_in_mem(data, filename):
     return send_file(f, as_attachment=True, attachment_filename=filename)
 
 
-def get_json_result(code=settings.RetCode.SUCCESS, message='success', data=None):
+def get_json_result(code=settings.RetCode.SUCCESS, message="success", data=None):
     response = {"code": code, "message": message, "data": data}
     return jsonify(response)
 
@@ -214,27 +212,24 @@ def get_json_result(code=settings.RetCode.SUCCESS, message='success', data=None)
 def apikey_required(func):
     @wraps(func)
     def decorated_function(*args, **kwargs):
-        token = flask_request.headers.get('Authorization').split()[1]
+        token = flask_request.headers.get("Authorization").split()[1]
         objs = APIToken.query(token=token)
         if not objs:
-            return build_error_result(
-                message='API-KEY is invalid!', code=settings.RetCode.FORBIDDEN
-            )
-        kwargs['tenant_id'] = objs[0].tenant_id
+            return build_error_result(message="API-KEY is invalid!", code=settings.RetCode.FORBIDDEN)
+        kwargs["tenant_id"] = objs[0].tenant_id
         return func(*args, **kwargs)
 
     return decorated_function
 
 
-def build_error_result(code=settings.RetCode.FORBIDDEN, message='success'):
+def build_error_result(code=settings.RetCode.FORBIDDEN, message="success"):
     response = {"code": code, "message": message}
     response = jsonify(response)
     response.status_code = code
     return response
 
 
-def construct_response(code=settings.RetCode.SUCCESS,
-                       message='success', data=None, auth=None):
+def construct_response(code=settings.RetCode.SUCCESS, message="success", data=None, auth=None):
     result_dict = {"code": code, "message": message, "data": data}
     response_dict = {}
     for key, value in result_dict.items():
@@ -253,7 +248,7 @@ def construct_response(code=settings.RetCode.SUCCESS,
     return response
 
 
-def construct_result(code=settings.RetCode.DATA_ERROR, message='data is missing'):
+def construct_result(code=settings.RetCode.DATA_ERROR, message="data is missing"):
     result_dict = {"code": code, "message": message}
     response = {}
     for key, value in result_dict.items():
@@ -264,7 +259,7 @@ def construct_result(code=settings.RetCode.DATA_ERROR, message='data is missing'
     return jsonify(response)
 
 
-def construct_json_result(code=settings.RetCode.SUCCESS, message='success', data=None):
+def construct_json_result(code=settings.RetCode.SUCCESS, message="success", data=None):
     if data is None:
         return jsonify({"code": code, "message": message})
     else:
@@ -286,7 +281,7 @@ def construct_error_response(e):
 def token_required(func):
     @wraps(func)
     def decorated_function(*args, **kwargs):
-        authorization_str = flask_request.headers.get('Authorization')
+        authorization_str = flask_request.headers.get("Authorization")
         if not authorization_str:
             return get_json_result(data=False, message="`Authorization` can't be empty")
         authorization_list = authorization_str.split()
@@ -295,11 +290,8 @@ def token_required(func):
         token = authorization_list[1]
         objs = APIToken.query(token=token)
         if not objs:
-            return get_json_result(
-                data=False, message='Authentication error: API key is invalid!',
-                code=settings.RetCode.AUTHENTICATION_ERROR
-            )
-        kwargs['tenant_id'] = objs[0].tenant_id
+            return get_json_result(data=False, message="Authentication error: API key is invalid!", code=settings.RetCode.AUTHENTICATION_ERROR)
+        kwargs["tenant_id"] = objs[0].tenant_id
         return func(*args, **kwargs)
 
     return decorated_function
@@ -316,11 +308,11 @@ def get_result(code=settings.RetCode.SUCCESS, message="", data=None):
     return jsonify(response)
 
 
-def get_error_data_result(message='Sorry! Data missing!', code=settings.RetCode.DATA_ERROR,
-                          ):
-    result_dict = {
-        "code": code,
-        "message": message}
+def get_error_data_result(
+    message="Sorry! Data missing!",
+    code=settings.RetCode.DATA_ERROR,
+):
+    result_dict = {"code": code, "message": message}
     response = {}
     for key, value in result_dict.items():
         if value is None and key != "code":
@@ -347,14 +339,17 @@ def valid_parameter(parameter, valid_values):
         return get_error_data_result(f"'{parameter}' is not in {valid_values}")
 
 
+def dataset_readonly_fields(field_name):
+    return field_name in ["chunk_count", "create_date", "create_time", "update_date", "update_time", "created_by", "document_count", "token_num", "status", "tenant_id", "id"]
+
+
 def get_parser_config(chunk_method, parser_config):
     if parser_config:
         return parser_config
     if not chunk_method:
         chunk_method = "naive"
     key_mapping = {
-        "naive": {"chunk_token_num": 128, "delimiter": "\\n!?;。；！？", "html4excel": False, "layout_recognize": "DeepDOC",
-                  "raptor": {"use_raptor": False}},
+        "naive": {"chunk_token_num": 128, "delimiter": "\\n!?;。；！？", "html4excel": False, "layout_recognize": "DeepDOC", "raptor": {"use_raptor": False}},
         "qa": {"raptor": {"use_raptor": False}},
         "tag": None,
         "resume": None,
@@ -365,38 +360,115 @@ def get_parser_config(chunk_method, parser_config):
         "laws": {"raptor": {"use_raptor": False}},
         "presentation": {"raptor": {"use_raptor": False}},
         "one": None,
-        "knowledge_graph": {"chunk_token_num": 8192, "delimiter": "\\n!?;。；！？",
-                            "entity_types": ["organization", "person", "location", "event", "time"]},
+        "knowledge_graph": {"chunk_token_num": 8192, "delimiter": "\\n!?;。；！？", "entity_types": ["organization", "person", "location", "event", "time"]},
         "email": None,
-        "picture": None}
+        "picture": None,
+    }
     parser_config = key_mapping[chunk_method]
     return parser_config
 
 
+def get_data_openai(id=None, 
+                    created=None, 
+                    model=None, 
+                    prompt_tokens= 0, 
+                    completion_tokens=0, 
+                    content = None, 
+                    finish_reason= None,
+                    object="chat.completion",
+                    param=None,
+):
+   
+    total_tokens= prompt_tokens + completion_tokens
+    return {
+        "id":f"{id}",
+        "object": object,
+        "created": int(time.time()) if created else None,
+        "model": model,
+        "param":param,
+        "usage": {
+            "prompt_tokens": prompt_tokens,
+            "completion_tokens": completion_tokens,
+            "total_tokens": total_tokens,
+            "completion_tokens_details": {
+                "reasoning_tokens": 0,
+                "accepted_prediction_tokens": 0,
+                "rejected_prediction_tokens": 0
+            }
+        },
+        "choices": [
+            {
+                "message": {
+                    "role": "assistant",
+                    "content": content
+                },
+                "logprobs": None,
+                "finish_reason": finish_reason,
+                "index": 0
+            }
+        ]
+    } 
 def valid_parser_config(parser_config):
     if not parser_config:
         return
-    scopes = set([
-        "chunk_token_num",
-        "delimiter",
-        "raptor",
-        "graphrag",
-        "layout_recognize",
-        "task_page_size",
-        "pages",
-        "html4excel",
-        "auto_keywords",
-        "auto_questions",
-        "tag_kb_ids",
-        "topn_tags",
-        "filename_embd_weight"
-    ])
+    scopes = set(
+        [
+            "chunk_token_num",
+            "delimiter",
+            "raptor",
+            "graphrag",
+            "layout_recognize",
+            "task_page_size",
+            "pages",
+            "html4excel",
+            "auto_keywords",
+            "auto_questions",
+            "tag_kb_ids",
+            "topn_tags",
+            "filename_embd_weight",
+        ]
+    )
     for k in parser_config.keys():
         assert k in scopes, f"Abnormal 'parser_config'. Invalid key: {k}"
 
+    assert isinstance(parser_config.get("chunk_token_num", 1), int), "chunk_token_num should be int"
     assert 1 <= parser_config.get("chunk_token_num", 1) < 100000000, "chunk_token_num should be in range from 1 to 100000000"
+    assert isinstance(parser_config.get("task_page_size", 1), int), "task_page_size should be int"
     assert 1 <= parser_config.get("task_page_size", 1) < 100000000, "task_page_size should be in range from 1 to 100000000"
+    assert isinstance(parser_config.get("auto_keywords", 1), int), "auto_keywords should be int"
     assert 0 <= parser_config.get("auto_keywords", 0) < 32, "auto_keywords should be in range from 0 to 32"
+    assert isinstance(parser_config.get("auto_questions", 1), int), "auto_questions should be int"
     assert 0 <= parser_config.get("auto_questions", 0) < 10, "auto_questions should be in range from 0 to 10"
+    assert isinstance(parser_config.get("topn_tags", 1), int), "topn_tags should be int"
     assert 0 <= parser_config.get("topn_tags", 0) < 10, "topn_tags should be in range from 0 to 10"
     assert isinstance(parser_config.get("html4excel", False), bool), "html4excel should be True or False"
+    assert isinstance(parser_config.get("delimiter", ""), str), "delimiter should be str"
+
+
+def check_duplicate_ids(ids, id_type="item"):
+    """
+    Check for duplicate IDs in a list and return unique IDs and error messages.
+
+    Args:
+        ids (list): List of IDs to check for duplicates
+        id_type (str): Type of ID for error messages (e.g., 'document', 'dataset', 'chunk')
+
+    Returns:
+        tuple: (unique_ids, error_messages)
+            - unique_ids (list): List of unique IDs
+            - error_messages (list): List of error messages for duplicate IDs
+    """
+    id_count = {}
+    duplicate_messages = []
+
+    # Count occurrences of each ID
+    for id_value in ids:
+        id_count[id_value] = id_count.get(id_value, 0) + 1
+
+    # Check for duplicates
+    for id_value, count in id_count.items():
+        if count > 1:
+            duplicate_messages.append(f"Duplicate {id_type} ids: {id_value}")
+
+    # Return unique IDs and error messages
+    return list(set(ids)), duplicate_messages

conf/infinity_mapping.json

@@ -5,10 +5,10 @@
 	"create_time": {"type": "varchar", "default": ""},
 	"create_timestamp_flt": {"type": "float", "default": 0.0},
 	"img_id": {"type": "varchar", "default": ""},
-	"docnm_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
+	"docnm_kwd": {"type": "varchar", "default": ""},
 	"title_tks": {"type": "varchar", "default": "", "analyzer": "whitespace"},
 	"title_sm_tks": {"type": "varchar", "default": "", "analyzer": "whitespace"},
-	"name_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
+	"name_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"important_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"tag_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"important_tks": {"type": "varchar", "default": "", "analyzer": "whitespace"},
@@ -27,16 +27,16 @@
 	"rank_int": {"type": "integer", "default": 0},
 	"rank_flt": {"type": "float", "default": 0},
 	"available_int": {"type": "integer", "default": 1},
-	"knowledge_graph_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
+	"knowledge_graph_kwd": {"type": "varchar", "default": ""},
 	"entities_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"pagerank_fea": {"type": "integer", "default":  0},
 	"tag_feas": {"type": "varchar", "default":  ""},
 
-	"from_entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
-	"to_entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
-	"entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
-	"entity_type_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"},
-	"source_id": {"type": "varchar", "default": ""},
+	"from_entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
+	"to_entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
+	"entity_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
+	"entity_type_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
+	"source_id": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"n_hop_with_weight": {"type": "varchar", "default": ""},
-	"removed_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace"}
+	"removed_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"}
 }

deepdoc/parser/excel_parser.py

@@ -45,7 +45,7 @@ class RAGFlowExcelParser:
                 raise Exception(f"****wxy: Failed to parse CSV and convert to Excel Workbook: {e_csv}")
 
         try:
-            return load_workbook(file_like_object)
+            return load_workbook(file_like_object,data_only= True)
         except Exception as e:
             logging.info(f"****wxy: openpyxl load error: {e}, try pandas instead")
             try:

deepdoc/parser/figure_parser.py

@@ -0,0 +1,91 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+
+from PIL import Image
+
+from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
+from rag.prompts import vision_llm_figure_describe_prompt
+
+
+def vision_figure_parser_figure_data_wraper(figures_data_without_positions):
+    return [(
+        (figure_data[1], [figure_data[0]]),
+        [(0, 0, 0, 0, 0)]
+    ) for figure_data in figures_data_without_positions if isinstance(figure_data[1], Image.Image)]
+
+
+class VisionFigureParser:
+    def __init__(self, vision_model, figures_data, *args, **kwargs):
+        self.vision_model = vision_model
+        self._extract_figures_info(figures_data)
+        assert len(self.figures) == len(self.descriptions)
+        assert not self.positions or (len(self.figures) == len(self.positions))
+
+    def _extract_figures_info(self, figures_data):
+        self.figures = []
+        self.descriptions = []
+        self.positions = []
+
+        for item in figures_data:
+            # position
+            if len(item) == 2 and isinstance(item[1], list) and len(item[1]) == 1 and isinstance(item[1][0], tuple) and len(item[1][0]) == 5:
+                img_desc = item[0]
+                assert len(img_desc) == 2 and isinstance(img_desc[0], Image.Image) and isinstance(img_desc[1], list), "Should be (figure, [description])"
+                self.figures.append(img_desc[0])
+                self.descriptions.append(img_desc[1])
+                self.positions.append(item[1])
+            else:
+                assert len(item) == 2 and isinstance(item, tuple) and isinstance(item[1], list), f"get {len(item)=}, {item=}"
+                self.figures.append(item[0])
+                self.descriptions.append(item[1])
+
+    def _assemble(self):
+        self.assembled = []
+        self.has_positions = len(self.positions) != 0
+        for i in range(len(self.figures)):
+            figure = self.figures[i]
+            desc = self.descriptions[i]
+            pos = self.positions[i] if self.has_positions else None
+
+            figure_desc = (figure, desc)
+
+            if pos is not None:
+                self.assembled.append((figure_desc, pos))
+            else:
+                self.assembled.append((figure_desc,))
+
+        return self.assembled
+
+    def __call__(self, **kwargs):
+        callback = kwargs.get("callback", lambda prog, msg: None)
+
+        for idx, img_binary in enumerate(self.figures or []):
+            figure_num = idx  # 0-based
+
+            txt = picture_vision_llm_chunk(
+                binary=img_binary,
+                vision_model=self.vision_model,
+                prompt=vision_llm_figure_describe_prompt(),
+                callback=callback,
+            )
+
+            if txt:
+                self.descriptions[figure_num] = txt + "\n".join(self.descriptions[figure_num])
+
+        self._assemble()
+
+        return self.assembled

deepdoc/parser/pdf_parser.py

@@ -17,31 +17,36 @@
 import logging
 import os
 import random
-from timeit import default_timer as timer
+import re
 import sys
 import threading
-
-import xgboost as xgb
+from copy import deepcopy
 from io import BytesIO
-import re
-import pdfplumber
-from PIL import Image
+from timeit import default_timer as timer
+
 import numpy as np
+import pdfplumber
+import trio
+import xgboost as xgb
+from huggingface_hub import snapshot_download
+from PIL import Image
 from pypdf import PdfReader as pdf2_read
 
 from api import settings
 from api.utils.file_utils import get_project_base_directory
-from deepdoc.vision import OCR, Recognizer, LayoutRecognizer, TableStructureRecognizer
+from deepdoc.vision import OCR, LayoutRecognizer, Recognizer, TableStructureRecognizer
+from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
 from rag.nlp import rag_tokenizer
-from copy import deepcopy
-from huggingface_hub import snapshot_download
+from rag.prompts import vision_llm_describe_prompt
+from rag.settings import PARALLEL_DEVICES
 
 LOCK_KEY_pdfplumber = "global_shared_lock_pdfplumber"
 if LOCK_KEY_pdfplumber not in sys.modules:
     sys.modules[LOCK_KEY_pdfplumber] = threading.Lock()
 
+
 class RAGFlowPdfParser:
-    def __init__(self):
+    def __init__(self, **kwargs):
         """
         If you have trouble downloading HuggingFace models, -_^ this might help!!
 
@@ -53,7 +58,12 @@ class RAGFlowPdfParser:
         ^_-
 
         """
+
         self.ocr = OCR()
+        self.parallel_limiter = None
+        if PARALLEL_DEVICES is not None and PARALLEL_DEVICES > 1:
+            self.parallel_limiter = [trio.CapacityLimiter(1) for _ in range(PARALLEL_DEVICES)]
+
         if hasattr(self, "model_speciess"):
             self.layouter = LayoutRecognizer("layout." + self.model_speciess)
         else:
@@ -63,7 +73,7 @@ class RAGFlowPdfParser:
         self.updown_cnt_mdl = xgb.Booster()
         if not settings.LIGHTEN:
             try:
-                import torch
+                import torch.cuda
                 if torch.cuda.is_available():
                     self.updown_cnt_mdl.set_param({"device": "cuda"})
             except Exception:
@@ -97,7 +107,7 @@ class RAGFlowPdfParser:
     def _y_dis(
             self, a, b):
         return (
-                       b["top"] + b["bottom"] - a["top"] - a["bottom"]) / 2
+            b["top"] + b["bottom"] - a["top"] - a["bottom"]) / 2
 
     def _match_proj(self, b):
         proj_patt = [
@@ -120,9 +130,9 @@ class RAGFlowPdfParser:
         tks_down = rag_tokenizer.tokenize(down["text"][:LEN]).split()
         tks_up = rag_tokenizer.tokenize(up["text"][-LEN:]).split()
         tks_all = up["text"][-LEN:].strip() \
-                  + (" " if re.match(r"[a-zA-Z0-9]+",
-                                     up["text"][-1] + down["text"][0]) else "") \
-                  + down["text"][:LEN].strip()
+            + (" " if re.match(r"[a-zA-Z0-9]+",
+                               up["text"][-1] + down["text"][0]) else "") \
+            + down["text"][:LEN].strip()
         tks_all = rag_tokenizer.tokenize(tks_all).split()
         fea = [
             up.get("R", -1) == down.get("R", -1),
@@ -144,7 +154,7 @@ class RAGFlowPdfParser:
             True if re.search(r"[，,][^。.]+$", up["text"]) else False,
             True if re.search(r"[，,][^。.]+$", up["text"]) else False,
             True if re.search(r"[\(（][^\)）]+$", up["text"])
-                    and re.search(r"[\)）]", down["text"]) else False,
+            and re.search(r"[\)）]", down["text"]) else False,
             self._match_proj(down),
             True if re.match(r"[A-Z]", down["text"]) else False,
             True if re.match(r"[A-Z]", up["text"][-1]) else False,
@@ -206,7 +216,7 @@ class RAGFlowPdfParser:
                 continue
             for tb in tbls:  # for table
                 left, top, right, bott = tb["x0"] - MARGIN, tb["top"] - MARGIN, \
-                                         tb["x1"] + MARGIN, tb["bottom"] + MARGIN
+                    tb["x1"] + MARGIN, tb["bottom"] + MARGIN
                 left *= ZM
                 top *= ZM
                 right *= ZM
@@ -283,9 +293,9 @@ class RAGFlowPdfParser:
                 b["H_right"] = spans[ii]["x1"]
                 b["SP"] = ii
 
-    def __ocr(self, pagenum, img, chars, ZM=3):
+    def __ocr(self, pagenum, img, chars, ZM=3, device_id: int | None = None):
         start = timer()
-        bxs = self.ocr.detect(np.array(img))
+        bxs = self.ocr.detect(np.array(img), device_id)
         logging.info(f"__ocr detecting boxes of a image cost ({timer() - start}s)")
 
         start = timer()
@@ -330,7 +340,7 @@ class RAGFlowPdfParser:
                 b["box_image"] = self.ocr.get_rotate_crop_image(img_np, np.array([[left, top], [right, top], [right, bott], [left, bott]], dtype=np.float32))
                 boxes_to_reg.append(b)
             del b["txt"]
-        texts = self.ocr.recognize_batch([b["box_image"] for b in boxes_to_reg])
+        texts = self.ocr.recognize_batch([b["box_image"] for b in boxes_to_reg], device_id)
         for i in range(len(boxes_to_reg)):
             boxes_to_reg[i]["text"] = texts[i]
             del boxes_to_reg[i]["box_image"]
@@ -448,7 +458,7 @@ class RAGFlowPdfParser:
                     b_["text"],
                     any(feats),
                     any(concatting_feats),
-                    ))
+                ))
                 i += 1
                 continue
             # merge up and down
@@ -643,8 +653,7 @@ class RAGFlowPdfParser:
             b_["top"] = b["top"]
             self.boxes.pop(i)
 
-    def _extract_table_figure(self, need_image, ZM,
-                              return_html, need_position):
+    def _extract_table_figure(self, need_image, ZM, return_html, need_position, separate_tables_figures=False):
         tables = {}
         figures = {}
         # extract figure and table boxes
@@ -656,7 +665,7 @@ class RAGFlowPdfParser:
                 i += 1
                 continue
             lout_no = str(self.boxes[i]["page_number"]) + \
-                      "-" + str(self.boxes[i]["layoutno"])
+                "-" + str(self.boxes[i]["layoutno"])
             if TableStructureRecognizer.is_caption(self.boxes[i]) or self.boxes[i]["layout_type"] in ["table caption",
                                                                                                       "title",
                                                                                                       "figure caption",
@@ -758,9 +767,6 @@ class RAGFlowPdfParser:
                     tk)
             self.boxes.pop(i)
 
-        res = []
-        positions = []
-
         def cropout(bxs, ltype, poss):
             nonlocal ZM
             pn = set([b["page_number"] - 1 for b in bxs])
@@ -808,6 +814,10 @@ class RAGFlowPdfParser:
                 height += img.size[1]
             return pic
 
+        res = []
+        positions = []
+        figure_results = []
+        figure_positions = []
         # crop figure out and add caption
         for k, bxs in figures.items():
             txt = "\n".join([b["text"] for b in bxs])
@@ -815,28 +825,46 @@ class RAGFlowPdfParser:
                 continue
 
             poss = []
-            res.append(
-                (cropout(
-                    bxs,
-                    "figure", poss),
-                 [txt]))
-            positions.append(poss)
+
+            if separate_tables_figures:
+                figure_results.append(
+                    (cropout(
+                        bxs,
+                        "figure", poss),
+                     [txt]))
+                figure_positions.append(poss)
+            else:
+                res.append(
+                    (cropout(
+                        bxs,
+                        "figure", poss),
+                     [txt]))
+                positions.append(poss)
 
         for k, bxs in tables.items():
             if not bxs:
                 continue
             bxs = Recognizer.sort_Y_firstly(bxs, np.mean(
                 [(b["bottom"] - b["top"]) / 2 for b in bxs]))
+
             poss = []
+
             res.append((cropout(bxs, "table", poss),
                         self.tbl_det.construct_table(bxs, html=return_html, is_english=self.is_english)))
             positions.append(poss)
 
-        assert len(positions) == len(res)
-
-        if need_position:
-            return list(zip(res, positions))
-        return res
+        if separate_tables_figures:
+            assert len(positions) + len(figure_positions) == len(res) + len(figure_results)
+            if need_position:
+                return list(zip(res, positions)), list(zip(figure_results, figure_positions))
+            else:
+                return res, figure_results
+        else:
+            assert len(positions) == len(res)
+            if need_position:
+                return list(zip(res, positions))
+            else:
+                return res
 
     def proj_match(self, line):
         if len(line) <= 2:
@@ -976,62 +1004,56 @@ class RAGFlowPdfParser:
         start = timer()
         try:
             with sys.modules[LOCK_KEY_pdfplumber]:
-                self.pdf = pdfplumber.open(fnm) if isinstance(
-                    fnm, str) else pdfplumber.open(BytesIO(fnm))
-                self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
-                                    enumerate(self.pdf.pages[page_from:page_to])]
-                try:
-                    self.page_chars = [[c for c in page.dedupe_chars().chars if self._has_color(c)] for page in self.pdf.pages[page_from:page_to]]
-                except Exception as e:
-                    logging.warning(f"Failed to extract characters for pages {page_from}-{page_to}: {str(e)}")
-                    self.page_chars = [[] for _ in range(page_to - page_from)]  # If failed to extract, using empty list instead.
+                with (pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm))) as pdf:
+                    self.pdf = pdf
+                    self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
+                                        enumerate(self.pdf.pages[page_from:page_to])]
+
+                    try:
+                        self.page_chars = [[c for c in page.dedupe_chars().chars if self._has_color(c)] for page in self.pdf.pages[page_from:page_to]]
+                    except Exception as e:
+                        logging.warning(f"Failed to extract characters for pages {page_from}-{page_to}: {str(e)}")
+                        self.page_chars = [[] for _ in range(page_to - page_from)]  # If failed to extract, using empty list instead.
+
+                    self.total_page = len(self.pdf.pages)
 
-                self.total_page = len(self.pdf.pages)
         except Exception:
             logging.exception("RAGFlowPdfParser __images__")
         logging.info(f"__images__ dedupe_chars cost {timer() - start}s")
 
         self.outlines = []
         try:
-            self.pdf = pdf2_read(fnm if isinstance(fnm, str) else BytesIO(fnm))
-            outlines = self.pdf.outline
+            with (pdf2_read(fnm if isinstance(fnm, str)
+                            else BytesIO(fnm))) as pdf:
+                self.pdf = pdf
 
-            def dfs(arr, depth):
-                for a in arr:
-                    if isinstance(a, dict):
-                        self.outlines.append((a["/Title"], depth))
-                        continue
-                    dfs(a, depth + 1)
+                outlines = self.pdf.outline
+                def dfs(arr, depth):
+                    for a in arr:
+                        if isinstance(a, dict):
+                            self.outlines.append((a["/Title"], depth))
+                            continue
+                        dfs(a, depth + 1)
+
+                dfs(outlines, 0)
 
-            dfs(outlines, 0)
         except Exception as e:
             logging.warning(f"Outlines exception: {e}")
-        finally:
-            self.pdf.close()
+
         if not self.outlines:
             logging.warning("Miss outlines")
 
-
         logging.debug("Images converted.")
         self.is_english = [re.search(r"[a-zA-Z0-9,/¸;:'\[\]\(\)!@#$%^&*\"?<>._-]{30,}", "".join(
             random.choices([c["text"] for c in self.page_chars[i]], k=min(100, len(self.page_chars[i]))))) for i in
-                           range(len(self.page_chars))]
+            range(len(self.page_chars))]
         if sum([1 if e else 0 for e in self.is_english]) > len(
                 self.page_images) / 2:
             self.is_english = True
         else:
             self.is_english = False
 
-        start = timer()
-        for i, img in enumerate(self.page_images):
-            chars = self.page_chars[i] if not self.is_english else []
-            self.mean_height.append(
-                np.median(sorted([c["height"] for c in chars])) if chars else 0
-            )
-            self.mean_width.append(
-                np.median(sorted([c["width"] for c in chars])) if chars else 8
-            )
-            self.page_cum_height.append(img.size[1] / zoomin)
+        async def __img_ocr(i, id, img, chars, limiter):
             j = 0
             while j + 1 < len(chars):
                 if chars[j]["text"] and chars[j + 1]["text"] \
@@ -1041,9 +1063,44 @@ class RAGFlowPdfParser:
                     chars[j]["text"] += " "
                 j += 1
 
-            self.__ocr(i + 1, img, chars, zoomin)
+            if limiter:
+                async with limiter:
+                    await trio.to_thread.run_sync(lambda: self.__ocr(i + 1, img, chars, zoomin, id))
+            else:
+                self.__ocr(i + 1, img, chars, zoomin, id)
+
             if callback and i % 6 == 5:
                 callback(prog=(i + 1) * 0.6 / len(self.page_images), msg="")
+
+        async def __img_ocr_launcher():
+            def __ocr_preprocess():
+                chars = self.page_chars[i] if not self.is_english else []
+                self.mean_height.append(
+                    np.median(sorted([c["height"] for c in chars])) if chars else 0
+                )
+                self.mean_width.append(
+                    np.median(sorted([c["width"] for c in chars])) if chars else 8
+                )
+                self.page_cum_height.append(img.size[1] / zoomin)
+                return chars
+
+            if self.parallel_limiter:
+                async with trio.open_nursery() as nursery:
+                    for i, img in enumerate(self.page_images):
+                        chars = __ocr_preprocess()
+
+                        nursery.start_soon(__img_ocr, i, i % PARALLEL_DEVICES, img, chars,
+                                           self.parallel_limiter[i % PARALLEL_DEVICES])
+                        await trio.sleep(0.1)
+            else:
+                for i, img in enumerate(self.page_images):
+                    chars = __ocr_preprocess()
+                    await __img_ocr(i, 0, img, chars, None)
+
+        start = timer()
+
+        trio.run(__img_ocr_launcher)
+
         logging.info(f"__images__ {len(self.page_images)} pages cost {timer() - start}s")
 
         if not self.is_english and not any(
@@ -1108,7 +1165,7 @@ class RAGFlowPdfParser:
                 self.page_images[pns[0]].crop((left * ZM, top * ZM,
                                                right *
                                                ZM, min(
-                    bottom, self.page_images[pns[0]].size[1])
+                                                   bottom, self.page_images[pns[0]].size[1])
                                                ))
             )
             if 0 < ii < len(poss) - 1:
@@ -1206,5 +1263,52 @@ class PlainParser:
         raise NotImplementedError
 
 
+class VisionParser(RAGFlowPdfParser):
+    def __init__(self, vision_model, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.vision_model = vision_model
+
+    def __images__(self, fnm, zoomin=3, page_from=0, page_to=299, callback=None):
+        try:
+            with sys.modules[LOCK_KEY_pdfplumber]:
+                self.pdf = pdfplumber.open(fnm) if isinstance(
+                    fnm, str) else pdfplumber.open(BytesIO(fnm))
+                self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
+                                    enumerate(self.pdf.pages[page_from:page_to])]
+                self.total_page = len(self.pdf.pages)
+        except Exception:
+            self.page_images = None
+            self.total_page = 0
+            logging.exception("VisionParser __images__")
+
+    def __call__(self, filename, from_page=0, to_page=100000, **kwargs):
+        callback = kwargs.get("callback", lambda prog, msg: None)
+
+        self.__images__(fnm=filename, zoomin=3, page_from=from_page, page_to=to_page, **kwargs)
+
+        total_pdf_pages = self.total_page
+
+        start_page = max(0, from_page)
+        end_page = min(to_page, total_pdf_pages)
+
+        all_docs = []
+
+        for idx, img_binary in enumerate(self.page_images or []):
+            pdf_page_num = idx  # 0-based
+            if pdf_page_num < start_page or pdf_page_num >= end_page:
+                continue
+
+            docs = picture_vision_llm_chunk(
+                binary=img_binary,
+                vision_model=self.vision_model,
+                prompt=vision_llm_describe_prompt(page=pdf_page_num+1),
+                callback=callback,
+            )
+
+            if docs:
+                all_docs.append(docs)
+        return [(doc, "") for doc in all_docs], []
+
+
 if __name__ == "__main__":
     pass

deepdoc/parser/ppt_parser.py

@@ -23,25 +23,56 @@ class RAGFlowPptParser:
     def __init__(self):
         super().__init__()
 
+    def __get_bulleted_text(self, paragraph):
+        is_bulleted = bool(paragraph._p.xpath("./a:pPr/a:buChar")) or bool(paragraph._p.xpath("./a:pPr/a:buAutoNum")) or bool(paragraph._p.xpath("./a:pPr/a:buBlip"))
+        if is_bulleted:
+            return f"{'  '* paragraph.level}.{paragraph.text}"
+        else:
+            return paragraph.text
+
     def __extract(self, shape):
-        if shape.shape_type == 19:
-            tb = shape.table
-            rows = []
-            for i in range(1, len(tb.rows)):
-                rows.append("; ".join([tb.cell(
-                    0, j).text + ": " + tb.cell(i, j).text for j in range(len(tb.columns)) if tb.cell(i, j)]))
-            return "\n".join(rows)
+        try:
+            # First try to get text content
+            if hasattr(shape, 'has_text_frame') and shape.has_text_frame:
+                text_frame = shape.text_frame
+                texts = []
+                for paragraph in text_frame.paragraphs:
+                    if paragraph.text.strip():
+                        texts.append(self.__get_bulleted_text(paragraph))
+                return "\n".join(texts)
 
-        if shape.has_text_frame:
-            return shape.text_frame.text
+            # Safely get shape_type
+            try:
+                shape_type = shape.shape_type
+            except NotImplementedError:
+                # If shape_type is not available, try to get text content
+                if hasattr(shape, 'text'):
+                    return shape.text.strip()
+                return ""
 
-        if shape.shape_type == 6:
-            texts = []
-            for p in sorted(shape.shapes, key=lambda x: (x.top // 10, x.left)):
-                t = self.__extract(p)
-                if t:
-                    texts.append(t)
-            return "\n".join(texts)
+            # Handle table
+            if shape_type == 19:
+                tb = shape.table
+                rows = []
+                for i in range(1, len(tb.rows)):
+                    rows.append("; ".join([tb.cell(
+                        0, j).text + ": " + tb.cell(i, j).text for j in range(len(tb.columns)) if tb.cell(i, j)]))
+                return "\n".join(rows)
+
+            # Handle group shape
+            if shape_type == 6:
+                texts = []
+                for p in sorted(shape.shapes, key=lambda x: (x.top // 10, x.left)):
+                    t = self.__extract_texts(p)
+                    if t:
+                        texts.append(t)
+                return "\n".join(texts)
+
+            return ""
+
+        except Exception as e:
+            logging.error(f"Error processing shape: {str(e)}")
+            return ""
 
     def __call__(self, fnm, from_page, to_page, callback=None):
         ppt = Presentation(fnm) if isinstance(

deepdoc/vision/layout_recognizer.py

@@ -46,8 +46,8 @@ class LayoutRecognizer(Recognizer):
     def __init__(self, domain):
         try:
             model_dir = os.path.join(
-                    get_project_base_directory(),
-                    "rag/res/deepdoc")
+                get_project_base_directory(),
+                "rag/res/deepdoc")
             super().__init__(self.labels, domain, model_dir)
         except Exception:
             model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc",
@@ -56,18 +56,23 @@ class LayoutRecognizer(Recognizer):
             super().__init__(self.labels, domain, model_dir)
 
         self.garbage_layouts = ["footer", "header", "reference"]
+        self.client = None
+        if os.environ.get("TENSORRT_DLA_SVR"):
+            from deepdoc.vision.dla_cli import DLAClient
+            self.client = DLAClient(os.environ["TENSORRT_DLA_SVR"])
 
-    def __call__(self, image_list, ocr_res, scale_factor=3,
-                 thr=0.2, batch_size=16, drop=True):
+    def __call__(self, image_list, ocr_res, scale_factor=3, thr=0.2, batch_size=16, drop=True):
         def __is_garbage(b):
-            patt = [r"^•+$", r"(版权归©|免责条款|地址[:：])", r"\.{3,}", "^[0-9]{1,2} / ?[0-9]{1,2}$",
+            patt = [r"^•+$", "^[0-9]{1,2} / ?[0-9]{1,2}$",
                     r"^[0-9]{1,2} of [0-9]{1,2}$", "^http://[^ ]{12,}",
-                    "(资料|数据)来源[:：]", "[0-9a-z._-]+@[a-z0-9-]+\\.[a-z]{2,3}",
                     "\\(cid *: *[0-9]+ *\\)"
                     ]
             return any([re.search(p, b["text"]) for p in patt])
 
-        layouts = super().__call__(image_list, thr, batch_size)
+        if self.client:
+            layouts = self.client.predict(image_list)
+        else:
+            layouts = super().__call__(image_list, thr, batch_size)
         # save_results(image_list, layouts, self.labels, output_dir='output/', threshold=0.7)
         assert len(image_list) == len(ocr_res)
         # Tag layout type
@@ -160,6 +165,7 @@ class LayoutRecognizer(Recognizer):
     def forward(self, image_list, thr=0.7, batch_size=16):
         return super().__call__(image_list, thr, batch_size)
 
+
 class LayoutRecognizer4YOLOv10(LayoutRecognizer):
     labels = [
         "title",
@@ -185,9 +191,9 @@ class LayoutRecognizer4YOLOv10(LayoutRecognizer):
 
     def preprocess(self, image_list):
         inputs = []
-        new_shape = self.input_shape # height, width
+        new_shape = self.input_shape  # height, width
         for img in image_list:
-            shape = img.shape[:2]# current shape [height, width]
+            shape = img.shape[:2]  # current shape [height, width]
             # Scale ratio (new / old)
             r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
             # Compute padding
@@ -242,4 +248,3 @@ class LayoutRecognizer4YOLOv10(LayoutRecognizer):
             "bbox": [float(t) for t in boxes[i].tolist()],
             "score": float(scores[i])
         } for i in indices]
-

deepdoc/vision/ocr.py

@@ -22,6 +22,7 @@ import os
 from huggingface_hub import snapshot_download
 
 from api.utils.file_utils import get_project_base_directory
+from rag.settings import PARALLEL_DEVICES
 from .operators import *  # noqa: F403
 from . import operators
 import math
@@ -66,10 +67,12 @@ def create_operators(op_param_list, global_config=None):
     return ops
 
 
-def load_model(model_dir, nm):
+def load_model(model_dir, nm, device_id: int | None = None):
     model_file_path = os.path.join(model_dir, nm + ".onnx")
+    model_cached_tag = model_file_path + str(device_id) if device_id is not None else model_file_path
+
     global loaded_models
-    loaded_model = loaded_models.get(model_file_path)
+    loaded_model = loaded_models.get(model_cached_tag)
     if loaded_model:
         logging.info(f"load_model {model_file_path} reuses cached model")
         return loaded_model
@@ -81,7 +84,7 @@ def load_model(model_dir, nm):
     def cuda_is_available():
         try:
             import torch
-            if torch.cuda.is_available():
+            if torch.cuda.is_available() and torch.cuda.device_count() > device_id:
                 return True
         except Exception:
             return False
@@ -98,7 +101,7 @@ def load_model(model_dir, nm):
     run_options = ort.RunOptions()
     if cuda_is_available():
         cuda_provider_options = {
-            "device_id": 0, # Use specific GPU
+            "device_id": device_id, # Use specific GPU
             "gpu_mem_limit": 512 * 1024 * 1024, # Limit gpu memory
             "arena_extend_strategy": "kNextPowerOfTwo",  # gpu memory allocation strategy
         }
@@ -108,7 +111,7 @@ def load_model(model_dir, nm):
             providers=['CUDAExecutionProvider'],
             provider_options=[cuda_provider_options]
             )
-        run_options.add_run_config_entry("memory.enable_memory_arena_shrinkage", "gpu:0")
+        run_options.add_run_config_entry("memory.enable_memory_arena_shrinkage", "gpu:" + str(device_id))
         logging.info(f"load_model {model_file_path} uses GPU")
     else:
         sess = ort.InferenceSession(
@@ -118,12 +121,12 @@ def load_model(model_dir, nm):
         run_options.add_run_config_entry("memory.enable_memory_arena_shrinkage", "cpu")
         logging.info(f"load_model {model_file_path} uses CPU")
     loaded_model = (sess, run_options)
-    loaded_models[model_file_path] = loaded_model
+    loaded_models[model_cached_tag] = loaded_model
     return loaded_model
 
 
 class TextRecognizer:
-    def __init__(self, model_dir):
+    def __init__(self, model_dir, device_id: int | None = None):
         self.rec_image_shape = [int(v) for v in "3, 48, 320".split(",")]
         self.rec_batch_num = 16
         postprocess_params = {
@@ -132,7 +135,7 @@ class TextRecognizer:
             "use_space_char": True
         }
         self.postprocess_op = build_post_process(postprocess_params)
-        self.predictor, self.run_options = load_model(model_dir, 'rec')
+        self.predictor, self.run_options = load_model(model_dir, 'rec', device_id)
         self.input_tensor = self.predictor.get_inputs()[0]
 
     def resize_norm_img(self, img, max_wh_ratio):
@@ -394,7 +397,7 @@ class TextRecognizer:
 
 
 class TextDetector:
-    def __init__(self, model_dir):
+    def __init__(self, model_dir, device_id: int | None = None):
         pre_process_list = [{
             'DetResizeForTest': {
                 'limit_side_len': 960,
@@ -418,7 +421,7 @@ class TextDetector:
                               "unclip_ratio": 1.5, "use_dilation": False, "score_mode": "fast", "box_type": "quad"}
 
         self.postprocess_op = build_post_process(postprocess_params)
-        self.predictor, self.run_options = load_model(model_dir, 'det')
+        self.predictor, self.run_options = load_model(model_dir, 'det', device_id)
         self.input_tensor = self.predictor.get_inputs()[0]
 
         img_h, img_w = self.input_tensor.shape[2:]
@@ -524,14 +527,33 @@ class OCR:
                 model_dir = os.path.join(
                         get_project_base_directory(),
                         "rag/res/deepdoc")
-                self.text_detector = TextDetector(model_dir)
-                self.text_recognizer = TextRecognizer(model_dir)
+                
+                # Append muti-gpus task to the list
+                if PARALLEL_DEVICES is not None and PARALLEL_DEVICES > 0:
+                    self.text_detector = []
+                    self.text_recognizer = []
+                    for device_id in range(PARALLEL_DEVICES):
+                        self.text_detector.append(TextDetector(model_dir, device_id))
+                        self.text_recognizer.append(TextRecognizer(model_dir, device_id))
+                else:
+                    self.text_detector = [TextDetector(model_dir, 0)]
+                    self.text_recognizer = [TextRecognizer(model_dir, 0)]
+
             except Exception:
                 model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc",
                                               local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"),
                                               local_dir_use_symlinks=False)
-                self.text_detector = TextDetector(model_dir)
-                self.text_recognizer = TextRecognizer(model_dir)
+                
+                if PARALLEL_DEVICES is not None:
+                    assert PARALLEL_DEVICES > 0, "Number of devices must be >= 1"
+                    self.text_detector = []
+                    self.text_recognizer = []
+                    for device_id in range(PARALLEL_DEVICES):
+                        self.text_detector.append(TextDetector(model_dir, device_id))
+                        self.text_recognizer.append(TextRecognizer(model_dir, device_id))
+                else:
+                    self.text_detector = [TextDetector(model_dir, 0)]
+                    self.text_recognizer = [TextRecognizer(model_dir, 0)]
 
         self.drop_score = 0.5
         self.crop_image_res_index = 0
@@ -593,14 +615,17 @@ class OCR:
                     break
         return _boxes
 
-    def detect(self, img):
+    def detect(self, img, device_id: int | None = None):
+        if device_id is None:
+            device_id = 0
+
         time_dict = {'det': 0, 'rec': 0, 'cls': 0, 'all': 0}
 
         if img is None:
             return None, None, time_dict
 
         start = time.time()
-        dt_boxes, elapse = self.text_detector(img)
+        dt_boxes, elapse = self.text_detector[device_id](img)
         time_dict['det'] = elapse
 
         if dt_boxes is None:
@@ -611,17 +636,22 @@ class OCR:
         return zip(self.sorted_boxes(dt_boxes), [
                    ("", 0) for _ in range(len(dt_boxes))])
 
-    def recognize(self, ori_im, box):
+    def recognize(self, ori_im, box, device_id: int | None = None):
+        if device_id is None:
+            device_id = 0
+
         img_crop = self.get_rotate_crop_image(ori_im, box)
 
-        rec_res, elapse = self.text_recognizer([img_crop])
+        rec_res, elapse = self.text_recognizer[device_id]([img_crop])
         text, score = rec_res[0]
         if score < self.drop_score:
             return ""
         return text
 
-    def recognize_batch(self, img_list):
-        rec_res, elapse = self.text_recognizer(img_list)
+    def recognize_batch(self, img_list, device_id: int | None = None):
+        if device_id is None:
+            device_id = 0
+        rec_res, elapse = self.text_recognizer[device_id](img_list)
         texts = []
         for i in range(len(rec_res)):
             text, score = rec_res[i]
@@ -630,15 +660,17 @@ class OCR:
             texts.append(text)
         return texts
 
-    def __call__(self, img, cls=True):
+    def __call__(self, img, device_id = 0, cls=True):
         time_dict = {'det': 0, 'rec': 0, 'cls': 0, 'all': 0}
+        if device_id is None:
+            device_id = 0
 
         if img is None:
             return None, None, time_dict
 
         start = time.time()
         ori_im = img.copy()
-        dt_boxes, elapse = self.text_detector(img)
+        dt_boxes, elapse = self.text_detector[device_id](img)
         time_dict['det'] = elapse
 
         if dt_boxes is None:
@@ -655,7 +687,7 @@ class OCR:
             img_crop = self.get_rotate_crop_image(ori_im, tmp_box)
             img_crop_list.append(img_crop)
 
-        rec_res, elapse = self.text_recognizer(img_crop_list)
+        rec_res, elapse = self.text_recognizer[device_id](img_crop_list)
 
         time_dict['rec'] = elapse
 

deepdoc/vision/recognizer.py

@@ -195,9 +195,8 @@ class Recognizer:
                 (im_info[0]['scale_factor'],)).astype('float32')
             return inputs
         
-        for e in im_info:
-            im_shape.append(np.array((e['im_shape'],)).astype('float32'))
-            scale_factor.append(np.array((e['scale_factor'],)).astype('float32'))
+        im_shape = np.array([info['im_shape'] for info in im_info], dtype='float32')
+        scale_factor = np.array([info['scale_factor'] for info in im_info], dtype='float32')
 
         inputs['im_shape'] = np.concatenate(im_shape, axis=0)
         inputs['scale_factor'] = np.concatenate(scale_factor, axis=0)

deepdoc/vision/t_ocr.py

@@ -28,14 +28,24 @@ from deepdoc.vision.seeit import draw_box
 from deepdoc.vision import OCR, init_in_out
 import argparse
 import numpy as np
+import trio
+
+# os.environ['CUDA_VISIBLE_DEVICES'] = '0,2' #2 gpus, uncontinuous
+os.environ['CUDA_VISIBLE_DEVICES'] = '0' #1 gpu
+# os.environ['CUDA_VISIBLE_DEVICES'] = '' #cpu
 
 
 def main(args):
+    import torch.cuda
+
+    cuda_devices = torch.cuda.device_count()
+    limiter = [trio.CapacityLimiter(1) for _ in range(cuda_devices)] if cuda_devices > 1 else None
     ocr = OCR()
     images, outputs = init_in_out(args)
 
-    for i, img in enumerate(images):
-        bxs = ocr(np.array(img))
+    def __ocr(i, id, img):
+        print("Task {} start".format(i))
+        bxs = ocr(np.array(img), id)
         bxs = [(line[0], line[1][0]) for line in bxs]
         bxs = [{
             "text": t,
@@ -47,6 +57,30 @@ def main(args):
         with open(outputs[i] + ".txt", "w+", encoding='utf-8') as f:
             f.write("\n".join([o["text"] for o in bxs]))
 
+        print("Task {} done".format(i))
+
+    async def __ocr_thread(i, id, img, limiter = None):
+        if limiter:
+            async with limiter:
+                print("Task {} use device {}".format(i, id))
+                await trio.to_thread.run_sync(lambda: __ocr(i, id, img))
+        else:
+            __ocr(i, id, img)
+
+    async def __ocr_launcher():
+        if cuda_devices > 1:
+            async with trio.open_nursery() as nursery:
+                for i, img in enumerate(images):
+                    nursery.start_soon(__ocr_thread, i, i % cuda_devices, img, limiter[i % cuda_devices])
+                    await trio.sleep(0.1)
+        else:
+            for i, img in enumerate(images):
+                await __ocr_thread(i, 0, img)
+
+    trio.run(__ocr_launcher)
+
+    print("OCR tasks are all done")
+
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()

deepdoc/vision/table_structure_recognizer.py

@@ -133,7 +133,7 @@ class TableStructureRecognizer(Recognizer):
         return "Ot"
 
     @staticmethod
-    def construct_table(boxes, is_english=False, html=False):
+    def construct_table(boxes, is_english=False, html=True, **kwargs):
         cap = ""
         i = 0
         while i < len(boxes):

docker/.env

@@ -80,28 +80,16 @@ REDIS_PASSWORD=infini_rag_flow
 SVR_HTTP_PORT=9380
 
 # The RAGFlow Docker image to download.
-# Defaults to the v0.17.2-slim edition, which is the RAGFlow Docker image without embedding models.
-RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2-slim
+# Defaults to the v0.18.0-slim edition, which is the RAGFlow Docker image without embedding models.
+RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0-slim
 #
 # To download the RAGFlow Docker image with embedding models, uncomment the following line instead:
-# RAGFLOW_IMAGE=infiniflow/ragflow:v0.17.2
+# RAGFLOW_IMAGE=infiniflow/ragflow:v0.18.0
 # 
-# The Docker image of the v0.17.2 edition includes:
-# - Built-in embedding models:
+# The Docker image of the v0.18.0 edition includes built-in embedding models:
 #   - BAAI/bge-large-zh-v1.5
-#   - BAAI/bge-reranker-v2-m3
 #   - maidalun1020/bce-embedding-base_v1
-#   - maidalun1020/bce-reranker-base_v1
-# - Embedding models that will be downloaded once you select them in the RAGFlow UI:
-#   - BAAI/bge-base-en-v1.5
-#   - BAAI/bge-large-en-v1.5
-#   - BAAI/bge-small-en-v1.5
-#   - BAAI/bge-small-zh-v1.5
-#   - jinaai/jina-embeddings-v2-base-en
-#   - jinaai/jina-embeddings-v2-small-en
-#   - nomic-ai/nomic-embed-text-v1.5
-#   - sentence-transformers/all-MiniLM-L6-v2
-#
+
 # 
 
 
@@ -125,10 +113,12 @@ TIMEZONE='Asia/Shanghai'
 # Uncomment the following line if your operating system is MacOS:
 # MACOS=1
 
-# The maximum file size for each uploaded file, in bytes.
-# You can uncomment this line and update the value if you wish to change the 128M file size limit
-# MAX_CONTENT_LENGTH=134217728
-# After making the change, ensure you update `client_max_body_size` in nginx/nginx.conf correspondingly.
+# The maximum file size limit (in bytes) for each upload to your knowledge base or File Management.
+# To change the 1GB file size limit, uncomment the line below and update as needed.
+# MAX_CONTENT_LENGTH=1073741824
+# After updating, ensure `client_max_body_size` in nginx/nginx.conf is updated accordingly.
+# Note that neither `MAX_CONTENT_LENGTH` nor `client_max_body_size` sets the maximum size for files uploaded to an agent.
+# See https://ragflow.io/docs/dev/begin_component for details.
 
 # The log level for the RAGFlow's owned packages and imported packages.
 # Available level:
@@ -146,3 +136,6 @@ TIMEZONE='Asia/Shanghai'
 # ENDPOINT=http://oss-cn-hangzhou.aliyuncs.com
 # REGION=cn-hangzhou
 # BUCKET=ragflow65536
+
+# user registration switch
+REGISTER_ENABLED=1

docker/README.md

@@ -78,22 +78,12 @@ The [.env](./.env) file contains important environment variables for Docker.
 - `RAGFLOW-IMAGE`  
   The Docker image edition. Available editions:  
   
-  - `infiniflow/ragflow:v0.17.2-slim` (default): The RAGFlow Docker image without embedding models.  
-  - `infiniflow/ragflow:v0.17.2`: The RAGFlow Docker image with embedding models including:
+  - `infiniflow/ragflow:v0.18.0-slim` (default): The RAGFlow Docker image without embedding models.  
+  - `infiniflow/ragflow:v0.18.0`: The RAGFlow Docker image with embedding models including:
     - Built-in embedding models:
       - `BAAI/bge-large-zh-v1.5` 
-      - `BAAI/bge-reranker-v2-m3`
       - `maidalun1020/bce-embedding-base_v1`
-      - `maidalun1020/bce-reranker-base_v1`
-    - Embedding models that will be downloaded once you select them in the RAGFlow UI:
-      - `BAAI/bge-base-en-v1.5`
-      - `BAAI/bge-large-en-v1.5`
-      - `BAAI/bge-small-en-v1.5`
-      - `BAAI/bge-small-zh-v1.5`
-      - `jinaai/jina-embeddings-v2-base-en`
-      - `jinaai/jina-embeddings-v2-small-en`
-      - `nomic-ai/nomic-embed-text-v1.5`
-      - `sentence-transformers/all-MiniLM-L6-v2`
+
   
 > [!TIP]  
 > If you cannot download the RAGFlow Docker image, try the following mirrors.  
@@ -146,6 +136,24 @@ The [.env](./.env) file contains important environment variables for Docker.
   - `password`: The password for MinIO.
   - `host`: The MinIO serving IP *and* port inside the Docker container. Defaults to `minio:9000`.
 
+- `oss`
+  - `access_key`: The access key ID used to authenticate requests to the OSS service.
+  - `secret_key`: The secret access key used to authenticate requests to the OSS service.
+  - `endpoint_url`: The URL of the OSS service endpoint.
+  - `region`: The OSS region where the bucket is located.
+  - `bucket`: The name of the OSS bucket where files will be stored. When you want to store all files in a specified bucket, you need this configuration item.
+  - `prefix_path`: Optional. A prefix path to prepend to file names in the OSS bucket, which can help organize files within the bucket.
+
+- `s3`:
+  - `access_key`: The access key ID used to authenticate requests to the S3 service.
+  - `secret_key`: The secret access key used to authenticate requests to the S3 service.
+  - `endpoint_url`: The URL of the S3-compatible service endpoint. This is necessary when using an S3-compatible protocol instead of the default AWS S3 endpoint.
+  - `bucket`: The name of the S3 bucket where files will be stored. When you want to store all files in a specified bucket, you need this configuration item.
+  - `region`: The AWS region where the S3 bucket is located. This is important for directing requests to the correct data center.
+  - `signature_version`: Optional. The version of the signature to use for authenticating requests. Common versions include `v4`.
+  - `addressing_style`: Optional. The style of addressing to use for the S3 endpoint. This can be `path` or `virtual`.
+  - `prefix_path`: Optional. A prefix path to prepend to file names in the S3 bucket, which can help organize files within the bucket.
+
 - `oauth`  
   The OAuth configuration for signing up or signing in to RAGFlow using a third-party account.  It is disabled by default. To enable this feature, uncomment the corresponding lines in **service_conf.yaml.template**.
   - `github`: The GitHub authentication settings for your application. Visit the [Github Developer Settings page](https://github.com/settings/developers) to obtain your client_id and secret_key.

docker/docker-compose-base.yml

@@ -81,6 +81,7 @@ services:
       --default-authentication-plugin=mysql_native_password
       --tls_version="TLSv1.2,TLSv1.3"
       --init-file /data/application/init.sql
+      --binlog_expire_logs_seconds=604800
     ports:
       - ${MYSQL_PORT}:3306
     volumes:

docker/docker-compose.yml

@@ -1,22 +1,38 @@
 include:
   - ./docker-compose-base.yml
 
+# To ensure that the container processes the locally modified `service_conf.yaml.template` instead of the one included in its image, you need to mount the local `service_conf.yaml.template` to the container.
 services:
   ragflow:
     depends_on:
       mysql:
         condition: service_healthy
     image: ${RAGFLOW_IMAGE}
+    # example to setup MCP server
+    # command:
+    #   - --enable-mcpserver
+    #   - --mcp-host=0.0.0.0
+    #   - --mcp-port=9382
+    #   - --mcp-base-url=http://127.0.0.1:9380
+    #   - --mcp-script-path=/ragflow/mcp/server/server.py
+    #   - --mcp-mode=self-host
+    #   - --mcp--host-api-key="ragflow-xxxxxxx"
     container_name: ragflow-server
     ports:
       - ${SVR_HTTP_PORT}:9380
       - 80:80
       - 443:443
+      - 5678:5678
+      - 5679:5679
+      - 9382:9382 # entry for MCP (host_port:docker_port). The docker_port should match with the value you set for `mcp-port` above
     volumes:
       - ./ragflow-logs:/ragflow/logs
       - ./nginx/ragflow.conf:/etc/nginx/conf.d/ragflow.conf
       - ./nginx/proxy.conf:/etc/nginx/proxy.conf
       - ./nginx/nginx.conf:/etc/nginx/nginx.conf
+      - ../history_data_agent:/ragflow/history_data_agent
+      - ./service_conf.yaml.template:/ragflow/conf/service_conf.yaml.template
+
     env_file: .env
     environment:
       - TZ=${TIMEZONE}

docker/entrypoint-parser.sh

@@ -1,29 +0,0 @@
-#!/bin/bash
-
-# replace env variables in the service_conf.yaml file
-rm -rf /ragflow/conf/service_conf.yaml
-while IFS= read -r line || [[ -n "$line" ]]; do
-    # Use eval to interpret the variable with default values
-    eval "echo \"$line\"" >> /ragflow/conf/service_conf.yaml
-done < /ragflow/conf/service_conf.yaml.template
-
-export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/
-
-PY=python3
-
-CONSUMER_NO_BEG=$1
-CONSUMER_NO_END=$2
-
-function task_exe(){
-    JEMALLOC_PATH=$(pkg-config --variable=libdir jemalloc)/libjemalloc.so
-    while [ 1 -eq 1 ]; do
-      LD_PRELOAD=$JEMALLOC_PATH $PY rag/svr/task_executor.py $1;
-    done
-}
-
-for ((i=CONSUMER_NO_BEG; i<CONSUMER_NO_END; i++))
-do
-  task_exe $i &
-done
-
-wait;

docker/entrypoint.sh

@@ -1,35 +1,192 @@
-#!/bin/bash
+#!/usr/bin/env bash
 
-# replace env variables in the service_conf.yaml file
-rm -rf /ragflow/conf/service_conf.yaml
-while IFS= read -r line || [[ -n "$line" ]]; do
-    # Use eval to interpret the variable with default values
-    eval "echo \"$line\"" >> /ragflow/conf/service_conf.yaml
-done < /ragflow/conf/service_conf.yaml.template
+set -e
 
-/usr/sbin/nginx
+# -----------------------------------------------------------------------------
+# Usage and command-line argument parsing
+# -----------------------------------------------------------------------------
+function usage() {
+    echo "Usage: $0 [--disable-webserver] [--disable-taskexecutor] [--consumer-no-beg=<num>] [--consumer-no-end=<num>] [--workers=<num>] [--host-id=<string>]"
+    echo
+    echo "  --disable-webserver             Disables the web server (nginx + ragflow_server)."
+    echo "  --disable-taskexecutor          Disables task executor workers."
+    echo "  --enable-mcpserver              Enables the MCP server."
+    echo "  --consumer-no-beg=<num>         Start range for consumers (if using range-based)."
+    echo "  --consumer-no-end=<num>         End range for consumers (if using range-based)."
+    echo "  --workers=<num>                 Number of task executors to run (if range is not used)."
+    echo "  --host-id=<string>              Unique ID for the host (defaults to \`hostname\`)."
+    echo
+    echo "Examples:"
+    echo "  $0 --disable-taskexecutor"
+    echo "  $0 --disable-webserver --consumer-no-beg=0 --consumer-no-end=5"
+    echo "  $0 --disable-webserver --workers=2 --host-id=myhost123"
+    echo "  $0 --enable-mcpserver"
+    exit 1
+}
 
-export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/
+ENABLE_WEBSERVER=1 # Default to enable web server
+ENABLE_TASKEXECUTOR=1  # Default to enable task executor
+ENABLE_MCP_SERVER=0
+CONSUMER_NO_BEG=0
+CONSUMER_NO_END=0
+WORKERS=1
 
-PY=python3
-if [[ -z "$WS" || $WS -lt 1 ]]; then
-  WS=1
+MCP_HOST="127.0.0.1"
+MCP_PORT=9382
+MCP_BASE_URL="http://127.0.0.1:9380"
+MCP_SCRIPT_PATH="/ragflow/mcp/server/server.py"
+MCP_MODE="self-host"
+MCP_HOST_API_KEY=""
+
+# -----------------------------------------------------------------------------
+# Host ID logic:
+#   1. By default, use the system hostname if length <= 32
+#   2. Otherwise, use the full MD5 hash of the hostname (32 hex chars)
+# -----------------------------------------------------------------------------
+CURRENT_HOSTNAME="$(hostname)"
+if [ ${#CURRENT_HOSTNAME} -le 32 ]; then
+  DEFAULT_HOST_ID="$CURRENT_HOSTNAME"
+else
+  DEFAULT_HOST_ID="$(echo -n "$CURRENT_HOSTNAME" | md5sum | cut -d ' ' -f 1)"
 fi
 
-function task_exe(){
-    JEMALLOC_PATH=$(pkg-config --variable=libdir jemalloc)/libjemalloc.so
-    while [ 1 -eq 1 ];do
-      LD_PRELOAD=$JEMALLOC_PATH $PY rag/svr/task_executor.py $1;
+HOST_ID="$DEFAULT_HOST_ID"
+
+# Parse arguments
+for arg in "$@"; do
+  case $arg in
+    --disable-webserver)
+      ENABLE_WEBSERVER=0
+      shift
+      ;;
+    --disable-taskexecutor)
+      ENABLE_TASKEXECUTOR=0
+      shift
+      ;;
+    --enable-mcpserver)
+      ENABLE_MCP_SERVER=1
+      shift
+      ;;
+    --mcp-host=*)
+      MCP_HOST="${arg#*=}"
+      shift
+      ;;
+    --mcp-port=*)
+      MCP_PORT="${arg#*=}"
+      shift
+      ;;
+    --mcp-base-url=*)
+      MCP_BASE_URL="${arg#*=}"
+      shift
+      ;;
+    --mcp-mode=*)
+      MCP_MODE="${arg#*=}"
+      shift
+      ;;
+    --mcp-host-api-key=*)
+      MCP_HOST_API_KEY="${arg#*=}"
+      shift
+      ;;
+    --mcp-script-path=*)
+      MCP_SCRIPT_PATH="${arg#*=}"
+      shift
+      ;;
+    --consumer-no-beg=*)
+      CONSUMER_NO_BEG="${arg#*=}"
+      shift
+      ;;
+    --consumer-no-end=*)
+      CONSUMER_NO_END="${arg#*=}"
+      shift
+      ;;
+    --workers=*)
+      WORKERS="${arg#*=}"
+      shift
+      ;;
+    --host-id=*)
+      HOST_ID="${arg#*=}"
+      shift
+      ;;
+    *)
+      usage
+      ;;
+  esac
+done
+
+# -----------------------------------------------------------------------------
+# Replace env variables in the service_conf.yaml file
+# -----------------------------------------------------------------------------
+CONF_DIR="/ragflow/conf"
+TEMPLATE_FILE="${CONF_DIR}/service_conf.yaml.template"
+CONF_FILE="${CONF_DIR}/service_conf.yaml"
+
+rm -f "${CONF_FILE}"
+while IFS= read -r line || [[ -n "$line" ]]; do
+    eval "echo \"$line\"" >> "${CONF_FILE}"
+done < "${TEMPLATE_FILE}"
+
+export LD_LIBRARY_PATH="/usr/lib/x86_64-linux-gnu/"
+PY=python3
+
+# -----------------------------------------------------------------------------
+# Function(s)
+# -----------------------------------------------------------------------------
+
+function task_exe() {
+    local consumer_id="$1"
+    local host_id="$2"
+
+    JEMALLOC_PATH="$(pkg-config --variable=libdir jemalloc)/libjemalloc.so"
+    while true; do
+        LD_PRELOAD="$JEMALLOC_PATH" \
+        "$PY" rag/svr/task_executor.py "${host_id}_${consumer_id}"
     done
 }
 
-for ((i=0;i<WS;i++))
-do
-  task_exe  $i &
-done
+function start_mcp_server() {
+    echo "Starting MCP Server on ${MCP_HOST}:${MCP_PORT} with base URL ${MCP_BASE_URL}..."
+    "$PY" "${MCP_SCRIPT_PATH}" \
+        --host="${MCP_HOST}" \
+        --port="${MCP_PORT}" \
+        --base_url="${MCP_BASE_URL}" \
+        --mode="${MCP_MODE}" \
+        --api_key="${MCP_HOST_API_KEY}" \ &
+}
 
-while [ 1 -eq 1 ];do
-    $PY api/ragflow_server.py
-done
+# -----------------------------------------------------------------------------
+# Start components based on flags
+# -----------------------------------------------------------------------------
 
-wait;
+if [[ "${ENABLE_WEBSERVER}" -eq 1 ]]; then
+    echo "Starting nginx..."
+    /usr/sbin/nginx
+
+    echo "Starting ragflow_server..."
+    while true; do
+        "$PY" api/ragflow_server.py
+    done &
+fi
+
+
+if [[ "${ENABLE_MCP_SERVER}" -eq 1 ]]; then
+    start_mcp_server
+fi
+
+if [[ "${ENABLE_TASKEXECUTOR}" -eq 1 ]]; then
+    if [[ "${CONSUMER_NO_END}" -gt "${CONSUMER_NO_BEG}" ]]; then
+        echo "Starting task executors on host '${HOST_ID}' for IDs in [${CONSUMER_NO_BEG}, ${CONSUMER_NO_END})..."
+        for (( i=CONSUMER_NO_BEG; i<CONSUMER_NO_END; i++ ))
+        do
+          task_exe "${i}" "${HOST_ID}" &
+        done
+    else
+        # Otherwise, start a fixed number of workers
+        echo "Starting ${WORKERS} task executor(s) on host '${HOST_ID}'..."
+        for (( i=0; i<WORKERS; i++ ))
+        do
+          task_exe "${i}" "${HOST_ID}" &
+        done
+    fi
+fi
+
+wait

docker/launch_backend_service.sh

@@ -3,6 +3,27 @@
 # Exit immediately if a command exits with a non-zero status
 set -e
 
+# Function to load environment variables from .env file
+load_env_file() {
+    # Get the directory of the current script
+    local script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+    local env_file="$script_dir/.env"
+
+    # Check if .env file exists
+    if [ -f "$env_file" ]; then
+        echo "Loading environment variables from: $env_file"
+        # Source the .env file
+        set -a
+        source "$env_file" 
+        set +a
+    else
+        echo "Warning: .env file not found at: $env_file"
+    fi
+}
+
+# Load environment variables
+load_env_file
+
 # Unset HTTP proxies that might be set by Docker daemon
 export http_proxy=""; export https_proxy=""; export no_proxy=""; export HTTP_PROXY=""; export HTTPS_PROXY=""; export NO_PROXY=""
 export PYTHONPATH=$(pwd)

docker/nginx/nginx.conf

@@ -26,7 +26,7 @@ http {
     keepalive_timeout  65;
 
     #gzip  on;
-    client_max_body_size 128M;
+    client_max_body_size 1024M;
 
     include /etc/nginx/conf.d/ragflow.conf;
 }

docker/service_conf.yaml.template

@@ -37,6 +37,11 @@ redis:
 #   access_key: 'access_key'
 #   secret_key: 'secret_key'
 #   region: 'region'
+#   endpoint_url: 'endpoint_url'
+#   bucket: 'bucket'
+#   prefix_path: 'prefix_path'
+#   signature_version: 'v4'
+#   addressing_style: 'path'
 # oss:
 #   access_key: '${ACCESS_KEY}'
 #   secret_key: '${SECRET_KEY}'

docs/configurations.md

@@ -74,6 +74,8 @@ The [.env](https://github.com/infiniflow/ragflow/blob/main/docker/.env) file con
 
 ### MinIO
 
+RAGFlow utilizes MinIO as its object storage solution, leveraging its scalability to store and manage all uploaded files.
+
 - `MINIO_CONSOLE_PORT`  
   The port used to expose the MinIO console interface to the host machine, allowing **external** access to the web-based console running inside the Docker container. Defaults to `9001`
 - `MINIO_PORT`  
@@ -97,22 +99,12 @@ The [.env](https://github.com/infiniflow/ragflow/blob/main/docker/.env) file con
 - `RAGFLOW-IMAGE`  
   The Docker image edition. Available editions:  
   
-  - `infiniflow/ragflow:v0.17.2-slim` (default): The RAGFlow Docker image without embedding models.  
-  - `infiniflow/ragflow:v0.17.2`: The RAGFlow Docker image with embedding models including:
+  - `infiniflow/ragflow:v0.18.0-slim` (default): The RAGFlow Docker image without embedding models.  
+  - `infiniflow/ragflow:v0.18.0`: The RAGFlow Docker image with embedding models including:
     - Built-in embedding models:
       - `BAAI/bge-large-zh-v1.5` 
-      - `BAAI/bge-reranker-v2-m3`
       - `maidalun1020/bce-embedding-base_v1`
-      - `maidalun1020/bce-reranker-base_v1`
-    - Embedding models that will be downloaded once you select them in the RAGFlow UI:
-      - `BAAI/bge-base-en-v1.5`
-      - `BAAI/bge-large-en-v1.5`
-      - `BAAI/bge-small-en-v1.5`
-      - `BAAI/bge-small-zh-v1.5`
-      - `jinaai/jina-embeddings-v2-base-en`
-      - `jinaai/jina-embeddings-v2-small-en`
-      - `nomic-ai/nomic-embed-text-v1.5`
-      - `sentence-transformers/all-MiniLM-L6-v2`
+
 
 :::tip NOTE  
 If you cannot download the RAGFlow Docker image, try the following mirrors.  

docs/develop/build_docker_image.mdx

@@ -77,7 +77,7 @@ After building the infiniflow/ragflow:nightly-slim image, you are ready to launc
 
 1. Edit Docker Compose Configuration
 
-Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.17.2-slim` to `infiniflow/ragflow:nightly-slim` to use the pre-built image.
+Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.18.0-slim` to `infiniflow/ragflow:nightly-slim` to use the pre-built image.
 
 
 2. Launch the Service

docs/develop/launch_ragflow_from_source.md

@@ -91,10 +91,16 @@ docker compose -f docker/docker-compose-base.yml up -d
    export HF_ENDPOINT=https://hf-mirror.com
    ```
 
-4. Run the **entrypoint.sh** script to launch the backend service:
+4. Check the configuration in **conf/service_conf.yaml**, ensuring all hosts and ports are correctly set.
    
+5. Run the **entrypoint.sh** script to launch the backend service:
+
+   ```shell
+   JEMALLOC_PATH=$(pkg-config --variable=libdir jemalloc)/libjemalloc.so;
+   LD_PRELOAD=$JEMALLOC_PATH python rag/svr/task_executor.py 1;
    ```
-   bash docker/entrypoint.sh
+   ```shell
+   python api/ragflow_server.py;
    ```
 
 ### Launch the RAGFlow frontend service

docs/develop/mcp.md

@@ -0,0 +1,195 @@
+---
+sidebar_position: 4
+slug: /mcp_server
+---
+
+# RAGFlow MCP server overview
+
+The RAGFlow Model Context Protocol (MCP) server operates as an independent component that complements the RAGFlow server. However, it requires a RAGFlow server to work functionally well, meaning, the MCP client and server communicate with each other in MCP HTTP+SSE mode (once the connection is established, server pushes messages to client only), and responses are expected from RAGFlow server.
+
+The MCP server currently offers a specific tool to assist users in searching for relevant information powered by RAGFlow DeepDoc technology:
+
+- **retrieve**: Fetches relevant chunks from specified `dataset_ids` and optional `document_ids` using the RAGFlow retrieve interface, based on a given question. Details of all available datasets, namely, `id` and `description`, are provided within the tool description for each individual dataset.
+
+## Launching the MCP Server
+
+Similar to launching the RAGFlow server, the MCP server can be started either from source code or via Docker.
+
+### Launch Modes
+
+The MCP server supports two launch modes:
+
+1. **Self-Host Mode**:
+
+   - In this mode, the MCP server is launched to access a specific tenant's datasets.
+   - This is the default mode.
+   - The `--api_key` argument is **required** to authenticate the server with the RAGFlow server.
+   - Example:
+     ```bash
+     uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --mode=self-host --api_key=ragflow-xxxxx
+     ```
+
+1. **Host Mode**:
+
+   - In this mode, the MCP server allows each user to access their own datasets.
+   - To ensure secure access, a valid API key must be included in the request headers to identify the user.
+   - The `--api_key` argument is **not required** during server launch but must be provided in the headers on each client request for user authentication.
+   - Example:
+     ```bash
+     uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --mode=host
+     ```
+
+### Launching from Source Code
+
+All you need to do is stand on the right place and strike out command, assuming you are on the project working directory.
+
+```bash
+uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --api_key=ragflow-xxxxx
+```
+
+For testing purposes, there is an [MCP client example](#example_mcp_client) provided, free to take!
+
+#### Required Arguments
+
+- **`host`**: Specifies the server's host address.
+- **`port`**: Defines the server's listening port.
+- **`base_url`**: The address of the RAGFlow server that is already running and ready to handle tasks.
+- **`mode`**: Launch mode, only accept `self-host` or `host`.
+- **`api_key`**: Required when `mode` is `self-host` to authenticate the MCP server with the RAGFlow server.
+
+Here are three augments required, the first two,`host` and `port`, are self-explained. The`base_url` is the address of the ready-to-serve RAGFlow server to actually perform the task.
+
+### Launching from Docker
+
+Building a standalone MCP server image is straightforward and easy, so we just proposed a way to launch it with RAGFlow server here.
+
+#### Alongside RAGFlow
+
+As MCP server is an extra and optional component of RAGFlow server, we consume that not everybody going to use it. Thus, it is disable by default.
+To enable it, simply find `docker/docker-compose.yml` to uncomment `services.ragflow.command` section.
+
+```yaml
+services:
+  ragflow:
+    ...
+    image: ${RAGFLOW_IMAGE}
+    # example to setup MCP server
+    command:
+      - --enable-mcpserver
+      - --mcp-host=0.0.0.0
+      - --mcp-port=9382
+      - --mcp-base-url=http://127.0.0.1:9380
+      - --mcp-script-path=/ragflow/mcp/server/server.py
+      - --mcp-mode=self-host # `self-host` or `host`
+      - --mcp--host-api-key="ragflow-xxxxxxx" # only need to privide when mode is `self-host`
+```
+
+Then launch it normally `docker compose -f docker-compose.yml`.
+
+```bash
+ragflow-server  | Starting MCP Server on 0.0.0.0:9382 with base URL http://127.0.0.1:9380...
+ragflow-server  | Starting 1 task executor(s) on host 'dd0b5e07e76f'...
+ragflow-server  | 2025-04-18 15:41:18,816 INFO     27 ragflow_server log path: /ragflow/logs/ragflow_server.log, log levels: {'peewee': 'WARNING', 'pdfminer': 'WARNING', 'root': 'INFO'}
+ragflow-server  | 
+ragflow-server  | __  __  ____ ____       ____  _____ ______     _______ ____
+ragflow-server  | |  \/  |/ ___|  _ \     / ___|| ____|  _ \ \   / / ____|  _ \
+ragflow-server  | | |\/| | |   | |_) |    \___ \|  _| | |_) \ \ / /|  _| | |_) |
+ragflow-server  | | |  | | |___|  __/      ___) | |___|  _ < \ V / | |___|  _ <
+ragflow-server  | |_|  |_|\____|_|        |____/|_____|_| \_\ \_/  |_____|_| \_\
+ragflow-server  |     
+ragflow-server  | MCP launch mode: self-host
+ragflow-server  | MCP host: 0.0.0.0
+ragflow-server  | MCP port: 9382
+ragflow-server  | MCP base_url: http://127.0.0.1:9380
+ragflow-server  | INFO:     Started server process [26]
+ragflow-server  | INFO:     Waiting for application startup.
+ragflow-server  | INFO:     Application startup complete.
+ragflow-server  | INFO:     Uvicorn running on http://0.0.0.0:9382 (Press CTRL+C to quit)
+ragflow-server  | 2025-04-18 15:41:20,469 INFO     27 found 0 gpus
+ragflow-server  | 2025-04-18 15:41:23,263 INFO     27 init database on cluster mode successfully
+ragflow-server  | 2025-04-18 15:41:25,318 INFO     27 load_model /ragflow/rag/res/deepdoc/det.onnx uses CPU
+ragflow-server  | 2025-04-18 15:41:25,367 INFO     27 load_model /ragflow/rag/res/deepdoc/rec.onnx uses CPU
+ragflow-server  |         ____   ___    ______ ______ __               
+ragflow-server  |        / __ \ /   |  / ____// ____// /____  _      __
+ragflow-server  |       / /_/ // /| | / / __ / /_   / // __ \| | /| / /
+ragflow-server  |      / _, _// ___ |/ /_/ // __/  / // /_/ /| |/ |/ / 
+ragflow-server  |     /_/ |_|/_/  |_|\____//_/    /_/ \____/ |__/|__/                             
+ragflow-server  | 
+ragflow-server  |     
+ragflow-server  | 2025-04-18 15:41:29,088 INFO     27 RAGFlow version: v0.18.0-285-gb2c299fa full
+ragflow-server  | 2025-04-18 15:41:29,088 INFO     27 project base: /ragflow
+ragflow-server  | 2025-04-18 15:41:29,088 INFO     27 Current configs, from /ragflow/conf/service_conf.yaml:
+ragflow-server  |  ragflow: {'host': '0.0.0.0', 'http_port': 9380}
+...
+ragflow-server  |  * Running on all addresses (0.0.0.0)
+ragflow-server  |  * Running on http://127.0.0.1:9380
+ragflow-server  |  * Running on http://172.19.0.6:9380
+ragflow-server  |   ______           __      ______                     __            
+ragflow-server  |  /_  __/___ ______/ /__   / ____/  _____  _______  __/ /_____  _____
+ragflow-server  |   / / / __ `/ ___/ //_/  / __/ | |/_/ _ \/ ___/ / / / __/ __ \/ ___/
+ragflow-server  |  / / / /_/ (__  ) ,<    / /____>  </  __/ /__/ /_/ / /_/ /_/ / /    
+ragflow-server  | /_/  \__,_/____/_/|_|  /_____/_/|_|\___/\___/\__,_/\__/\____/_/                               
+ragflow-server  |     
+ragflow-server  | 2025-04-18 15:41:34,501 INFO     32 TaskExecutor: RAGFlow version: v0.18.0-285-gb2c299fa full
+ragflow-server  | 2025-04-18 15:41:34,501 INFO     32 Use Elasticsearch http://es01:9200 as the doc engine.
+...
+```
+
+You are ready to brew🍺!
+
+## Testing and Usage
+
+Typically, there are various ways to utilize an MCP server. You can integrate it with LLMs or use it as a standalone tool. You find the way.
+
+### Example MCP Client {#example_mcp_client}
+
+```python
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+
+from mcp.client.session import ClientSession
+from mcp.client.sse import sse_client
+
+
+async def main():
+    try:
+        # To access RAGFlow server in `host` mode, you need to attach `api_key` for each request to indicate identification.
+        # async with sse_client("http://localhost:9382/sse", headers={"api_key": "ragflow-IyMGI1ZDhjMTA2ZTExZjBiYTMyMGQ4Zm"}) as streams:
+        async with sse_client("http://localhost:9382/sse") as streams:
+            async with ClientSession(
+                streams[0],
+                streams[1],
+            ) as session:
+                await session.initialize()
+                tools = await session.list_tools()
+                print(f"{tools.tools=}")
+                response = await session.call_tool(name="ragflow_retrieval", arguments={"dataset_ids": ["ce3bb17cf27a11efa69751e139332ced"], "document_ids": [], "question": "How to install neovim?"})
+                print(f"Tool response: {response.model_dump()}")
+
+    except Exception as e:
+        print(e)
+
+
+if __name__ == "__main__":
+    from anyio import run
+
+    run(main)
+```
+
+## Security and Concerns
+
+Since MCP technology is still in booming age and there are still no official Authentication and Authorization best practices to follow, RAGFlow uses `api_key` to validate the identification, and it is required to perform any operations mentioned in the preview section. Obviously, this is not a premium solution to do so, thus this RAGFlow MCP server is not expected to exposed to public use as it could be highly venerable to be attacked. For local SSE server, bind only to localhost (127.0.0.1) instead of all interfaces (0.0.0.0). For additional guidance, you can refer to [MCP official website](https://modelcontextprotocol.io/docs/concepts/transports#security-considerations).

docs/faq.mdx

@@ -9,6 +9,10 @@ Answers to questions about general features, troubleshooting, usage, and more.
 
 ---
 
+import TOCInline from '@theme/TOCInline';
+
+<TOCInline toc={toc} />
+
 ## General features
 
 ---
@@ -71,10 +75,10 @@ We officially support x86 CPU and nvidia GPU. While we also test RAGFlow on ARM6
 
 ### Which embedding models can be deployed locally?
 
-RAGFlow offers two Docker image editions, `v0.17.2-slim` and `v0.17.2`:  
+RAGFlow offers two Docker image editions, `v0.18.0-slim` and `v0.18.0`:  
   
-- `infiniflow/ragflow:v0.17.2-slim` (default): The RAGFlow Docker image without embedding models.  
-- `infiniflow/ragflow:v0.17.2`: The RAGFlow Docker image with embedding models including:
+- `infiniflow/ragflow:v0.18.0-slim` (default): The RAGFlow Docker image without embedding models.  
+- `infiniflow/ragflow:v0.18.0`: The RAGFlow Docker image with embedding models including:
   - Built-in embedding models:
     - `BAAI/bge-large-zh-v1.5`
     - `BAAI/bge-reranker-v2-m3`
@@ -104,7 +108,7 @@ Yes, we do.
 
 ---
 
-### Is it possible to share dialogue through URL?
+### Do you support sharing dialogue through URL?
 
 No, this feature is not supported.
 
@@ -115,30 +119,29 @@ No, this feature is not supported.
 Yes, we support enhancing user queries based on existing context of an ongoing conversation:
 
 1. On the **Chat** page, hover over the desired assistant and select **Edit**.
-2. In the **Chat Configuration** popup, click the **Prompt Engine** tab.
+2. In the **Chat Configuration** popup, click the **Prompt engine** tab.
 3. Switch on **Multi-turn optimization** to enable this feature.
 
 ---
 
+### Key differences between AI search and chat?
+
+- **AI search**: This is a single-turn AI conversation using a predefined retrieval strategy (a hybrid search of weighted keyword similarity and weighted vector similarity) and the system's default chat model. It does not involve advanced RAG strategies like knowledge graph, auto-keyword, or auto-question. Retrieved chunks will be listed below the chat model's response.
+- **AI chat**: This is a multi-turn AI conversation where you can define your retrieval strategy (a weighted reranking score can be used to replace the weighted vector similarity in a hybrid search) and choose your chat model. In an AI chat, you can configure advanced RAG strategies, such as knowledge graphs, auto-keyword, and auto-question, for your specific case. Retrieved chunks are not displayed along with the answer.
+
+When debugging your chat assistant, you can use AI search as a reference to verify your model settings and retrieval strategy.
+
+---
+
 ## Troubleshooting
 
 ---
 
-### Issues with Docker images
-
----
-
-#### How to build the RAGFlow image from scratch?
+### How to build the RAGFlow image from scratch?
 
 See [Build a RAGFlow Docker image](./develop/build_docker_image.mdx).
 
----
-
-### Issues with huggingface models
-
----
-
-#### Cannot access https://huggingface.co
+### Cannot access https://huggingface.co
 
 A locally deployed RAGflow downloads OCR and embedding modules from [Huggingface website](https://huggingface.co) by default. If your machine is unable to access this site, the following error occurs and PDF parsing fails:
 
@@ -169,7 +172,7 @@ To fix this issue, use https://hf-mirror.com instead:
 
 ---
 
-#### `MaxRetryError: HTTPSConnectionPool(host='hf-mirror.com', port=443)`
+### `MaxRetryError: HTTPSConnectionPool(host='hf-mirror.com', port=443)`
 
 This error suggests that you do not have Internet access or are unable to connect to hf-mirror.com. Try the following:
 
@@ -182,17 +185,13 @@ This error suggests that you do not have Internet access or are unable to connec
 
 ---
 
-### Issues with RAGFlow servers
-
----
-
-#### `WARNING: can't find /raglof/rag/res/borker.tm`
+### `WARNING: can't find /raglof/rag/res/borker.tm`
 
 Ignore this warning and continue. All system warnings can be ignored.
 
 ---
 
-#### `network anomaly There is an abnormality in your network and you cannot connect to the server.`
+### `network anomaly There is an abnormality in your network and you cannot connect to the server.`
 
 ![anomaly](https://github.com/infiniflow/ragflow/assets/93570324/beb7ad10-92e4-4a58-8886-bfb7cbd09e5d)
 
@@ -215,11 +214,7 @@ You will not log in to RAGFlow unless the server is fully initialized. Run `dock
 
 ---
 
-### Issues with RAGFlow backend services
-
----
-
-#### `Realtime synonym is disabled, since no redis connection`
+### `Realtime synonym is disabled, since no redis connection`
 
 Ignore this warning and continue. All system warnings can be ignored.
 
@@ -227,7 +222,7 @@ Ignore this warning and continue. All system warnings can be ignored.
 
 ---
 
-#### Why does my document parsing stall at under one percent?
+### Why does my document parsing stall at under one percent?
 
 ![stall](https://github.com/infiniflow/ragflow/assets/93570324/3589cc25-c733-47d5-bbfc-fedb74a3da50)
 
@@ -244,7 +239,7 @@ Click the red cross beside the 'parsing status' bar, then restart the parsing pr
 
 ---
 
-#### Why does my pdf parsing stall near completion, while the log does not show any error?
+### Why does my pdf parsing stall near completion, while the log does not show any error?
 
 Click the red cross beside the 'parsing status' bar, then restart the parsing process to see if the issue remains. If the issue persists and your RAGFlow is deployed locally, the parsing process is likely killed due to insufficient RAM. Try increasing your memory allocation by increasing the `MEM_LIMIT` value in **docker/.env**.
 
@@ -265,13 +260,13 @@ docker compose up -d
 
 ---
 
-#### `Index failure`
+### `Index failure`
 
 An index failure usually indicates an unavailable Elasticsearch service.
 
 ---
 
-#### How to check the log of RAGFlow?
+### How to check the log of RAGFlow?
 
 ```bash
 tail -f ragflow/docker/ragflow-logs/*.log
@@ -279,7 +274,7 @@ tail -f ragflow/docker/ragflow-logs/*.log
 
 ---
 
-#### How to check the status of each component in RAGFlow?
+### How to check the status of each component in RAGFlow?
 
 1. Check the status of the Elasticsearch Docker container:
 
@@ -304,7 +299,7 @@ The status of a Docker container status does not necessarily reflect the status
 
 ---
 
-#### `Exception: Can't connect to ES cluster`
+### `Exception: Can't connect to ES cluster`
 
 1. Check the status of the Elasticsearch Docker container:
 
@@ -328,19 +323,19 @@ The status of a Docker container status does not necessarily reflect the status
 
 ---
 
-#### Can't start ES container and get `Elasticsearch did not exit normally`
+### Can't start ES container and get `Elasticsearch did not exit normally`
 
 This is because you forgot to update the `vm.max_map_count` value in **/etc/sysctl.conf** and your change to this value was reset after a system reboot.
 
 ---
 
-#### `{"data":null,"code":100,"message":"<NotFound '404: Not Found'>"}`
+### `{"data":null,"code":100,"message":"<NotFound '404: Not Found'>"}`
 
 Your IP address or port number may be incorrect. If you are using the default configurations, enter `http://<IP_OF_YOUR_MACHINE>` (**NOT 9380, AND NO PORT NUMBER REQUIRED!**) in your browser. This should work.
 
 ---
 
-#### `Ollama - Mistral instance running at 127.0.0.1:11434 but cannot add Ollama as model in RagFlow`
+### `Ollama - Mistral instance running at 127.0.0.1:11434 but cannot add Ollama as model in RagFlow`
 
 A correct Ollama IP address and port is crucial to adding models to Ollama:
 
@@ -351,37 +346,13 @@ See [Deploy a local LLM](./guides/models/deploy_local_llm.mdx) for more informat
 
 ---
 
-#### Do you offer examples of using DeepDoc to parse PDF or other files?
+### Do you offer examples of using DeepDoc to parse PDF or other files?
 
 Yes, we do. See the Python files under the **rag/app** folder.
 
 ---
 
-#### Why did I fail to upload a 128MB+ file to my locally deployed RAGFlow?
-
-Ensure that you update the **MAX_CONTENT_LENGTH** environment variable:
-
-1. In **ragflow/docker/.env**, uncomment environment variable `MAX_CONTENT_LENGTH`:
-
-   ```
-   MAX_CONTENT_LENGTH=176160768 # 168MB
-   ```
-
-2. Update **ragflow/docker/nginx/nginx.conf**:
-
-   ```
-   client_max_body_size 168M
-   ```
-
-3. Restart the RAGFlow server:
-
-   ```
-   docker compose up ragflow -d
-   ```
-
----
-
-#### `FileNotFoundError: [Errno 2] No such file or directory`
+### `FileNotFoundError: [Errno 2] No such file or directory`
 
 1. Check the status of the MinIO Docker container:
 
@@ -407,21 +378,13 @@ The status of a Docker container status does not necessarily reflect the status
 
 ---
 
-### How to increase the length of RAGFlow responses?
-
-1. Right-click the desired dialog to display the **Chat Configuration** window.
-2. Switch to the **Model Setting** tab and adjust the **Max Tokens** slider to get the desired length.
-3. Click **OK** to confirm your change.
-
----
-
 ### How to run RAGFlow with a locally deployed LLM?
 
 You can use Ollama or Xinference to deploy local LLM. See [here](./guides/models/deploy_local_llm.mdx) for more information.
 
 ---
 
-### Is it possible to add an LLM that is not supported?
+### How to add an LLM that is not supported?
 
 If your model is not currently supported but has APIs compatible with those of OpenAI, click **OpenAI-API-Compatible** on the **Model providers** page to configure your model:
 
@@ -429,7 +392,7 @@ If your model is not currently supported but has APIs compatible with those of O
 
 ---
 
-### How to interconnect RAGFlow with Ollama?
+### How to integrate RAGFlow with Ollama?
 
 - If RAGFlow is locally deployed, ensure that your RAGFlow and Ollama are in the same LAN.
 - If you are using our online demo, ensure that the IP address of your Ollama server is public and accessible.
@@ -438,12 +401,25 @@ See [here](./guides/models/deploy_local_llm.mdx) for more information.
 
 ---
 
+### How to change the file size limit?
+
+For a locally deployed RAGFlow: the total file size limit per upload is 1GB, with a batch upload limit of 32 files. There is no cap on the total number of files per account. To update this 1GB file size limit:
+
+- In **docker/.env**, upcomment `# MAX_CONTENT_LENGTH=1073741824`, adjust the value as needed, and note that `1073741824` represents 1GB in bytes.
+- If you update the value of `MAX_CONTENT_LENGTH` in **docker/.env**, ensure that you update `client_max_body_size` in **nginx/nginx.conf** accordingly.
+
+:::tip NOTE
+It is not recommended to manually change the 32-file batch upload limit. However, if you use RAGFlow's HTTP API or Python SDK to upload files, the 32-file batch upload limit is automatically removed.
+:::
+
+---
+
 ### `Error: Range of input length should be [1, 30000]`
 
 This error occurs because there are too many chunks matching your search criteria. Try reducing the **TopN** and increasing **Similarity threshold** to fix this issue:
 
 1. Click **Chat** in the middle top of the page.
-2. Right-click the desired conversation > **Edit** > **Prompt Engine**
+2. Right-click the desired conversation > **Edit** > **Prompt engine**
 3. Reduce the **TopN** and/or raise **Similarity threshold**.
 4. Click **OK** to confirm your changes.
 
@@ -462,3 +438,31 @@ See [Acquire a RAGFlow API key](./develop/acquire_ragflow_api_key.md).
 See [Upgrade RAGFlow](./guides/upgrade_ragflow.mdx) for more information.
 
 ---
+
+### How to switch the document engine to Infinity?
+
+To switch your document engine from Elasticsearch to [Infinity](https://github.com/infiniflow/infinity):
+
+1. Stop all running containers:  
+
+   ```bash
+   $ docker compose -f docker/docker-compose.yml down -v
+   ```
+:::caution WARNING
+`-v` will delete all Docker container volumes, and the existing data will be cleared.
+:::
+
+2. In **docker/.env**, set `DOC_ENGINE=${DOC_ENGINE:-infinity}`
+3. Restart your Docker image: 
+
+   ```bash
+   $ docker compose -f docker-compose.yml up -d
+   ```
+
+---
+
+### Where are my uploaded files stored in RAGFlow's image?
+
+All uploaded files are stored in Minio, RAGFlow's object storage solution. For instance, if you upload your file directly to a knowledge base, it is located at `<knowledgebase_id>/filename`.
+
+---

docs/guides/agent/agent_component_reference/begin.mdx

@@ -46,6 +46,17 @@ You can set global variables within the **Begin** component, which can be either
   - **boolean**: Requires the user to toggle between on and off.
 - **Optional**: A toggle indicating whether the variable is optional. 
 
+:::tip NOTE
+To pass in parameters from a client, call:
+- HTTP method [Converse with agent](../../../references/http_api_reference.md#converse-with-agent), or
+- Python method [Converse with agent](../../../references/python_api_reference.md#converse-with-agent).
+:::
+
+:::danger IMPORTANT
+- If you set the key type as **file**, ensure the token count of the uploaded file does not exceed your model provider's maximum token limit; otherwise, the plain text in your file will be truncated and incomplete.
+- If your agent's **Begin** component takes a variable, you *cannot* embed it into a webpage.
+:::
+
 ## Examples
 
 As mentioned earlier, the **Begin** component is indispensable for an agent. Still, you can take a look at our three-step interpreter agent template, where the **Begin** component takes two global variables:
@@ -60,7 +71,7 @@ As mentioned earlier, the **Begin** component is indispensable for an agent. Sti
 
 ### Is the uploaded file in a knowledge base?
 
-No. Files uploaded to an agent as input are not stored in a knowledge base and will not be chunked using RAGFlow's built-in chunk methods. However, RAGFlow's built-in OSR, DLR, and TSR models will still be applied to process the document.
+No. Files uploaded to an agent as input are not stored in a knowledge base and hence will not be processed using RAGFlow's built-in OCR, DLR or TSR models, or chunked using RAGFlow's built-in chunk methods. 
 
 ### How to upload a webpage or file from a URL?
 
@@ -70,5 +81,8 @@ If you set the type of a variable as **file**, your users will be able to upload
 
 ### File size limit for an uploaded file
 
-The maximum file size for each uploaded file is determined by the variable `MAX_CONTENT_LENGTH` in `/docker/.env`. It defaults to 128 MB. If you change the default file size, ensure you also update the value of `client_max_body_size` in `/docker/nginx/nginx.conf` accordingly.
+There is no *specific* file size limit for a file uploaded to an agent. However, note that model providers typically have a default or explicit maximum token setting, which can range from 8196 to 128k: The plain text part of the uploaded file will be passed in as the key value, but if the file's token count exceeds this limit, the string will be truncated and incomplete.
 
+:::tip NOTE
+The variables `MAX_CONTENT_LENGTH` in `/docker/.env` and `client_max_body_size` in `/docker/nginx/nginx.conf` set the file size limit for each upload to a knowledge base or **File Management**. These settings DO NOT apply in this scenario.
+:::

docs/guides/agent/agent_component_reference/categorize.mdx

@@ -33,7 +33,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
   - Ensure you set the chat model correctly on the **Model providers** page.
   - You can use different models for different components to increase flexibility or improve overall performance.
-- **Preset configurations**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.  
+- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.  
   This parameter has three options:  
   - **Improvise**: Produces more creative responses.
   - **Precise**: (Default) Produces more conservative responses.
@@ -52,9 +52,6 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
   - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
   - Defaults to 0.7.
-- **Max tokens**: Sets the maximum length of the model's output, measured in the number of tokens.  
-  - Defaults to 512. 
-  - If disabled, you lift the maximum token limit, allowing the model to determine the number of tokens in its responses.
 
 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.

docs/guides/agent/agent_component_reference/generate.mdx

@@ -24,7 +24,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
   - Ensure you set the chat model correctly on the **Model providers** page.
   - You can use different models for different components to increase flexibility or improve overall performance.
-- **Preset configurations**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
   This parameter has three options:
   - **Improvise**: Produces more creative responses.
   - **Precise**: (Default) Produces more conservative responses.
@@ -43,9 +43,6 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
   - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
   - Defaults to 0.7.
-- **Max tokens**: Sets the maximum length of the model's output, measured in the number of tokens.  
-  - Defaults to 512. 
-  - If disabled, you lift the maximum token limit, allowing the model to determine the number of tokens in its responses.
 
 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.
@@ -56,7 +53,11 @@ Click the dropdown menu of **Model** to show the model configuration window.
 
 Typically, you use the system prompt to describe the task for the LLM, specify how it should respond, and outline other miscellaneous requirements. We do not plan to elaborate on this topic, as it can be as extensive as prompt engineering. However, please be aware that the system prompt is often used in conjunction with keys (variables), which serve as various data inputs for the LLM. 
 
-Keys in a system prompt should be enclosed in curly braces. Below is a prompt excerpt of a **Generate** component from the **Interpreter** template (component ID: **Reflect**):
+:::danger IMPORTANT
+A **Generate** component relies on keys (variables) to specify its data inputs. Its immediate upstream component is *not* necessarily its data input, and the arrows in the workflow indicate *only* the processing sequence. Keys in a **Generate** component are used in conjunction with the system prompt to specify data inputs for the LLM. Use a forward slash `/` or the **(x)** button to show the keys to use.
+:::
+
+Below is a prompt excerpt of a **Generate** component from the **Interpreter** template (component ID: **Reflect**):
 
 ```text
 Your task is to read a source text and a translation to {target_lang}, and give constructive suggestions to improve the translation. The source text and initial translation, delimited by XML tags <SOURCE_TEXT></SOURCE_TEXT> and <TRANSLATION></TRANSLATION>, are as follows:
@@ -76,11 +77,6 @@ When writing suggestions, pay attention to whether there are ways to improve the
 
 Where `{source_text}` and `{target_lang}` are global variables defined by the **Begin** component, while `{translation_1}` is the output of another **Generate** component with the component ID **Translate directly**.
 
-
-:::danger IMPORTANT
-A **Generate** component relies on keys (variables) to specify its data inputs. Its immediate upstream component is *not* necessarily its data input, and the arrows in the workflow indicate *only* the processing sequence. Keys in a **Generate** component are used in conjunction with the system prompt to specify data inputs for the LLM. Use a forward slash `/` to show the keys to use.
-:::
-
 ### Cite 
 
 This toggle sets whether to cite the original text as reference. 

docs/guides/agent/agent_component_reference/keyword.mdx

@@ -34,7 +34,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
   - Ensure you set the chat model correctly on the **Model providers** page.
   - You can use different models for different components to increase flexibility or improve overall performance.
-- **Preset configurations**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
   This parameter has three options:
   - **Improvise**: Produces more creative responses.
   - **Precise**: (Default) Produces more conservative responses.
@@ -53,9 +53,6 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
   - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
   - Defaults to 0.7.
-- **Max tokens**: Sets the maximum length of the model's output, measured in the number of tokens.  
-  - Defaults to 512. 
-  - If disabled, you lift the maximum token limit, allowing the model to determine the number of tokens in its responses.
 
 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.

docs/guides/agent/agent_component_reference/retrieval.mdx

@@ -51,19 +51,34 @@ If a rerank model is selected, a combination of weighted keyword similarity and
 Using a rerank model will *significantly* increase the system's response time.
 :::
 
+### Tavily API key
+
+*Optional*
+
+Enter your Tavily API key here to enable Tavily web search during retrieval. See [here](https://app.tavily.com/home) for instructions on getting a Tavily API key.
+
+### Use knowledge graph
+
+Whether to use knowledge graph(s) in the specified knowledge base(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
 
 ### Knowledge bases 
 
-*Required*
+*Optional*
 
-You are required to select the knowledge base(s) to retrieve data from.
+Select the knowledge base(s) to retrieve data from.
 
-:::danger IMPORTANT 
-If you select multiple knowledge bases, you must ensure that the knowledge bases (datasets) you select use the same embedding model; otherwise, an error message would occur.
+- If no knowledge base is selected, meaning conversations with the agent will not be based on any knowledge base, ensure that the **Empty response** field is left blank to avoid an error.
+- If you select multiple knowledge bases, you must ensure that the knowledge bases (datasets) you select use the same embedding model; otherwise, an error message would occur.
+
+### Empty response
+
+- Set this as a response if no results are retrieved from the knowledge base(s) for your query, or 
+- Leave this field blank to allow the chat model to improvise when nothing is found.
+
+:::caution WARNING
+If you do not specify a knowledge base, you must leave this field blank; otherwise, an error would occur.
 :::
 
-
-
 ## Examples
 
 Explore our customer service agent template, where the **Retrieval** component (component ID: **Search product info**) is used to search the dataset and send the Top N results to the LLM:

docs/guides/agent/agent_component_reference/rewrite.mdx

@@ -32,7 +32,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
   - Ensure you set the chat model correctly on the **Model providers** page.
   - You can use different models for different components to increase flexibility or improve overall performance.
-- **Preset configurations**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
   This parameter has three options:
   - **Improvise**: Produces more creative responses.
   - **Precise**: (Default) Produces more conservative responses.
@@ -51,9 +51,6 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
   - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
   - Defaults to 0.7.
-- **Max tokens**: Sets the maximum length of the model's output, measured in the number of tokens.  
-  - Defaults to 512. 
-  - If disabled, you lift the maximum token limit, allowing the model to determine the number of tokens in its responses.
 
 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.

docs/guides/agent/embed_agent_into_webpage.md

@@ -8,7 +8,7 @@ slug: /embed_agent_into_webpage
 You can use iframe to embed an agent into a third-party webpage.
 
 :::caution WARNING
-If your agent's **Begin** component takes a key of **file** type (a **file** type variable), you *cannot* embed it into a webpage.
+If your agent's **Begin** component takes a variable, you *cannot* embed it into a webpage.
 :::
 
 1. Before proceeding, you must [acquire an API key](../models/llm_api_key_setup.md); otherwise, an error message would appear.

docs/guides/chat/accelerate_question_answering.mdx

@@ -12,9 +12,15 @@ A checklist to speed up question answering.
 
 Please note that some of your settings may consume a significant amount of time. If you often find that your question answering is time-consuming, here is a checklist to consider:
 
-- In the **Prompt Engine** tab of your **Chat Configuration** dialogue, disabling **Multi-turn optimization** will reduce the time required to get an answer from the LLM.
-- In the **Prompt Engine** tab of your **Chat Configuration** dialogue, leaving the **Rerank model** field empty will significantly decrease retrieval time.
-- In the **Assistant Setting** tab of your **Chat Configuration** dialogue, disabling **Keyword analysis** will reduce the time to receive an answer from the LLM.
+- In the **Prompt engine** tab of your **Chat Configuration** dialogue, disabling **Multi-turn optimization** will reduce the time required to get an answer from the LLM.
+- In the **Prompt engine** tab of your **Chat Configuration** dialogue, leaving the **Rerank model** field empty will significantly decrease retrieval time.
+- When using a rerank model, ensure you have a GPU for acceleration; otherwise, the reranking process will be *prohibitively* slow.
+
+:::tip NOTE 
+Please note that rerank models are essential in certain scenarios. There is always a trade-off between speed and performance; you must weigh the pros against cons for your specific case.
+:::
+
+- In the **Assistant settings** tab of your **Chat Configuration** dialogue, disabling **Keyword analysis** will reduce the time to receive an answer from the LLM.
 - When chatting with your chat assistant, click the light bulb icon above the *current* dialogue and scroll down the popup window to view the time taken for each task:  
    ![enlighten](https://github.com/user-attachments/assets/fedfa2ee-21a7-451b-be66-20125619923c)  
 

docs/guides/chat/implement_deep_research.md

@@ -15,11 +15,11 @@ From v0.17.0 onward, RAGFlow supports integrating agentic reasoning in an AI cha
 
 To activate this feature:
 
-1. Enable the **Reasoning** toggle under the **Prompt Engine** tab of your chat assistant dialogue.
+1. Enable the **Reasoning** toggle under the **Prompt engine** tab of your chat assistant dialogue.
 
 ![Image](https://github.com/user-attachments/assets/4a1968d0-0128-4371-879f-77f3a70197f5)
 
-2. Enter the correct Tavily API key under the **Assistant Setting** tab of your chat assistant dialogue to leverage Tavily-based web search
+2. Enter the correct Tavily API key under the **Assistant settings** tab of your chat assistant dialogue to leverage Tavily-based web search
 
 ![Image](https://github.com/user-attachments/assets/e8787532-7e72-49ef-8951-169ae544512f)
 

docs/guides/chat/set_chat_variables.md

@@ -0,0 +1,114 @@
+---
+sidebar_position: 4
+slug: /set_chat_variables
+---
+
+# Set variables
+
+Set variables to be used together with the system prompt for your LLM.
+
+---
+
+When configuring the system prompt for a chat model, variables play an important role in enhancing flexibility and reusability. With variables, you can dynamically adjust the system prompt to be sent to your model. In the context of RAGFlow, if you have defined variables in the **Chat Configuration** dialogue, except for the system's reserved variable `{knowledge}`, you are required to pass in values for them from RAGFlow's [HTTP API](../../references/http_api_reference.md#converse-with-chat-assistant) or through its [Python SDK](../../references/python_api_reference.md#converse-with-chat-assistant).
+
+:::danger IMPORTANT
+In RAGFlow, variables are closely linked with the system prompt. When you add a variable in the **Variable** section, include it in the system prompt. Conversely, when deleting a variable, ensure it is removed from the system prompt; otherwise, an error would occur.
+:::
+
+## Where to set variables
+
+Hover your mouse over your chat assistant, click **Edit** to open its **Chat Configuration** dialogue, then click the **Prompt engine** tab. Here, you can work on your variables in the **System prompt** field and the **Variable** section:
+
+![set_variables](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/prompt_engine.jpg)
+
+## 1. Manage variables
+
+In the **Variable** section, you add, remove, or update variables.
+
+### `{knowledge}` - a reserved variable
+
+`{knowledge}` is the system's reserved variable, representing the chunks retrieved from the knowledge base(s) specified by **Knowledge bases** under the **Assistant settings** tab. If your chat assistant is associated with certain knowledge bases, you can keep it as is.
+
+:::info NOTE
+It does not currently make a difference whether you set `{knowledge}` to optional or mandatory, but note that this design will be updated at a later point.
+:::
+
+From v0.17.0 onward, you can start an AI chat without specifying knowledge bases. In this case, we recommend removing the `{knowledge}` variable to prevent unnecessary reference and keeping the **Empty response** field empty to avoid errors.
+
+### Custom variables
+
+Besides `{knowledge}`, you can also define your own variables to pair with the system prompt. To use these custom variables, you must pass in their values through RAGFlow's official APIs. The **Optional** toggle determines whether these variables are required in the corresponding APIs:
+
+- **Disabled** (Default): The variable is mandatory and must be provided.
+- **Enabled**: The variable is optional and can be omitted if not needed.
+
+
+
+## 2. Update system prompt
+
+After you add or remove variables in the **Variable** section, ensure your changes are reflected in the system prompt to avoid inconsistencies or errors. Here's an example:
+
+```
+You are an intelligent assistant. Please answer the question by summarizing chunks from the specified knowledge base(s)...
+
+Your answers should follow a professional and {style} style.
+
+...
+
+Here is the knowledge base:
+{knowledge}
+The above is the knowledge base.
+```
+
+:::tip NOTE
+If you have removed `{knowledge}`, ensure that you thoroughly review and update the entire system prompt to achieve optimal results.
+:::
+
+## APIs
+
+The *only* way to pass in values for the custom variables defined in the **Chat Configuration** dialogue is to call RAGFlow's [HTTP API](../../references/http_api_reference.md#converse-with-chat-assistant) or through its [Python SDK](../../references/python_api_reference.md#converse-with-chat-assistant).
+
+### HTTP API
+
+See [Converse with chat assistant](../../references/http_api_reference.md#converse-with-chat-assistant). Here's an example:
+
+```json {9}
+curl --request POST \
+     --url http://{address}/api/v1/chats/{chat_id}/completions \
+     --header 'Content-Type: application/json' \
+     --header 'Authorization: Bearer <YOUR_API_KEY>' \
+     --data-binary '
+     {
+          "question": "xxxxxxxxx",
+          "stream": true,
+          "style":"hilarious"
+     }'
+```
+
+### Python API
+
+See [Converse with chat assistant](../../references/python_api_reference.md#converse-with-chat-assistant). Here's an example:
+
+```python {18}
+from ragflow_sdk import RAGFlow
+
+rag_object = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
+assistant = rag_object.list_chats(name="Miss R")
+assistant = assistant[0]
+session = assistant.create_session()    
+
+print("\n==================== Miss R =====================\n")
+print("Hello. What can I do for you?")
+
+while True:
+    question = input("\n==================== User =====================\n> ")
+    style = input("Please enter your preferred style (e.g., formal, informal, hilarious): ")
+    
+    print("\n==================== Miss R =====================\n")
+    
+    cont = ""
+    for ans in session.ask(question, stream=True, style=style):
+        print(ans.content[len(cont):], end='', flush=True)
+        cont = ans.content
+```
+

docs/guides/chat/start_chat.md

@@ -19,7 +19,7 @@ You start an AI conversation by creating an assistant.
 
    > RAGFlow offers you the flexibility of choosing a different chat model for each dialogue, while allowing you to set the default models in **System Model Settings**.
 
-2. Update **Assistant Setting**:
+2. Update **Assistant settings**:
 
    - **Assistant name** is the name of your chat assistant. Each assistant corresponds to a dialogue with a unique combination of knowledge bases, prompts, hybrid search configurations, and large model settings.
    - **Empty response**:
@@ -28,7 +28,7 @@ You start an AI conversation by creating an assistant.
    - **Show quote**: This is a key feature of RAGFlow and enabled by default. RAGFlow does not work like a black box. Instead, it clearly shows the sources of information that its responses are based on.
    - Select the corresponding knowledge bases. You can select one or multiple knowledge bases, but ensure that they use the same embedding model, otherwise an error would occur.
 
-3. Update **Prompt Engine**:
+3. Update **Prompt engine**:
 
    - In **System**, you fill in the prompts for your LLM, you can also leave the default prompt as-is for the beginning.
    - **Similarity threshold** sets the similarity "bar" for each chunk of text. The default is 0.2. Text chunks with lower similarity scores are filtered out of the final response.
@@ -37,19 +37,39 @@ You start an AI conversation by creating an assistant.
      - If **Rerank model** is selected, the hybrid score system uses keyword similarity and reranker score, and the default weight assigned to the reranker score is 1-0.7=0.3.
    - **Top N** determines the *maximum* number of chunks to feed to the LLM. In other words, even if more chunks are retrieved, only the top N chunks are provided as input.
    - **Multi-turn optimization** enhances user queries using existing context in a multi-round conversation. It is enabled by default. When enabled, it will consume additional LLM tokens and significantly increase the time to generate answers.
+   - **Use knowledge graph** indicates whether to use knowledge graph(s) in the specified knowledge base(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
+   - **Reasoning** indicates whether to generate answers through reasoning processes like Deepseek-R1/OpenAI o1. Once enabled, the chat model autonomously integrates Deep Research during question answering when encountering an unknown topic. This involves the chat model dynamically searching external knowledge and generating final answers through reasoning.
    - **Rerank model** sets the reranker model to use. It is left empty by default.
      - If **Rerank model** is left empty, the hybrid score system uses keyword similarity and vector similarity, and the default weight assigned to the vector similarity component is 1-0.7=0.3.
      - If **Rerank model** is selected, the hybrid score system uses keyword similarity and reranker score, and the default weight assigned to the reranker score is 1-0.7=0.3.
    - **Variable** refers to the variables (keys) to be used in the system prompt. `{knowledge}` is a reserved variable. Click **Add** to add more variables for the system prompt.
       - If you are uncertain about the logic behind **Variable**, leave it *as-is*.
+      - As of v0.18.0, if you add custom variables here, the only way you can pass in their values is to call:
+         - HTTP method [Converse with chat assistant](../../references/http_api_reference.md#converse-with-chat-assistant), or
+         - Python method [Converse with chat assistant](../../references/python_api_reference.md#converse-with-chat-assistant).
 
 4. Update **Model Setting**:
 
    - In **Model**: you select the chat model. Though you have selected the default chat model in **System Model Settings**, RAGFlow allows you to choose an alternative chat model for your dialogue.
-   - **Preset configurations** refers to the level that the LLM improvises. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.
-   - **Temperature**: Level of the prediction randomness of the LLM. The higher the value, the more creative the LLM is.
-   - **Top P** is also known as "nucleus sampling". See [here](https://en.wikipedia.org/wiki/Top-p_sampling) for more information.
-   - **Max Tokens**: The maximum length of the LLM's responses. Note that the responses may be curtailed if this value is set too low.
+   - **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+   This parameter has three options:
+      - **Improvise**: Produces more creative responses.
+      - **Precise**: (Default) Produces more conservative responses.
+      - **Balance**: A middle ground between **Improvise** and **Precise**.
+   - **Temperature**: The randomness level of the model's output.  
+   Defaults to 0.1.
+      - Lower values lead to more deterministic and predictable outputs.
+      - Higher values lead to more creative and varied outputs.
+      - A temperature of zero results in the same output for the same prompt.
+   - **Top P**: Nucleus sampling.  
+      - Reduces the likelihood of generating repetitive or unnatural text by setting a threshold *P* and restricting the sampling to tokens with a cumulative probability exceeding *P*.
+      - Defaults to 0.3.
+   - **Presence penalty**: Encourages the model to include a more diverse range of tokens in the response.  
+      - A higher **presence penalty** value results in the model being more likely to generate tokens not yet been included in the generated text.
+      - Defaults to 0.4.
+   - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
+      - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
+      - Defaults to 0.7.
 
 5. Now, let's start the show:
 

docs/guides/dataset/accelerate_doc_indexing.mdx

@@ -15,5 +15,5 @@ Please note that some of your settings may consume a significant amount of time.
 - Use GPU to reduce embedding time.
 - On the configuration page of your knowledge base, switch off **Use RAPTOR to enhance retrieval**.
 - Extracting knowledge graph (GraphRAG) is time-consuming.
-- Disable **Auto-keyword** and **Auto-question** on the configuration page of yor knowledge base, as both depend on the LLM.
-- **v0.17.0:** If your document is plain text PDF and does not require GPU-intensive processes like OCR (Optical Character Recognition), TSR (Table Structure Recognition), or DLA (Document Layout Analysis), you can choose **Naive** over **DeepDoc** or other time-consuming large model options in the **Document parser** dropdown. This will substantially reduce document parsing time.
+- Disable **Auto-keyword** and **Auto-question** on the configuration page of your knowledge base, as both depend on the LLM.
+- **v0.17.0+:** If your document is plain text PDF and does not require GPU-intensive processes like OCR (Optical Character Recognition), TSR (Table Structure Recognition), or DLA (Document Layout Analysis), you can choose **Naive** over **DeepDoc** or other time-consuming large model options in the **Document parser** dropdown. This will substantially reduce document parsing time.

docs/guides/dataset/configure_knowledge_base.md

@@ -39,18 +39,20 @@ This section covers the following topics:
 
 RAGFlow offers multiple chunking template to facilitate chunking files of different layouts and ensure semantic integrity. In **Chunk method**, you can choose the default template that suits the layouts and formats of your files. The following table shows the descriptions and the compatible file formats of each supported chunk template:
 
-| **Template** | Description                                                           | File format                                                                  |
-|--------------|-----------------------------------------------------------------------|------------------------------------------------------------------------------|
-| General      | Files are consecutively chunked based on a preset chunk token number. | DOCX, XLSX, XLS (Excel97~2003), PPT, PDF, TXT, JPEG, JPG, PNG, TIF, GIF, CSV |
-| Q&A          |                                                                       | XLSX, XLS (Excel97~2003), CSV/TXT                                            |
-| Manual       |                                                                       | PDF                                                                          |
-| Table        |                                                                       | XLSX, XLS (Excel97~2003), CSV/TXT                                            |
-| Paper        |                                                                       | PDF                                                                          |
-| Book         |                                                                       | DOCX, PDF, TXT                                                               |
-| Laws         |                                                                       | DOCX, PDF, TXT                                                               |
-| Presentation |                                                                       | PDF, PPTX                                                                    |
-| Picture      |                                                                       | JPEG, JPG, PNG, TIF, GIF                                                     |
-| One          | The entire document is chunked as one.                                | DOCX, XLSX, XLS (Excel97~2003), PDF, TXT                                     |
+| **Template** | Description                                                           | File format                                                                                   |
+|--------------|-----------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|
+| General      | Files are consecutively chunked based on a preset chunk token number. | DOCX, XLSX, XLS (Excel97~2003), PPT, PDF, TXT, JPEG, JPG, PNG, TIF, GIF, CSV, JSON, EML, HTML |
+| Q&A          |                                                                       | XLSX, XLS (Excel97~2003), CSV/TXT                                                             |
+| Resume       | Enterprise edition only. You can also try it out on demo.ragflow.io.  | DOCX, PDF, TXT                                                                                |
+| Manual       |                                                                       | PDF                                                                                           |
+| Table        |                                                                       | XLSX, XLS (Excel97~2003), CSV/TXT                                                             |
+| Paper        |                                                                       | PDF                                                                                           |
+| Book         |                                                                       | DOCX, PDF, TXT                                                                                |
+| Laws         |                                                                       | DOCX, PDF, TXT                                                                                |
+| Presentation |                                                                       | PDF, PPTX                                                                                     |
+| Picture      |                                                                       | JPEG, JPG, PNG, TIF, GIF                                                                      |
+| One          | Each document is chunked in its entirety (as one).                    | DOCX, XLSX, XLS (Excel97~2003), PDF, TXT                                                      |
+| Tag          | The knowledge base functions as a tag set for the others.             | XLSX, CSV/TXT                                                                                 |
 
 You can also change a file's chunk method on the **Datasets** page.
 
@@ -63,14 +65,6 @@ An embedding model converts chunks into embeddings. It cannot be changed once th
 The following embedding models can be deployed locally:
 
 - BAAI/bge-large-zh-v1.5
-- BAAI/bge-base-en-v1.5
-- BAAI/bge-large-en-v1.5
-- BAAI/bge-small-en-v1.5
-- BAAI/bge-small-zh-v1.5
-- jinaai/jina-embeddings-v2-base-en
-- jinaai/jina-embeddings-v2-small-en
-- nomic-ai/nomic-embed-text-v1.5
-- sentence-transformers/all-MiniLM-L6-v2
 - maidalun1020/bce-embedding-base_v1
 
 ### Upload file
@@ -130,7 +124,7 @@ See [Run retrieval test](./run_retrieval_test.md) for details.
 
 ## Search for knowledge base
 
-As of RAGFlow v0.17.2, the search feature is still in a rudimentary form, supporting only knowledge base search by name.
+As of RAGFlow v0.18.0, the search feature is still in a rudimentary form, supporting only knowledge base search by name.
 
 ![search knowledge base](https://github.com/infiniflow/ragflow/assets/93570324/836ae94c-2438-42be-879e-c7ad2a59693e)
 

docs/guides/dataset/construct_knowledge_graph.md

@@ -1,5 +1,5 @@
 ---
-sidebar_position: 2
+sidebar_position: 8
 slug: /construct_knowledge_graph
 ---
 
@@ -23,6 +23,10 @@ Constructing a knowledge graph requires significant memory, computational resour
 
 Knowledge graphs are especially useful for multi-hop question-answering involving *nested* logic. They outperform traditional extraction approaches when you are performing question answering on books or works with complex entities and relationships.
 
+:::tip NOTE
+RAPTOR (Recursive Abstractive Processing for Tree Organized Retrieval) can also be used for multi-hop question-answering tasks. See [Enable RAPTOR](./enable_raptor.md) for details. You may use either approach or both, but ensure you understand the memory, computational, and token costs involved.
+:::
+
 ## Prerequisites
 
 The system's default chat model is used to generate knowledge graph. Before proceeding, ensure that you have a chat model properly configured:
@@ -68,6 +72,10 @@ In a knowledge graph, a community is a cluster of entities linked by relationshi
    _A **Knowledge graph** entry appears under **Configuration** once a knowledge graph is created._
 
 3. Click **Knowledge graph** to view the details of the generated graph.
+4. To use the created knowledge graph, do either of the following:
+   
+   - In your **Chat Configuration** dialogue, click the **Assistant settings** tab to add the corresponding knowledge base(s) and click the **Prompt engine** tab to switch on the **Use knowledge graph** toggle.
+   - If you are using an agent, click the **Retrieval** agent component to specify the knowledge base(s) and switch on the **Use knowledge graph** toggle.
 
 ## Frequently asked questions
 

docs/guides/dataset/enable_raptor.md

@@ -0,0 +1,74 @@
+---
+sidebar_position: 7
+slug: /enable_raptor
+---
+
+# Enable RAPTOR
+
+A recursive abstractive method used in long-context knowledge retrieval and summarization, balancing broad semantic understanding with fine details.
+
+---
+
+RAPTOR (Recursive Abstractive Processing for Tree Organized Retrieval) is an enhanced document preprocessing technique introduced in a [2024 paper](https://arxiv.org/html/2401.18059v1). Designed to tackle multi-hop question-answering issues, RAPTOR performs recursive clustering and summarization of document chunks to build a hierarchical tree structure. This enables more context-aware retrieval across lengthy documents. RAGFlow v0.6.0 integrates RAPTOR for document clustering as part of its data preprocessing pipeline between data extraction and indexing, as illustrated below.
+
+![document_clustering](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/document_clustering_as_preprocessing.jpg)
+
+Our tests with this new approach demonstrate state-of-the-art (SOTA) results on question-answering tasks requiring complex, multi-step reasoning. By combining RAPTOR retrieval with our built-in chunking methods and/or other retrieval-augmented generation (RAG) approaches, you can further improve your question-answering accuracy.
+
+:::danger WARNING
+Enabling RAPTOR requires significant memory, computational resources, and tokens.
+:::
+
+## Basic principles
+
+After the original documents are divided into chunks, the chunks are clustered by semantic similarity rather than by their original order in the text. Clusters are then summarized into higher-level chunks by your system's default chat model. This process is applied recursively, forming a tree structure with various levels of summarization from the bottom up. As illustrated in the figure below, the initial chunks form the leaf nodes (shown in blue) and are progressively summarized into a root node (shown in orange).
+
+![raptor](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/clustering_and_summarizing.jpg)
+
+The recursive clustering and summarization capture a broad understanding (by the root node) as well as fine details (by the leaf nodes) necessary for multi-hop question-answering.
+
+## Scenarios
+
+For multi-hop question-answering tasks involving complex, multi-step reasoning, a semantic gap often exists between the question and its answer. As a result, searching with the question often fails to retrieve the relevant chunks that contribute to the correct answer. RAPTOR addresses this challenge by providing the chat model with richer and more context-aware and relevant chunks to summarize, enabling a holistic understanding without losing granular details.
+
+:::tip NOTE
+Knowledge graphs can also be used for multi-hop question-answering tasks. See [Construct knowledge graph](./construct_knowledge_graph.md) for details. You may use either approach or both, but ensure you understand the memory, computational, and token costs involved.
+:::
+
+## Prerequisites
+
+The system's default chat model is used to summarize clustered content. Before proceeding, ensure that you have a chat model properly configured:
+
+![Image](https://github.com/user-attachments/assets/6bc34279-68c3-4d99-8d20-b7bd1dafc1c1)
+
+## Configurations
+
+The RAPTOR feature is disabled by default. To enable it, manually switch on the **Use RAPTOR to enhance retrieval** toggle on your knowledge base's **Configuration** page.
+
+### Prompt
+
+The following prompt will be applied recursively for cluster summarization, with `{cluster_content}` serving as an internal parameter. We recommend that you keep it as-is for now. The design will be updated at a later point.
+
+```
+Please summarize the following paragraphs... Paragraphs as following:
+      {cluster_content}
+The above is the content you need to summarize.
+```
+
+### Max token
+
+The maximum number of tokens per generated summary chunk. Defaults to 256, with a maximum limit of 2048.
+
+### Threshold
+
+In RAPTOR, chunks are clustered by their semantic similarity. The **Threshold** parameter sets the minimum similarity required for chunks to be grouped together.
+
+It defaults to 0.1, with a maximum limit of 1. A higher **Threshold** means fewer chunks in each cluster, while a lower one means more.
+
+### Max cluster
+
+The maximum number of clusters to create. Defaults to 108, with a maximum limit of 1024.
+
+### Random seed
+
+A random seed. Click the **+** button to change the seed value.

docs/guides/dataset/run_retrieval_test.md

@@ -27,7 +27,7 @@ In contrast, chunks created from [knowledge graph construction](./construct_know
 
 ### Similarity threshold
 
-This sets the bar for retrieving chunks: chunks with similarities below the threshold will be filtered out. By default, the threshold is set to 0.2.
+This sets the bar for retrieving chunks: chunks with similarities below the threshold will be filtered out. By default, the threshold is set to 0.2. This means that only chunks with hybrid similarity score of 20 or higher will be retrieved.
 
 ### Keyword similarity weight
 

docs/guides/dataset/set_page_rank.md

@@ -0,0 +1,39 @@
+---
+sidebar_position: 3
+slug: /set_page_rank
+---
+
+# Set page rank
+
+Create a step-retrieval strategy using page rank.
+
+---
+
+## Scenario
+
+In an AI-powered chat, you can configure a chat assistant or an agent to respond using knowledge retrieved from multiple specified knowledge bases (datasets), provided that they employ the same embedding model. In situations where you prefer information from certain knowledge base(s) to take precedence or to be retrieved first, you can use RAGFlow's page rank feature to increase the ranking of chunks from these knowledge bases. For example, if you have configured a chat assistant to draw from two knowledge bases, knowledge base A for 2024 news and knowledge base B for 2023 news, but wish to prioritize news from year 2024, this feature is particularly useful.
+
+:::info NOTE
+It is important to note that this 'page rank' feature operates at the level of the entire knowledge base rather than on individual files or documents.
+:::
+
+## Configuration
+
+On the **Configuration** page of your knowledge base, drag the slider under **Page rank** to set the page rank value for your knowledge base. You are also allowed to input the intended page rank value in the field next to the slider.
+
+:::info NOTE
+The page rank value must be an integer. Range: [0,100]
+
+- 0: Disabled (Default)
+- A specific value: enabled
+:::
+
+:::tip NOTE
+If you set the page rank value to a non-integer, say 1.7, it will be rounded down to the nearest integer, which in this case is 1.
+:::
+
+## Mechanism
+
+If you configure a chat assistant's **similarity threshold** to 0.2, only chunks with a hybrid score greater than 0.2 x 100 = 20 will be retrieved and sent to the chat model for content generation. This initial filtering step is crucial for narrowing down relevant information.
+
+If you have assigned a page rank of 1 to knowledge base A (2024 news) and 0 to knowledge base B (2023 news), the final hybrid scores of the retrieved chunks will be adjusted accordingly. A chunk retrieved from knowledge base A with an initial score of 50 will receive a boost of 1 x 100 = 100 points, resulting in a final score of 50 + 1 x 100 = 150. In this way, chunks retrieved from knowledge base A will always precede chunks from knowledge base B.

docs/guides/dataset/use_tag_sets.md

@@ -11,11 +11,15 @@ Use a tag set to tag chunks in your datasets.
 
 Retrieval accuracy is the touchstone for a production-ready RAG framework. In addition to retrieval-enhancing approaches like auto-keyword, auto-question, and knowledge graph, RAGFlow introduces an auto-tagging feature to address semantic gaps. The auto-tagging feature automatically maps tags in the user-defined tag sets to relevant chunks within your knowledge base based on similarity with each chunk. This automation mechanism allows you to apply an additional "layer" of domain-specific knowledge to existing datasets, which is particularly useful when dealing with a large number of chunks.
 
-To use this feature, ensure you have at least one properly configured tag set, specify the tag set(s) on the **Configuration** page of your knowledge base (dataset), and then re-parse your documents to initiate the auto-tag process. During this process, each chunk in your dataset is compared with every entry in the specified tag set(s), and tags are automatically applied based on similarity.
+To use this feature, ensure you have at least one properly configured tag set, specify the tag set(s) on the **Configuration** page of your knowledge base (dataset), and then re-parse your documents to initiate the auto-tagging process. During this process, each chunk in your dataset is compared with every entry in the specified tag set(s), and tags are automatically applied based on similarity.
+
+:::caution NOTE
+The auto-tagging feature is *unavailable* on the [Infinity](https://github.com/infiniflow/infinity) document engine.
+:::
 
 ## Scenarios
 
-Auto-tagging applies in situations where chunks are so similar to each other that the intended chunks cannot be distinguished from the rest. For example, when you have a few chunks about iPhone and a majority about iPhone case or iPhone accessaries, it becomes difficult to retrieve the iPhone-specific chunks without additional information.
+Auto-tagging applies in situations where chunks are so similar to each other that the intended chunks cannot be distinguished from the rest. For example, when you have a few chunks about iPhone and a majority about iPhone case or iPhone accessaries, it becomes difficult to retrieve those chunks about iPhone without additional information.
 
 ## Create tag set
 

docs/guides/manage_files.md

@@ -1,5 +1,5 @@
 ---
-sidebar_position: 5
+sidebar_position: 6
 slug: /manage_files
 ---
 
@@ -7,7 +7,7 @@ slug: /manage_files
 
 Knowledge base, hallucination-free chat, and file management are the three pillars of RAGFlow. RAGFlow's file management allows you to upload files individually or in bulk. You can then link an uploaded file to multiple target knowledge bases. This guide showcases some basic usages of the file management feature.
 
-:::danger IMPORTANT
+:::info IMPORTANT
 Compared to uploading files directly to various knowledge bases, uploading them to RAGFlow's file management and then linking them to different knowledge bases is *not* an unnecessary step, particularly when you want to delete some parsed files or an entire knowledge base but retain the original files.
 :::
 
@@ -17,7 +17,9 @@ RAGFlow's file management allows you to establish your file system with nested f
 
 ![create new folder](https://github.com/infiniflow/ragflow/assets/93570324/3a37a5f4-43a6-426d-a62a-e5cd2ff7a533)
 
-> Each knowledge base in RAGFlow has a corresponding folder under the **root/.knowledgebase** directory. You are not allowed to create a subfolder within it.
+:::caution NOTE
+Each knowledge base in RAGFlow has a corresponding folder under the **root/.knowledgebase** directory. You are not allowed to create a subfolder within it.
+:::
 
 ## Upload file
 
@@ -85,4 +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 v0.17.2, bulk download is not supported, nor can you download an entire folder. 
+> As of RAGFlow v0.18.0, bulk download is not supported, nor can you download an entire folder. 

docs/guides/manage_team_members.md

@@ -1,41 +0,0 @@
----
-sidebar_position: 4
-slug: /manage_team_members
----
-
-# Team
-
-Invite or remove team members, join or leave a team.
-
----
-
-By default, each RAGFlow user is assigned a single team named after their name. RAGFlow allows you to invite RAGFlow users to your team. Your team members can help you:
-
-- Upload documents to your datasets (knowledge bases).
-- Update document configurations in your datasets.
-- Update the default configurations for your datasets.
-- Parse documents in your datasets.
-
-:::tip NOTE
-Team members are currently *not* allowed to invite users to your team, and only you, the team owner, is permitted to do so.
-:::
-
-To enter the **Team** page, click on your avatar in the top right corner of the page **>** Team:
-
-![team](https://github.com/user-attachments/assets/0eac2503-26bc-4568-b3f2-bcd84069a07a)
-
-_On the **Team** page, you can view the information about members of your team and the teams you have joined._
-
-## Invite team members
-
-You are, by default, the owner of your own team and the only person permitted to invite users to join your team or remove team members.
-
-![invite_team_member](https://github.com/user-attachments/assets/d85b55c3-7e86-4f04-a414-ca18a9ee8963)
-
-## Remove team members
-
-![remove_members](https://github.com/user-attachments/assets/5c1a6ab5-8862-47a0-ad09-77fe88866508)
-
-## Accept or decline team invite
-
-![accept_or_decline_team_invite](https://github.com/user-attachments/assets/6a2cb61f-03d5-4423-9ed1-71df97ff4114)

docs/guides/models/deploy_local_llm.mdx

@@ -3,11 +3,11 @@ sidebar_position: 2
 slug: /deploy_local_llm
 ---
 
-# Deploy LLM locally
+# Deploy local models
 import Tabs from '@theme/Tabs';
 import TabItem from '@theme/TabItem';
 
-Run models locally using Ollama, Xinference, or other frameworks.
+Deploy and run local models using Ollama, Xinference, or other frameworks.
 
 ---
 
@@ -28,7 +28,7 @@ This user guide does not intend to cover much of the installation or configurati
 - For a complete list of supported models and variants, see the [Ollama model library](https://ollama.com/library).
 :::
 
-### 1. Deploy ollama using docker
+### 1. Deploy Ollama using Docker
 
 ```bash
 sudo docker run --name ollama -p 11434:11434 ollama/ollama
@@ -36,14 +36,14 @@ time=2024-12-02T02:20:21.360Z level=INFO source=routes.go:1248 msg="Listening on
 time=2024-12-02T02:20:21.360Z level=INFO source=common.go:49 msg="Dynamic LLM libraries" runners="[cpu cpu_avx cpu_avx2 cuda_v11 cuda_v12]"
 ```
 
-Ensure ollama is listening on all IP address:
+Ensure Ollama is listening on all IP address:
 ```bash
 sudo ss -tunlp | grep 11434
 tcp   LISTEN 0      4096                  0.0.0.0:11434      0.0.0.0:*    users:(("docker-proxy",pid=794507,fd=4))
 tcp   LISTEN 0      4096                     [::]:11434         [::]:*    users:(("docker-proxy",pid=794513,fd=4))
 ```
 
-Pull models as you need. It's recommended to start with `llama3.2` (a 3B chat model) and `bge-m3` (a 567M embedding model):
+Pull models as you need. We recommend that you start with `llama3.2` (a 3B chat model) and `bge-m3` (a 567M embedding model):
 ```bash
 sudo docker exec ollama ollama pull llama3.2
 pulling dde5aa3fc5ff... 100% ▕████████████████▏ 2.0 GB
@@ -58,20 +58,20 @@ success
 
 ### 2. Ensure Ollama is accessible
 
-If RAGFlow runs in Docker and Ollama runs on the same host machine, check if ollama is accessible from inside the RAGFlow container:
+- If RAGFlow runs in Docker and Ollama runs on the same host machine, check if Ollama is accessible from inside the RAGFlow container:
 ```bash
 sudo docker exec -it ragflow-server bash
-root@8136b8c3e914:/ragflow# curl  http://host.docker.internal:11434/
+curl  http://host.docker.internal:11434/
 Ollama is running
 ```
 
-If RAGFlow runs from source code and Ollama runs on the same host machine, check if ollama is accessible from RAGFlow host machine:
+- If RAGFlow is launched from source code and Ollama runs on the same host machine as RAGFlow, check if Ollama is accessible from RAGFlow's host machine:
 ```bash
 curl  http://localhost:11434/
 Ollama is running
 ```
 
-If RAGFlow and Ollama run on different machines, check if ollama is accessible from RAGFlow host machine:
+- If RAGFlow and Ollama run on different machines, check if Ollama is accessible from RAGFlow's host machine:
 ```bash
 curl  http://${IP_OF_OLLAMA_MACHINE}:11434/
 Ollama is running
@@ -88,8 +88,8 @@ In RAGFlow, click on your logo on the top right of the page **>** **Model provid
 
 In the popup window, complete basic settings for Ollama:
 
-1. Ensure model name and type match those been pulled at step 1, For example, (`llama3.2`, `chat`), (`bge-m3`, `embedding`).
-2. Ensure that the base URL match which been determined at step 2.
+1. Ensure that your model name and type match those been pulled at step 1 (Deploy Ollama using Docker). For example, (`llama3.2` and `chat`) or (`bge-m3` and `embedding`).
+2. Ensure that the base URL match the URL determined at step 2 (Ensure Ollama is accessible).
 3. OPTIONAL: Switch on the toggle under **Does it support Vision?** if your model includes an image-to-text model.
 
 
@@ -104,15 +104,12 @@ Max retries exceeded with url: /api/chat (Caused by NewConnectionError('<urllib3
 
 Click on your logo **>** **Model providers** **>** **System Model Settings** to update your model:
    
-*You should now be able to find **llama3.2** from the dropdown list under **Chat model**, and **bge-m3** from the dropdown list under **Embedding model**.*
-
-> If your local model is an embedding model, you should find your local model under **Embedding model**.
+- *You should now be able to find **llama3.2** from the dropdown list under **Chat model**, and **bge-m3** from the dropdown list under **Embedding model**.*
+- _If your local model is an embedding model, you should find it under **Embedding model**._
 
 ### 7. Update Chat Configuration
 
-Update your chat model accordingly in **Chat Configuration**:
-
-> If your local model is an embedding model, update it on the configuration page of your knowledge base.
+Update your model(s) accordingly in **Chat Configuration**.
 
 ## Deploy a local model using Xinference