feat: add response body schemas for various API endpoints

--bug=1060032 --user=王孝刚 【桂物智慧】 api 调用客户端修改对话摘要接口，报错403，没有访问权限 https://www.tapd.cn/57709429/s/1757188

.dockerignore

@@ -1 +1,2 @@
-.git*
+.git*
+.idea*

.github/ISSUE_TEMPLATE/bug.yml

@@ -0,0 +1,50 @@
+name: 'Bug Report'
+description: 'Report an Bug'
+title: "[Bug] "
+assignees: zyyfit
+body:
+  - type: markdown
+    attributes:
+      value: "## Contact Information"
+  - type: input
+    validations:
+      required: false
+    attributes:
+      label: "Contact Information"
+      description: "The ways to quickly contact you: WeChat group number and nickname, email, etc."
+  - type: markdown
+    attributes:
+      value: "## Environment Information"
+  - type: input
+    validations:
+      required: true
+    attributes:
+      label: "MaxKB Version"
+      description: "Log in to the MaxKB Web Console and check the current version on the `About` page in the top right corner."
+  - type: markdown
+    attributes:
+      value: "## Detailed information"
+  - type: textarea
+    attributes:
+      label: "Problem Description"
+      description: "Briefly describe the issue you’ve encountered."
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: "Steps to Reproduce"
+      description: "How can this issue be reproduced."
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: "The expected correct result"
+  - type: textarea
+    attributes:
+      label: "Related log output"
+      description: "Please paste any relevant log output here. It will automatically be formatted as code, so no backticks are necessary."
+      render: shell
+  - type: textarea
+    attributes:
+      label: "Additional Information"
+      description: "If you have any additional information to provide, you can include it here (screenshots, videos, etc., are welcome)."

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,5 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Questions & Discussions
+    url: https://github.com/1Panel-dev/MaxKB/discussions
+    about: Raise questions about the installation, deployment, use and other aspects of the project.

.github/ISSUE_TEMPLATE/feature.yml

@@ -0,0 +1,29 @@
+name: 'Feature Request'
+description: 'Suggest an idea'
+title: '[Feature] '
+assignees: baixin513
+body:
+  - type: markdown
+    attributes:
+      value: "## Environment Information"
+  - type: input
+    validations:
+      required: true
+    attributes:
+      label: "MaxKB Version"
+      description: "Log in to the MaxKB Web Console and check the current version on the `About` page in the top right corner."
+  - type: markdown
+    attributes:
+      value: "## Detailed information"
+  - type: textarea
+    attributes:
+      label: "Please describe your needs or suggestions for improvements"
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: "Please describe the solution you suggest"
+  - type: textarea
+    attributes:
+      label: "Additional Information"
+      description: "If you have any additional information to provide, you can include it here (screenshots, videos, etc., are welcome)."

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,9 @@
+#### What this PR does / why we need it?
+
+#### Summary of your change
+
+#### Please indicate you've done the following:
+
+- [ ] Made sure tests are passing and test coverage is added if needed.
+- [ ] Made sure commit message follow the rule of [Conventional Commits specification](https://www.conventionalcommits.org/).
+- [ ] Considered the docs impact and opened a new docs issue or PR with docs changes if needed.

.github/dependabot.yml

@@ -0,0 +1,17 @@
+version: 2
+updates:
+  - package-ecosystem: "pip"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+      timezone: "Asia/Shanghai"
+      day: "friday"
+    target-branch: "v2"
+    groups:
+      python-dependencies:
+        patterns:
+          - "*"
+#    ignore:
+#      - dependency-name: "pymupdf"
+#        versions: ["*"]
+

.github/workflows/build-and-push-python-pg.yml

@@ -14,7 +14,7 @@ on:
           - linux/amd64,linux/arm64
 jobs:
   build-and-push-python-pg-to-ghcr:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
       - name: Check Disk Space
         run: df -h
@@ -39,7 +39,7 @@ jobs:
         run: |
           DOCKER_IMAGE=ghcr.io/1panel-dev/maxkb-python-pg
           DOCKER_PLATFORMS=${{ github.event.inputs.architecture }}
-          TAG_NAME=python3.11.8-pg15.3
+          TAG_NAME=python3.11-pg15.8
           DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME} --tag ${DOCKER_IMAGE}:latest"
           echo ::set-output name=docker_image::${DOCKER_IMAGE}
           echo ::set-output name=version::${TAG_NAME}
@@ -50,6 +50,9 @@ jobs:
             ${DOCKER_IMAGE_TAGS} .
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
+        with:
+          # Until https://github.com/tonistiigi/binfmt/issues/215
+          image: tonistiigi/binfmt:qemu-v7.0.0-28
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
       - name: Login to GitHub Container Registry

.github/workflows/build-and-push-vector-model.yml

@@ -5,7 +5,7 @@ on:
     inputs:
       dockerImageTag:
         description: 'Docker Image Tag'
-        default: 'v1.0.0'
+        default: 'v1.0.1'
         required: true
       architecture:
         description: 'Architecture'
@@ -19,7 +19,7 @@ on:
 
 jobs:
   build-and-push-vector-model-to-ghcr:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
       - name: Check Disk Space
         run: df -h
@@ -55,6 +55,9 @@ jobs:
             ${DOCKER_IMAGE_TAGS} .
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
+        with:
+          # Until https://github.com/tonistiigi/binfmt/issues/215
+          image: tonistiigi/binfmt:qemu-v7.0.0-28
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
       - name: Login to GitHub Container Registry

.github/workflows/build-and-push.yml

@@ -1,12 +1,19 @@
 name: build-and-push
 
+run-name: 构建镜像并推送仓库 ${{ github.event.inputs.dockerImageTag }} (${{ github.event.inputs.registry }})
+
 on:
   workflow_dispatch:
     inputs:
       dockerImageTag:
-        description: 'Docker Image Tag'
-        default: 'v1.0.0-dev'
+        description: 'Image Tag'
+        default: 'v1.10.7-dev'
         required: true
+      dockerImageTagWithLatest:
+        description: '是否发布latest tag（正式发版时选择，测试版本切勿选择）'
+        default: false
+        required: true
+        type: boolean
       architecture:
         description: 'Architecture'
         required: true
@@ -19,11 +26,11 @@ on:
       registry:
         description: 'Push To Registry'
         required: true
-        default: 'dockerhub, fit2cloud-registry'
+        default: 'fit2cloud-registry'
         type: choice
         options:
-          - dockerhub
           - fit2cloud-registry
+          - dockerhub
           - dockerhub, fit2cloud-registry
 
 jobs:
@@ -48,19 +55,21 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v4
         with:
-          ref: main
+          ref: ${{ github.ref_name }}
       - name: Prepare
         id: prepare
         run: |
-          DOCKER_IMAGE=registry-hkproxy.fit2cloud.com/maxkb/maxkb
+          DOCKER_IMAGE=${{ secrets.FIT2CLOUD_REGISTRY_HOST }}/maxkb/maxkb
           DOCKER_PLATFORMS=${{ github.event.inputs.architecture }}
           TAG_NAME=${{ github.event.inputs.dockerImageTag }}
-          if [[ ${TAG_NAME} == *dev* ]]; then
-            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME}"
+          TAG_NAME_WITH_LATEST=${{ github.event.inputs.dockerImageTagWithLatest }}
+          if [[ ${TAG_NAME_WITH_LATEST} == 'true' ]]; then            
+            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME} --tag ${DOCKER_IMAGE}:${TAG_NAME%%.*}"
           else
-            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME} --tag ${DOCKER_IMAGE}:latest"
+            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME}"
           fi
-          echo ::set-output name=buildx_args::--platform ${DOCKER_PLATFORMS} --no-cache \
+          echo ::set-output name=buildx_args::--platform ${DOCKER_PLATFORMS} --memory-swap -1 \
+            --build-arg DOCKER_IMAGE_TAG=${{ github.event.inputs.dockerImageTag }} --build-arg BUILD_AT=$(TZ=Asia/Shanghai date +'%Y-%m-%dT%H:%M') --build-arg GITHUB_COMMIT=`git rev-parse --short HEAD` --no-cache \
             ${DOCKER_IMAGE_TAGS} .
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
@@ -75,11 +84,12 @@ jobs:
       - name: Login to FIT2CLOUD Registry
         uses: docker/login-action@v3
         with:
-          registry: registry-hkproxy.fit2cloud.com
+          registry: ${{ secrets.FIT2CLOUD_REGISTRY_HOST }}
           username: ${{ secrets.FIT2CLOUD_REGISTRY_USERNAME }}
           password: ${{ secrets.FIT2CLOUD_REGISTRY_PASSWORD }}
       - name: Docker Buildx (build-and-push)
         run: |
+          sudo sync && echo 3 | sudo tee /proc/sys/vm/drop_caches && free -m
           docker buildx build --output "type=image,push=true" ${{ steps.prepare.outputs.buildx_args }} -f installer/Dockerfile
 
   build-and-push-to-dockerhub:
@@ -103,21 +113,22 @@ jobs:
       - name: Checkout
         uses: actions/checkout@v4
         with:
-          ref: main
+          ref: ${{ github.ref_name }}
       - name: Prepare
         id: prepare
         run: |
           DOCKER_IMAGE=1panel/maxkb
           DOCKER_PLATFORMS=${{ github.event.inputs.architecture }}
-          TAG_NAME=${{ github.event.inputs.dockerImageTag }}
-          if [[ ${TAG_NAME} == *dev* ]]; then
-            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME}"
+          TAG_NAME=${{ github.event.inputs.dockerImageTag }}          
+          TAG_NAME_WITH_LATEST=${{ github.event.inputs.dockerImageTagWithLatest }}
+          if [[ ${TAG_NAME_WITH_LATEST} == 'true' ]]; then            
+            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME} --tag ${DOCKER_IMAGE}:${TAG_NAME%%.*}"
           else
-            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME} --tag ${DOCKER_IMAGE}:latest"
+            DOCKER_IMAGE_TAGS="--tag ${DOCKER_IMAGE}:${TAG_NAME}"
           fi
-          echo ::set-output name=buildx_args::--platform ${DOCKER_PLATFORMS} --no-cache \
+          echo ::set-output name=buildx_args::--platform ${DOCKER_PLATFORMS} --memory-swap -1 \
+            --build-arg DOCKER_IMAGE_TAG=${{ github.event.inputs.dockerImageTag }} --build-arg BUILD_AT=$(TZ=Asia/Shanghai date +'%Y-%m-%dT%H:%M') --build-arg GITHUB_COMMIT=`git rev-parse --short HEAD` --no-cache \
             ${DOCKER_IMAGE_TAGS} .
-          env
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
       - name: Set up Docker Buildx
@@ -135,4 +146,5 @@ jobs:
           password: ${{ secrets.DOCKERHUB_TOKEN }}
       - name: Docker Buildx (build-and-push)
         run: |
+          sudo sync && echo 3 | sudo tee /proc/sys/vm/drop_caches && free -m
           docker buildx build --output "type=image,push=true" ${{ steps.prepare.outputs.buildx_args }} -f installer/Dockerfile

.github/workflows/create-pr-from-push.yml

@@ -0,0 +1,17 @@
+on:
+  push:
+    branches:    
+      - 'pr@**'
+      - 'repr@**'
+
+name: 针对特定分支名自动创建 PR
+
+jobs:
+  generic_handler:
+    name: 自动创建 PR
+    runs-on: ubuntu-latest
+    steps:
+      - name: Create pull request
+        uses: jumpserver/action-generic-handler@master
+        env:
+          GITHUB_TOKEN: ${{ secrets.GH_TOKEN }}

.github/workflows/issue-translator.yml

@@ -0,0 +1,14 @@
+name: Issue Translator
+on: 
+  issue_comment: 
+    types: [created]
+  issues: 
+    types: [opened]
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: usthe/issues-translate-action@v2.7
+        with:
+          IS_MODIFY_TITLE: true
+          BOT_GITHUB_TOKEN: ${{ secrets.FIT2CLOUDRD_LLM_CODE_REVIEW_TOKEN }}

.github/workflows/llm-code-review.yml

@@ -0,0 +1,28 @@
+name: LLM Code Review
+
+permissions:
+  contents: read
+  pull-requests: write
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+
+jobs:
+  llm-code-review:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: fit2cloud/LLM-CodeReview-Action@main
+        env:
+          GITHUB_TOKEN: ${{ secrets.FIT2CLOUDRD_LLM_CODE_REVIEW_TOKEN }}
+          OPENAI_API_KEY: ${{ secrets.ALIYUN_LLM_API_KEY }}
+          LANGUAGE: English
+          OPENAI_API_ENDPOINT: https://dashscope.aliyuncs.com/compatible-mode/v1
+          MODEL: qwen2.5-coder-3b-instruct
+          PROMPT: "Please check the following code for any irregularities, potential issues, or optimization suggestions, and provide your answers in English."
+          top_p: 1
+          temperature: 1
+          # max_tokens: 10000
+          MAX_PATCH_LENGTH: 10000
+          IGNORE_PATTERNS: "/node_modules,*.md,/dist,/.github"
+          FILE_PATTERNS: "*.java,*.go,*.py,*.vue,*.ts,*.js,*.css,*.scss,*.html"

.github/workflows/sync2gitee.yml

@@ -0,0 +1,16 @@
+name: sync2gitee
+on: [push]
+
+jobs:
+  repo-sync:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Mirror the Github organization repos to Gitee.
+        uses: Yikun/hub-mirror-action@master
+        with:
+          src: 'github/1Panel-dev'
+          dst: 'gitee/fit2cloud-feizhiyun'
+          dst_key: ${{ secrets.GITEE_PRIVATE_KEY }}
+          dst_token:  ${{ secrets.GITEE_TOKEN }}
+          static_list: "MaxKB"
+          force_update: true

.github/workflows/typos_check.yml

@@ -0,0 +1,18 @@
+name: Typos Check
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    types: [opened, synchronize, reopened]
+
+jobs:
+  run:
+    name: Spell Check with Typos
+    runs-on: ubuntu-latest
+    steps:
+    - name: Checkout Actions Repository
+      uses: actions/checkout@v2
+
+    - name: Check spelling
+      uses: crate-ci/typos@master

.gitignore

@@ -8,7 +8,8 @@
 *.factorypath
 
 # IntelliJ IDEA
-.idea
+.idea/*
+!.idea/icon.png
 *.iws
 *.iml
 *.ipr
@@ -41,7 +42,7 @@ share/python-wheels/
 MANIFEST
 
 # PyInstaller
-#  Usually these files are written by a python script froms a template
+#  Usually these files are written by a python script forms a template
 #  before PyInstaller builds the exe, so as to inject date/other infos into it.
 *.manifest
 *.spec
@@ -177,7 +178,10 @@ ui/node_modules
 ui/dist
 apps/static
 models/
+apps/xpack
+!apps/**/models/
 data
-.idea
 .dev
 poetry.lock
+apps/setting/models_provider/impl/*/icon/
+tmp/

.typos.toml

@@ -0,0 +1,4 @@
+[files]
+extend-exclude = [
+    'apps/setting/models_provider/impl/*/icon/*'
+]

CODE_OF_CONDUCT.md

@@ -0,0 +1,128 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and leaders pledge to make participation in our
+community a harassment-free experience for everyone, regardless of age, body
+size, visible or invisible disability, ethnicity, sex characteristics, gender
+identity and expression, level of experience, education, socio-economic status,
+nationality, personal appearance, race, religion, or sexual identity
+and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+## Our Standards
+
+Examples of behavior that contributes to a positive environment for our
+community include:
+
+* Demonstrating empathy and kindness toward other people
+* Being respectful of differing opinions, viewpoints, and experiences
+* Giving and gracefully accepting constructive feedback
+* Accepting responsibility and apologizing to those affected by our mistakes,
+  and learning from the experience
+* Focusing on what is best not just for us as individuals, but for the
+  overall community
+
+Examples of unacceptable behavior include:
+
+* The use of sexualized language or imagery, and sexual attention or
+  advances of any kind
+* Trolling, insulting or derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or email
+  address, without their explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of
+acceptable behavior and will take appropriate and fair corrective action in
+response to any behavior that they deem inappropriate, threatening, offensive,
+or harmful.
+
+Community leaders have the right and responsibility to remove, edit, or reject
+comments, commits, code, wiki edits, issues, and other contributions that are
+not aligned to this Code of Conduct, and will communicate reasons for moderation
+decisions when appropriate.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Examples of representing our community include using an official e-mail address,
+posting via an official social media account, or acting as an appointed
+representative at an online or offline event.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported to the community leaders responsible for enforcement at
+support@fit2cloud.com.
+All complaints will be reviewed and investigated promptly and fairly.
+
+All community leaders are obligated to respect the privacy and security of the
+reporter of any incident.
+
+## Enforcement Guidelines
+
+Community leaders will follow these Community Impact Guidelines in determining
+the consequences for any action they deem in violation of this Code of Conduct:
+
+### 1. Correction
+
+**Community Impact**: Use of inappropriate language or other behavior deemed
+unprofessional or unwelcome in the community.
+
+**Consequence**: A private, written warning from community leaders, providing
+clarity around the nature of the violation and an explanation of why the
+behavior was inappropriate. A public apology may be requested.
+
+### 2. Warning
+
+**Community Impact**: A violation through a single incident or series
+of actions.
+
+**Consequence**: A warning with consequences for continued behavior. No
+interaction with the people involved, including unsolicited interaction with
+those enforcing the Code of Conduct, for a specified period of time. This
+includes avoiding interactions in community spaces as well as external channels
+like social media. Violating these terms may lead to a temporary or
+permanent ban.
+
+### 3. Temporary Ban
+
+**Community Impact**: A serious violation of community standards, including
+sustained inappropriate behavior.
+
+**Consequence**: A temporary ban from any sort of interaction or public
+communication with the community for a specified period of time. No public or
+private interaction with the people involved, including unsolicited interaction
+with those enforcing the Code of Conduct, is allowed during this period.
+Violating these terms may lead to a permanent ban.
+
+### 4. Permanent Ban
+
+**Community Impact**: Demonstrating a pattern of violation of community
+standards, including sustained inappropriate behavior,  harassment of an
+individual, or aggression toward or disparagement of classes of individuals.
+
+**Consequence**: A permanent ban from any sort of public interaction within
+the community.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage],
+version 2.0, available at
+https://www.contributor-covenant.org/version/2/0/code_of_conduct.html.
+
+Community Impact Guidelines were inspired by [Mozilla's code of conduct
+enforcement ladder](https://github.com/mozilla/diversity).
+
+[homepage]: https://www.contributor-covenant.org
+
+For answers to common questions about this code of conduct, see the FAQ at
+https://www.contributor-covenant.org/faq. Translations are available at
+https://www.contributor-covenant.org/translations.

CONTRIBUTING.md

@@ -0,0 +1,30 @@
+# Contributing
+
+As a contributor, you should agree that:
+
+- The producer can adjust the open-source agreement to be more strict or relaxed as deemed necessary.
+- Your contributed code may be used for commercial purposes, including but not limited to its cloud business operations.
+
+## Create pull request
+PR are always welcome, even if they only contain small fixes like typos or a few lines of code. If there will be a significant effort, please document it as an issue and get a discussion going before starting to work on it.
+
+Please submit a PR broken down into small changes bit by bit. A PR consisting of a lot of features and code changes may be hard to review. It is recommended to submit PRs in an incremental fashion.
+
+This [development guideline](https://github.com/1Panel-dev/MaxKB/wiki/3-%E5%BC%80%E5%8F%91%E7%8E%AF%E5%A2%83%E6%90%AD%E5%BB%BA) contains information about repository structure, how to set up development environment, how to run it, and more.
+
+Note: If you split your pull request to small changes, please make sure any of the changes goes to master will not break anything. Otherwise, it can not be merged until this feature complete.
+
+## Report issues
+It is a great way to contribute by reporting an issue. Well-written and complete bug reports are always welcome! Please open an issue and follow the template to fill in required information.
+
+Before opening any issue, please look up the existing issues to avoid submitting a duplication.
+If you find a match, you can "subscribe" to it to get notified on updates. If you have additional helpful information about the issue, please leave a comment.
+
+When reporting issues, always include:
+
+* Which version you are using.
+* Steps to reproduce the issue.
+* Snapshots or log files if needed
+
+Because the issues are open to the public, when submitting files, be sure to remove any sensitive information, e.g. user name, password, IP address, and company name. You can
+replace those parts with "REDACTED" or other strings like "****".

README.md

@@ -1,56 +1,126 @@
 <p align="center"><img src= "https://github.com/1Panel-dev/maxkb/assets/52996290/c0694996-0eed-40d8-b369-322bf2a380bf" alt="MaxKB" width="300" /></p>
-<h3 align="center">基于 LLM 大语言模型的知识库问答系统</h3>
+<h3 align="center">Open-source platform for building enterprise-grade agents</h3>
+<h3 align="center">强大易用的企业级智能体平台</h3>
+<p align="center"><a href="https://trendshift.io/repositories/9113" target="_blank"><img src="https://trendshift.io/api/badge/repositories/9113" alt="1Panel-dev%2FMaxKB | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a></p>
 <p align="center">
-  <a href="https://www.gnu.org/licenses/old-licenses/gpl-3.0"><img src="https://img.shields.io/github/license/1Panel-dev/maxkb?color=%231890FF" alt="License: GPL v3"></a>
-  <a href="https://app.codacy.com/gh/1Panel-dev/maxkb?utm_source=github.com&utm_medium=referral&utm_content=1Panel-dev/maxkb&utm_campaign=Badge_Grade_Dashboard"><img src="https://app.codacy.com/project/badge/Grade/da67574fd82b473992781d1386b937ef" alt="Codacy"></a>
+  <a href="https://www.gnu.org/licenses/gpl-3.0.html#license-text"><img src="https://img.shields.io/github/license/1Panel-dev/maxkb?color=%231890FF" alt="License: GPL v3"></a>
   <a href="https://github.com/1Panel-dev/maxkb/releases/latest"><img src="https://img.shields.io/github/v/release/1Panel-dev/maxkb" alt="Latest release"></a>
   <a href="https://github.com/1Panel-dev/maxkb"><img src="https://img.shields.io/github/stars/1Panel-dev/maxkb?color=%231890FF&style=flat-square" alt="Stars"></a>    
+  <a href="https://hub.docker.com/r/1panel/maxkb"><img src="https://img.shields.io/docker/pulls/1panel/maxkb?label=downloads" alt="Download"></a><br/>
+ [<a href="/README_CN.md">中文(简体)</a>] | [<a href="/README.md">English</a>] 
 </p>
 <hr/>
 
-MaxKB 是一款基于 LLM 大语言模型的知识库问答系统。   
+MaxKB = Max Knowledge Brain, it is an open-source platform for building enterprise-grade agents. MaxKB integrates Retrieval-Augmented Generation (RAG) pipelines, supports robust workflows, and provides advanced MCP tool-use capabilities. MaxKB is widely applied in scenarios such as intelligent customer service, corporate internal knowledge bases, academic research, and education.
 
-- **多模型支持**：支持对接主流的大模型，包括本地私有大模型（如 Llama 2）、Azure OpenAI 和百度千帆大模型等；
-- **开箱即用**：支持直接上传文档、自动爬取在线文档，支持文本自动拆分、向量化，智能问答交互体验好；
-- **无缝嵌入**：支持零编码快速嵌入到第三方业务系统。
+- **RAG Pipeline**: Supports direct uploading of documents / automatic crawling of online documents, with features for automatic text splitting, vectorization. This effectively reduces hallucinations in large models, providing a superior smart Q&A interaction experience.
+- **Agentic Workflow**: Equipped with a powerful workflow engine, function library and MCP tool-use, enabling the orchestration of AI processes to meet the needs of complex business scenarios. 
+- **Seamless Integration**: Facilitates zero-coding rapid integration into third-party business systems, quickly equipping existing systems with intelligent Q&A capabilities to enhance user satisfaction.
+- **Model-Agnostic**: Supports various large models, including private models (such as DeepSeek, Llama, Qwen, etc.) and public models (like OpenAI, Claude, Gemini, etc.).
+- **Multi Modal**: Native support for input and output text, image, audio and video.
 
-## 快速开始
+## Quick start
 
-```
-docker run -d --name=maxkb -p 8080:8080 -v ~/.maxkb:/var/lib/postgresql/data 1panel/maxkb
+Execute the script below to start a MaxKB container using Docker:
+
+```bash
+docker run -d --name=maxkb --restart=always -p 8080:8080 -v ~/.maxkb:/var/lib/postgresql/data -v ~/.python-packages:/opt/maxkb/app/sandbox/python-packages 1panel/maxkb
 ```
 
-也可以通过 [1Panel 应用商店](https://apps.fit2cloud.com/1panel) 快速部署 MaxKB + Ollama + Llama 2，30 分钟内即可上线基于本地大模型的知识库问答系统。
+Access MaxKB web interface at `http://your_server_ip:8080` with default admin credentials:
 
-## UI 展示
+- username: admin
+- password: MaxKB@123..
+
+中国用户如遇到 Docker 镜像 Pull 失败问题，请参照该 [离线安装文档](https://maxkb.cn/docs/installation/offline_installtion/) 进行安装。
+
+## Screenshots
 
 <table style="border-collapse: collapse; border: 1px solid black;">
   <tr>
-    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/80892890/2b893a25-ae46-48da-b6a1-61d23015565e" alt="MaxKB Demo1"   /></td>
-    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/80892890/3e50e7ff-cdc4-4a37-b430-d84975f11d4e" alt="MaxKB Demo2"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://maxkb.hk/images/overview.png" alt="MaxKB Demo1"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://maxkb.hk/images/screenshot-models.png" alt="MaxKB Demo2"   /></td>
   </tr>
   <tr>
-    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/80892890/dfdcc03f-ef36-4f75-bb82-797c0f9da1ad" alt="MaxKB Demo3"   /></td>
-    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/80892890/884a9db1-3f93-4013-bc8f-a3f0dbcfeb2f" alt="MaxKB Demo4"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://maxkb.hk/images/screenshot-knowledge.png" alt="MaxKB Demo3"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://maxkb.hk/images/screenshot-function.png" alt="MaxKB Demo4"   /></td>
   </tr>
 </table>
 
-## 社区交流
+## Technical stack
 
-[论坛](https://bbs.fit2cloud.com/c/mk/11)
+- Frontend：[Vue.js](https://vuejs.org/)
+- Backend：[Python / Django](https://www.djangoproject.com/)
+- LLM Framework：[LangChain](https://www.langchain.com/)
+- Database：[PostgreSQL + pgvector](https://www.postgresql.org/)
 
-## 技术栈
+## Feature Comparison
 
--   前端：[Vue.js](https://cn.vuejs.org/)
--   后端：[Python / Django](https://www.djangoproject.com/)
--   Langchain：[Langchain](https://www.langchain.com/)
--   向量数据库：[PostgreSQL / pgvector](https://www.postgresql.org/)
--   大模型：Azure OpenAI、百度千帆大模型、[Ollama](https://github.com/ollama/ollama)
+<table style="width: 100%;">
+  <tr>
+    <th align="center">Feature</th>
+    <th align="center">LangChain</th>
+    <th align="center">Dify.AI</th>
+    <th align="center">Flowise</th>
+    <th align="center">MaxKB <br>（Built upon LangChain）</th>
+  </tr>
+  <tr>
+    <td align="center">Supported LLMs</td>
+    <td align="center">Rich Variety</td>
+    <td align="center">Rich Variety</td>
+    <td align="center">Rich Variety</td>
+    <td align="center">Rich Variety</td>
+  </tr>
+  <tr>
+    <td align="center">RAG Engine</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+  </tr>
+  <tr>
+    <td align="center">Agent</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">❌</td>
+    <td align="center">✅</td>
+  </tr>
+  <tr>
+    <td align="center">Workflow</td>
+    <td align="center">❌</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+  </tr>
+  <tr>
+    <td align="center">Observability</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">❌</td>
+    <td align="center">✅</td>
+  </tr>
+  <tr>
+    <td align="center">SSO/Access control</td>
+    <td align="center">❌</td>
+    <td align="center">✅</td>
+    <td align="center">❌</td>
+    <td align="center">✅ (Pro)</td>
+  </tr>
+  <tr>
+    <td align="center">On-premise Deployment</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+    <td align="center">✅</td>
+  </tr>
+</table>
+
+## Star History
+
+[![Star History Chart](https://api.star-history.com/svg?repos=1Panel-dev/MaxKB&type=Date)](https://star-history.com/#1Panel-dev/MaxKB&Date)
 
 ## License
 
-Copyright (c) 2014-2024 飞致云 FIT2CLOUD, All rights reserved.
-
 Licensed under The GNU General Public License version 3 (GPLv3)  (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
 
 <https://www.gnu.org/licenses/gpl-3.0.html>

README_CN.md

@@ -0,0 +1,89 @@
+<p align="center"><img src= "https://github.com/1Panel-dev/maxkb/assets/52996290/c0694996-0eed-40d8-b369-322bf2a380bf" alt="MaxKB" width="300" /></p>
+<h3 align="center">强大易用的企业级智能体平台</h3>
+<p align="center">
+    <a href="https://trendshift.io/repositories/9113" target="_blank"><img src="https://trendshift.io/api/badge/repositories/9113" alt="1Panel-dev%2FMaxKB | Trendshift" style="width: 250px; height: auto;" /></a>
+</p>
+<p align="center">
+  <a href="README_EN.md"><img src="https://img.shields.io/badge/English_README-blue" alt="English README"></a>
+  <a href="https://www.gnu.org/licenses/gpl-3.0.html#license-text"><img src="https://img.shields.io/github/license/1Panel-dev/maxkb?color=%231890FF" alt="License: GPL v3"></a>
+  <a href="https://github.com/1Panel-dev/maxkb/releases/latest"><img src="https://img.shields.io/github/v/release/1Panel-dev/maxkb" alt="Latest release"></a>
+  <a href="https://github.com/1Panel-dev/maxkb"><img src="https://img.shields.io/github/stars/1Panel-dev/maxkb?style=flat-square" alt="Stars"></a>
+  <a href="https://hub.docker.com/r/1panel/maxkb"><img src="https://img.shields.io/docker/pulls/1panel/maxkb?label=downloads" alt="Download"></a>
+  <a href="https://gitee.com/fit2cloud-feizhiyun/MaxKB"><img src="https://gitee.com/fit2cloud-feizhiyun/MaxKB/badge/star.svg?theme=gvp" alt="Gitee Stars"></a>
+  <a href="https://gitcode.com/feizhiyun/MaxKB"><img src="https://gitcode.com/feizhiyun/MaxKB/star/badge.svg" alt="GitCode Stars"></a>
+</p>
+<hr/>
+
+MaxKB = Max Knowledge Brain，是一款强大易用的企业级智能体平台，支持 RAG 检索增强生成、工作流编排、MCP 工具调用能力。MaxKB 支持对接各种主流大语言模型，广泛应用于智能客服、企业内部知识库问答、员工助手、学术研究与教育等场景。
+
+- **RAG 检索增强生成**：高效搭建本地 AI 知识库，支持直接上传文档 / 自动爬取在线文档，支持文本自动拆分、向量化，有效减少大模型幻觉，提升问答效果；
+- **灵活编排**：内置强大的工作流引擎、函数库和 MCP 工具调用能力，支持编排 AI 工作过程，满足复杂业务场景下的需求；
+- **无缝嵌入**：支持零编码快速嵌入到第三方业务系统，让已有系统快速拥有智能问答能力，提高用户满意度；
+- **模型中立**：支持对接各种大模型，包括本地私有大模型（DeepSeek R1 / Llama 3 / Qwen 2 等）、国内公共大模型（通义千问 / 腾讯混元 / 字节豆包 / 百度千帆 / 智谱 AI / Kimi 等）和国外公共大模型（OpenAI / Claude / Gemini 等）。
+
+MaxKB 三分钟视频介绍：https://www.bilibili.com/video/BV18JypYeEkj/
+
+## 快速开始
+
+```
+# Linux 机器
+docker run -d --name=maxkb --restart=always -p 8080:8080 -v ~/.maxkb:/var/lib/postgresql/data -v ~/.python-packages:/opt/maxkb/app/sandbox/python-packages registry.fit2cloud.com/maxkb/maxkb
+
+# Windows 机器
+docker run -d --name=maxkb --restart=always -p 8080:8080 -v C:/maxkb:/var/lib/postgresql/data -v C:/python-packages:/opt/maxkb/app/sandbox/python-packages registry.fit2cloud.com/maxkb/maxkb
+
+# 用户名: admin
+# 密码: MaxKB@123..
+```
+
+- 你也可以通过 [1Panel 应用商店](https://apps.fit2cloud.com/1panel) 快速部署 MaxKB；
+- 如果是内网环境，推荐使用 [离线安装包](https://community.fit2cloud.com/#/products/maxkb/downloads) 进行安装部署；
+- MaxKB 产品版本分为社区版和专业版，详情请参见：[MaxKB 产品版本对比](https://maxkb.cn/pricing.html)；
+- 如果您需要向团队介绍 MaxKB，可以使用这个 [官方 PPT 材料](https://maxkb.cn/download/introduce-maxkb_202503.pdf)。
+
+如你有更多问题，可以查看使用手册，或者通过论坛与我们交流。
+
+- [案例展示](USE-CASES.md)
+- [使用手册](https://maxkb.cn/docs/)
+- [论坛求助](https://bbs.fit2cloud.com/c/mk/11)
+- 技术交流群
+
+<image height="150px" width="150px" src="https://github.com/1Panel-dev/MaxKB/assets/52996290/a083d214-02be-4178-a1db-4f428124153a"/>
+
+## UI 展示
+
+<table style="border-collapse: collapse; border: 1px solid black;">
+  <tr>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/52996290/d87395fa-a8d7-401c-82bf-c6e475d10ae9" alt="MaxKB Demo1"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/1Panel-dev/MaxKB/assets/52996290/47c35ee4-3a3b-4bd4-9f4f-ee20788b2b9a" alt="MaxKB Demo2"   /></td>
+  </tr>
+  <tr>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/user-attachments/assets/9a1043cb-fa62-4f71-b9a3-0b46fa59a70e" alt="MaxKB Demo3"   /></td>
+    <td style="padding: 5px;background-color:#fff;"><img src= "https://github.com/user-attachments/assets/3407ce9a-779c-4eb4-858e-9441a2ddc664" alt="MaxKB Demo4"   /></td>
+  </tr>
+</table>
+
+## 技术栈
+
+- 前端：[Vue.js](https://cn.vuejs.org/)
+- 后端：[Python / Django](https://www.djangoproject.com/)
+- LangChain：[LangChain](https://www.langchain.com/)
+- 向量数据库：[PostgreSQL / pgvector](https://www.postgresql.org/)
+
+## 飞致云的其他明星项目
+
+- [1Panel](https://github.com/1panel-dev/1panel/) - 现代化、开源的 Linux 服务器运维管理面板
+- [JumpServer](https://github.com/jumpserver/jumpserver/) - 广受欢迎的开源堡垒机
+- [DataEase](https://github.com/dataease/dataease/) - 人人可用的开源数据可视化分析工具
+- [MeterSphere](https://github.com/metersphere/metersphere/) - 新一代的开源持续测试工具
+- [Halo](https://github.com/halo-dev/halo/) - 强大易用的开源建站工具
+
+## License
+
+Copyright (c) 2014-2025 飞致云 FIT2CLOUD, All rights reserved.
+
+Licensed under The GNU General Public License version 3 (GPLv3)  (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
+
+<https://www.gnu.org/licenses/gpl-3.0.html>
+
+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.

SECURITY.md

@@ -0,0 +1,17 @@
+# 安全说明
+
+如果您发现安全问题，请直接联系我们：
+
+- support@fit2cloud.com
+- 400-052-0755
+
+感谢您的支持！
+
+# Security Policy
+
+All security bugs should be reported to the contact as below:
+
+- support@fit2cloud.com
+- 400-052-0755
+
+Thanks for your support!

USE-CASES.md

@@ -0,0 +1,39 @@
+<h3 align="center">MaxKB 应用案例，持续更新中...</h3>
+
+------------------------------
+
+- [MaxKB 应用案例：中国农业大学-小鹉哥](https://mp.weixin.qq.com/s/4g_gySMBQZCJ9OZ-yBkmvw)
+- [MaxKB 应用案例：东北财经大学-小银杏](https://mp.weixin.qq.com/s/3BoxkY7EMomMmmvFYxvDIA)
+- [MaxKB 应用案例：中铁水务](https://mp.weixin.qq.com/s/voNAddbK2CJOrJJs1ewZ8g)
+- [MaxKB 应用案例：解放军总医院](https://mp.weixin.qq.com/s/ETrZC-vrA4Aap0eF-15EeQ)
+- [MaxKB 应用案例：无锡市数据局](https://mp.weixin.qq.com/s/enfUFLevvL_La74PQ0kIXw)
+- [MaxKB 应用案例：中核西仪研究院-西仪睿答](https://mp.weixin.qq.com/s/CbKr4mev8qahKLAtV6Dxdg)
+- [MaxKB 应用案例：南京中医药大学](https://mp.weixin.qq.com/s/WUmAKYbZjp3272HIecpRFA)
+- [MaxKB 应用案例：西北电力设计院-AI数字助理Memex](https://mp.weixin.qq.com/s/ezHFdB7C7AVL9MTtDwYGSA)
+- [MaxKB 应用案例：西安国际医院中心医院-国医小助](https://mp.weixin.qq.com/s/DSOUvwrQrxbqQxKBilTCFQ)
+- [MaxKB 应用案例：华莱士智能AI客服助手上线啦！](https://www.bilibili.com/video/BV1hQtVeXEBL)
+- [MaxKB 应用案例：把医疗行业知识转化为知识库问答助手！](https://www.bilibili.com/video/BV157wme9EgB)
+- [MaxKB 应用案例：会展AI智能客服体验](https://www.bilibili.com/video/BV1J7BqY6EKA)
+- [MaxKB 应用案例：孩子要上幼儿园了，AI 智能助手择校好帮手](https://www.bilibili.com/video/BV1wKrhYvEer)
+- [MaxKB 应用案例：产品使用指南AI助手，新手小白也能轻松搞定！](https://www.bilibili.com/video/BV1Yz6gYtEqX)
+- [MaxKB 应用案例：生物医药AI客服智能体验!](https://www.bilibili.com/video/BV13JzvYsE3e)
+- [MaxKB 应用案例：高校行政管理AI小助手](https://www.bilibili.com/video/BV1yvBMYvEdy)
+- [MaxKB 应用案例：岳阳市人民医院-OA小助手](https://mp.weixin.qq.com/s/O94Qo3UH-MiUtDdWCVg8sQ)
+- [MaxKB 应用案例：常熟市第一人民医院](https://mp.weixin.qq.com/s/s5XXGTR3_MUo41NbJ8WzZQ)
+- [MaxKB 应用案例：华北水利水电大学](https://mp.weixin.qq.com/s/PoOFAcMCr9qJdvSj8c08qg)
+- [MaxKB 应用案例：唐山海事局-“小海”AI语音助手](https://news.qq.com/rain/a/20250223A030BE00)
+- [MaxKB 应用案例：湖南汉寿政务](http://hsds.hsdj.gov.cn:19999/ui/chat/a2c976736739aadc)
+- [MaxKB 应用案例：广州市妇女儿童医疗中心-AI医疗数据分类分级小助手](https://mp.weixin.qq.com/s/YHUMkUOAaUomBV8bswpK3g)
+- [MaxKB 应用案例：苏州热工研究院有限公司-维修大纲评估质量自查AI小助手](https://mp.weixin.qq.com/s/Ts5FQdnv7Tu9Jp7bvofCVA)
+- [MaxKB 应用案例：国核自仪系统工程有限公司-NuCON AI帮](https://mp.weixin.qq.com/s/HNPc7u5xVfGLJr8IQz3vjQ)
+- [MaxKB 应用案例：深圳通开启Deep Seek智能应用新篇章](https://mp.weixin.qq.com/s/SILN0GSescH9LyeQqYP0VQ)
+- [MaxKB 应用案例：南通智慧出行领跑长三角！首款接入DeepSeek的"畅行南通"APP上线AI新场景](https://mp.weixin.qq.com/s/WEC9UQ6msY0VS8LhTZh-Ew)
+- [MaxKB 应用案例：中船动力人工智能"智慧动力云助手"及首批数字员工正式上线](https://mp.weixin.qq.com/s/OGcEkjh9DzGO1Tkc9nr7qg)
+- [MaxKB 应用案例：AI+矿山：DeepSeek助力绿色智慧矿山智慧“升级”](https://mp.weixin.qq.com/s/SZstxTvVoLZg0ECbZbfpIA)
+- [MaxKB 应用案例：DeepSeek落地弘盛铜业：国产大模型点亮"黑灯工厂"新引擎](https://mp.weixin.qq.com/s/Eczdx574MS5RMF7WfHN7_A)
+- [MaxKB 应用案例：拥抱智能时代！中国五矿以 “AI+”赋能企业发展](https://mp.weixin.qq.com/s/D5vBtlX2E81pWE3_2OgWSw)
+- [MaxKB 应用案例：DeepSeek赋能中冶武勘AI智能体](https://mp.weixin.qq.com/s/8m0vxGcWXNdZazziQrLyxg)
+- [MaxKB 应用案例：重磅！陕西广电网络“秦岭云”平台实现DeepSeek本地化部署](https://mp.weixin.qq.com/s/ZKmEU_wWShK1YDomKJHQeA)
+- [MaxKB 应用案例：粤海集团完成DeepSeek私有化部署，助力集团智能化管理](https://mp.weixin.qq.com/s/2JbVp0-kr9Hfp-0whH4cvg)
+- [MaxKB 应用案例：建筑材料工业信息中心完成DeepSeek本地化部署，推动行业数智化转型新发展](https://mp.weixin.qq.com/s/HThGSnND3qDF8ySEqiM4jw)
+- [MaxKB 应用案例：一起DeepSeek！福建设计以AI大模型开启新篇章](https://mp.weixin.qq.com/s/m67e-H7iQBg3d24NM82UjA)

apps/application/chat_pipeline/I_base_chat_pipeline.py

@@ -8,7 +8,7 @@
 """
 import time
 from abc import abstractmethod
-from typing import Type, Dict
+from typing import Type
 
 from rest_framework import serializers
 
@@ -18,7 +18,8 @@ from dataset.models import Paragraph
 class ParagraphPipelineModel:
 
     def __init__(self, _id: str, document_id: str, dataset_id: str, content: str, title: str, status: str,
-                 is_active: bool, comprehensive_score: float, similarity: float, dataset_name: str, document_name: str):
+                 is_active: bool, comprehensive_score: float, similarity: float, dataset_name: str, document_name: str,
+                 hit_handling_method: str, directly_return_similarity: float, meta: dict = None):
         self.id = _id
         self.document_id = document_id
         self.dataset_id = dataset_id
@@ -30,6 +31,9 @@ class ParagraphPipelineModel:
         self.similarity = similarity
         self.dataset_name = dataset_name
         self.document_name = document_name
+        self.hit_handling_method = hit_handling_method
+        self.directly_return_similarity = directly_return_similarity
+        self.meta = meta
 
     def to_dict(self):
         return {
@@ -43,7 +47,8 @@ class ParagraphPipelineModel:
             'comprehensive_score': self.comprehensive_score,
             'similarity': self.similarity,
             'dataset_name': self.dataset_name,
-            'document_name': self.document_name
+            'document_name': self.document_name,
+            'meta': self.meta,
         }
 
     class builder:
@@ -53,6 +58,9 @@ class ParagraphPipelineModel:
             self.comprehensive_score = None
             self.document_name = None
             self.dataset_name = None
+            self.hit_handling_method = None
+            self.directly_return_similarity = 0.9
+            self.meta = {}
 
         def add_paragraph(self, paragraph):
             if isinstance(paragraph, Paragraph):
@@ -76,6 +84,14 @@ class ParagraphPipelineModel:
             self.document_name = document_name
             return self
 
+        def add_hit_handling_method(self, hit_handling_method):
+            self.hit_handling_method = hit_handling_method
+            return self
+
+        def add_directly_return_similarity(self, directly_return_similarity):
+            self.directly_return_similarity = directly_return_similarity
+            return self
+
         def add_comprehensive_score(self, comprehensive_score: float):
             self.comprehensive_score = comprehensive_score
             return self
@@ -84,6 +100,10 @@ class ParagraphPipelineModel:
             self.similarity = similarity
             return self
 
+        def add_meta(self, meta: dict):
+            self.meta = meta
+            return self
+
         def build(self):
             return ParagraphPipelineModel(str(self.paragraph.get('id')), str(self.paragraph.get('document_id')),
                                           str(self.paragraph.get('dataset_id')),
@@ -91,7 +111,8 @@ class ParagraphPipelineModel:
                                           self.paragraph.get('status'),
                                           self.paragraph.get('is_active'),
                                           self.comprehensive_score, self.similarity, self.dataset_name,
-                                          self.document_name)
+                                          self.document_name, self.hit_handling_method, self.directly_return_similarity,
+                                          self.meta)
 
 
 class IBaseChatPipelineStep:

apps/application/chat_pipeline/pipeline_manage.py

@@ -11,14 +11,18 @@ from functools import reduce
 from typing import List, Type, Dict
 
 from application.chat_pipeline.I_base_chat_pipeline import IBaseChatPipelineStep
+from common.handle.base_to_response import BaseToResponse
+from common.handle.impl.response.system_to_response import SystemToResponse
 
 
-class PiplineManage:
-    def __init__(self, step_list: List[Type[IBaseChatPipelineStep]]):
+class PipelineManage:
+    def __init__(self, step_list: List[Type[IBaseChatPipelineStep]],
+                 base_to_response: BaseToResponse = SystemToResponse()):
         # 步骤执行器
         self.step_list = [step() for step in step_list]
         # 上下文
         self.context = {'message_tokens': 0, 'answer_tokens': 0}
+        self.base_to_response = base_to_response
 
     def run(self, context: Dict = None):
         self.context['start_time'] = time.time()
@@ -33,13 +37,21 @@ class PiplineManage:
                                                 filter(lambda r: r is not None,
                                                        [row.get_details(self) for row in self.step_list])], {})
 
+    def get_base_to_response(self):
+        return self.base_to_response
+
     class builder:
         def __init__(self):
             self.step_list: List[Type[IBaseChatPipelineStep]] = []
+            self.base_to_response = SystemToResponse()
 
         def append_step(self, step: Type[IBaseChatPipelineStep]):
             self.step_list.append(step)
             return self
 
+        def add_base_to_response(self, base_to_response: BaseToResponse):
+            self.base_to_response = base_to_response
+            return self
+
         def build(self):
-            return PiplineManage(step_list=self.step_list)
+            return PipelineManage(step_list=self.step_list, base_to_response=self.base_to_response)

apps/application/chat_pipeline/step/chat_step/i_chat_step.py

@@ -9,21 +9,22 @@
 from abc import abstractmethod
 from typing import Type, List
 
+from django.utils.translation import gettext_lazy as _
 from langchain.chat_models.base import BaseChatModel
 from langchain.schema import BaseMessage
 from rest_framework import serializers
 
 from application.chat_pipeline.I_base_chat_pipeline import IBaseChatPipelineStep, ParagraphPipelineModel
-from application.chat_pipeline.pipeline_manage import PiplineManage
+from application.chat_pipeline.pipeline_manage import PipelineManage
+from application.serializers.application_serializers import NoReferencesSetting
 from common.field.common import InstanceField
 from common.util.field_message import ErrMessage
-from dataset.models import Paragraph
 
 
 class ModelField(serializers.Field):
     def to_internal_value(self, data):
         if not isinstance(data, BaseChatModel):
-            self.fail('模型类型错误', value=data)
+            self.fail(_('Model type error'), value=data)
         return data
 
     def to_representation(self, value):
@@ -33,7 +34,7 @@ class ModelField(serializers.Field):
 class MessageField(serializers.Field):
     def to_internal_value(self, data):
         if not isinstance(data, BaseMessage):
-            self.fail('message类型错误', value=data)
+            self.fail(_('Message type error'), value=data)
         return data
 
     def to_representation(self, value):
@@ -44,7 +45,7 @@ class PostResponseHandler:
     @abstractmethod
     def handler(self, chat_id, chat_record_id, paragraph_list: List[ParagraphPipelineModel], problem_text: str,
                 answer_text,
-                manage, step, padding_problem_text: str = None, **kwargs):
+                manage, step, padding_problem_text: str = None, client_id=None, **kwargs):
         pass
 
 
@@ -52,36 +53,47 @@ class IChatStep(IBaseChatPipelineStep):
     class InstanceSerializer(serializers.Serializer):
         # 对话列表
         message_list = serializers.ListField(required=True, child=MessageField(required=True),
-                                             error_messages=ErrMessage.list("对话列表"))
-        # 大语言模型
-        chat_model = ModelField(error_messages=ErrMessage.list("大语言模型"))
+                                             error_messages=ErrMessage.list(_("Conversation list")))
+        model_id = serializers.UUIDField(required=False, allow_null=True, error_messages=ErrMessage.uuid(_("Model id")))
         # 段落列表
-        paragraph_list = serializers.ListField(error_messages=ErrMessage.list("段落列表"))
+        paragraph_list = serializers.ListField(error_messages=ErrMessage.list(_("Paragraph List")))
         # 对话id
-        chat_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid("对话id"))
+        chat_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid(_("Conversation ID")))
         # 用户问题
-        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.uuid("用户问题"))
+        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.uuid(_("User Questions")))
         # 后置处理器
         post_response_handler = InstanceField(model_type=PostResponseHandler,
-                                              error_messages=ErrMessage.base("用户问题"))
+                                              error_messages=ErrMessage.base(_("Post-processor")))
         # 补全问题
-        padding_problem_text = serializers.CharField(required=False, error_messages=ErrMessage.base("补全问题"))
+        padding_problem_text = serializers.CharField(required=False,
+                                                     error_messages=ErrMessage.base(_("Completion Question")))
         # 是否使用流的形式输出
-        stream = serializers.BooleanField(required=False, error_messages=ErrMessage.base("流式输出"))
-        client_id = serializers.CharField(required=True, error_messages=ErrMessage.char("客户端id"))
-        client_type = serializers.CharField(required=True, error_messages=ErrMessage.char("客户端类型"))
+        stream = serializers.BooleanField(required=False, error_messages=ErrMessage.base(_("Streaming Output")))
+        client_id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Client id")))
+        client_type = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Client Type")))
+        # 未查询到引用分段
+        no_references_setting = NoReferencesSetting(required=True,
+                                                    error_messages=ErrMessage.base(_("No reference segment settings")))
+
+        user_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid(_("User ID")))
+
+        model_setting = serializers.DictField(required=True, allow_null=True,
+                                              error_messages=ErrMessage.dict(_("Model settings")))
+
+        model_params_setting = serializers.DictField(required=False, allow_null=True,
+                                                     error_messages=ErrMessage.dict(_("Model parameter settings")))
 
         def is_valid(self, *, raise_exception=False):
             super().is_valid(raise_exception=True)
             message_list: List = self.initial_data.get('message_list')
             for message in message_list:
                 if not isinstance(message, BaseMessage):
-                    raise Exception("message 类型错误")
+                    raise Exception(_("message type error"))
 
-    def get_step_serializer(self, manage: PiplineManage) -> Type[serializers.Serializer]:
+    def get_step_serializer(self, manage: PipelineManage) -> Type[serializers.Serializer]:
         return self.InstanceSerializer
 
-    def _run(self, manage: PiplineManage):
+    def _run(self, manage: PipelineManage):
         chat_result = self.execute(**self.context['step_args'], manage=manage)
         manage.context['chat_result'] = chat_result
 
@@ -89,8 +101,10 @@ class IChatStep(IBaseChatPipelineStep):
     def execute(self, message_list: List[BaseMessage],
                 chat_id, problem_text,
                 post_response_handler: PostResponseHandler,
-                chat_model: BaseChatModel = None,
+                model_id: str = None,
+                user_id: str = None,
                 paragraph_list=None,
-                manage: PiplineManage = None,
-                padding_problem_text: str = None, stream: bool = True, client_id=None, client_type=None, **kwargs):
+                manage: PipelineManage = None,
+                padding_problem_text: str = None, stream: bool = True, client_id=None, client_type=None,
+                no_references_setting=None, model_params_setting=None, model_setting=None, **kwargs):
         pass

apps/application/chat_pipeline/step/chat_step/impl/base_chat_step.py

@@ -6,7 +6,6 @@
     @date：2024/1/9 18:25
     @desc: 对话step Base实现
 """
-import json
 import logging
 import time
 import traceback
@@ -15,27 +14,44 @@ from typing import List
 
 from django.db.models import QuerySet
 from django.http import StreamingHttpResponse
+from django.utils.translation import gettext as _
 from langchain.chat_models.base import BaseChatModel
 from langchain.schema import BaseMessage
-from langchain.schema.messages import BaseMessageChunk, HumanMessage, AIMessage
+from langchain.schema.messages import HumanMessage, AIMessage
+from langchain_core.messages import AIMessageChunk
+from rest_framework import status
 
 from application.chat_pipeline.I_base_chat_pipeline import ParagraphPipelineModel
-from application.chat_pipeline.pipeline_manage import PiplineManage
+from application.chat_pipeline.pipeline_manage import PipelineManage
 from application.chat_pipeline.step.chat_step.i_chat_step import IChatStep, PostResponseHandler
+from application.flow.tools import Reasoning
 from application.models.api_key_model import ApplicationPublicAccessClient
 from common.constants.authentication_type import AuthenticationType
-from common.response import result
+from setting.models_provider.tools import get_model_instance_by_model_user_id
 
 
-def add_access_num(client_id=None, client_type=None):
-    if client_type == AuthenticationType.APPLICATION_ACCESS_TOKEN.value:
-        application_public_access_client = QuerySet(ApplicationPublicAccessClient).filter(id=client_id).first()
+def add_access_num(client_id=None, client_type=None, application_id=None):
+    if client_type == AuthenticationType.APPLICATION_ACCESS_TOKEN.value and application_id is not None:
+        application_public_access_client = (QuerySet(ApplicationPublicAccessClient).filter(client_id=client_id,
+                                                                                           application_id=application_id)
+                                            .first())
         if application_public_access_client is not None:
             application_public_access_client.access_num = application_public_access_client.access_num + 1
             application_public_access_client.intraday_access_num = application_public_access_client.intraday_access_num + 1
             application_public_access_client.save()
 
 
+def write_context(step, manage, request_token, response_token, all_text):
+    step.context['message_tokens'] = request_token
+    step.context['answer_tokens'] = response_token
+    current_time = time.time()
+    step.context['answer_text'] = all_text
+    step.context['run_time'] = current_time - step.context['start_time']
+    manage.context['run_time'] = current_time - manage.context['start_time']
+    manage.context['message_tokens'] = manage.context['message_tokens'] + request_token
+    manage.context['answer_tokens'] = manage.context['answer_tokens'] + response_token
+
+
 def event_content(response,
                   chat_id,
                   chat_record_id,
@@ -47,34 +63,95 @@ def event_content(response,
                   message_list: List[BaseMessage],
                   problem_text: str,
                   padding_problem_text: str = None,
-                  client_id=None, client_type=None):
+                  client_id=None, client_type=None,
+                  is_ai_chat: bool = None,
+                  model_setting=None):
+    if model_setting is None:
+        model_setting = {}
+    reasoning_content_enable = model_setting.get('reasoning_content_enable', False)
+    reasoning_content_start = model_setting.get('reasoning_content_start', '<think>')
+    reasoning_content_end = model_setting.get('reasoning_content_end', '</think>')
+    reasoning = Reasoning(reasoning_content_start,
+                          reasoning_content_end)
     all_text = ''
+    reasoning_content = ''
     try:
+        response_reasoning_content = False
         for chunk in response:
-            all_text += chunk.content
-            yield 'data: ' + json.dumps({'chat_id': str(chat_id), 'id': str(chat_record_id), 'operate': True,
-                                         'content': chunk.content, 'is_end': False}) + "\n\n"
-
+            reasoning_chunk = reasoning.get_reasoning_content(chunk)
+            content_chunk = reasoning_chunk.get('content')
+            if 'reasoning_content' in chunk.additional_kwargs:
+                response_reasoning_content = True
+                reasoning_content_chunk = chunk.additional_kwargs.get('reasoning_content', '')
+            else:
+                reasoning_content_chunk = reasoning_chunk.get('reasoning_content')
+            all_text += content_chunk
+            if reasoning_content_chunk is None:
+                reasoning_content_chunk = ''
+            reasoning_content += reasoning_content_chunk
+            yield manage.get_base_to_response().to_stream_chunk_response(chat_id, str(chat_record_id), 'ai-chat-node',
+                                                                         [], content_chunk,
+                                                                         False,
+                                                                         0, 0, {'node_is_end': False,
+                                                                                'view_type': 'many_view',
+                                                                                'node_type': 'ai-chat-node',
+                                                                                'real_node_id': 'ai-chat-node',
+                                                                                'reasoning_content': reasoning_content_chunk if reasoning_content_enable else ''})
+        reasoning_chunk = reasoning.get_end_reasoning_content()
+        all_text += reasoning_chunk.get('content')
+        reasoning_content_chunk = ""
+        if not response_reasoning_content:
+            reasoning_content_chunk = reasoning_chunk.get(
+                'reasoning_content')
+        yield manage.get_base_to_response().to_stream_chunk_response(chat_id, str(chat_record_id), 'ai-chat-node',
+                                                                     [], reasoning_chunk.get('content'),
+                                                                     False,
+                                                                     0, 0, {'node_is_end': False,
+                                                                            'view_type': 'many_view',
+                                                                            'node_type': 'ai-chat-node',
+                                                                            'real_node_id': 'ai-chat-node',
+                                                                            'reasoning_content'
+                                                                            : reasoning_content_chunk if reasoning_content_enable else ''})
         # 获取token
-        request_token = chat_model.get_num_tokens_from_messages(message_list)
-        response_token = chat_model.get_num_tokens(all_text)
-        step.context['message_tokens'] = request_token
-        step.context['answer_tokens'] = response_token
-        current_time = time.time()
-        step.context['answer_text'] = all_text
-        step.context['run_time'] = current_time - step.context['start_time']
-        manage.context['run_time'] = current_time - manage.context['start_time']
-        manage.context['message_tokens'] = manage.context['message_tokens'] + request_token
-        manage.context['answer_tokens'] = manage.context['answer_tokens'] + response_token
+        if is_ai_chat:
+            try:
+                request_token = chat_model.get_num_tokens_from_messages(message_list)
+                response_token = chat_model.get_num_tokens(all_text)
+            except Exception as e:
+                request_token = 0
+                response_token = 0
+        else:
+            request_token = 0
+            response_token = 0
+        write_context(step, manage, request_token, response_token, all_text)
+        asker = manage.context.get('form_data', {}).get('asker', None)
         post_response_handler.handler(chat_id, chat_record_id, paragraph_list, problem_text,
-                                      all_text, manage, step, padding_problem_text)
-        yield 'data: ' + json.dumps({'chat_id': str(chat_id), 'id': str(chat_record_id), 'operate': True,
-                                     'content': '', 'is_end': True}) + "\n\n"
-        add_access_num(client_id, client_type)
+                                      all_text, manage, step, padding_problem_text, client_id,
+                                      reasoning_content=reasoning_content if reasoning_content_enable else ''
+                                      , asker=asker)
+        yield manage.get_base_to_response().to_stream_chunk_response(chat_id, str(chat_record_id), 'ai-chat-node',
+                                                                     [], '', True,
+                                                                     request_token, response_token,
+                                                                     {'node_is_end': True, 'view_type': 'many_view',
+                                                                      'node_type': 'ai-chat-node'})
+        add_access_num(client_id, client_type, manage.context.get('application_id'))
     except Exception as e:
         logging.getLogger("max_kb_error").error(f'{str(e)}:{traceback.format_exc()}')
-        yield 'data: ' + json.dumps({'chat_id': str(chat_id), 'id': str(chat_record_id), 'operate': True,
-                                     'content': '异常' + str(e), 'is_end': True}) + "\n\n"
+        all_text = 'Exception:' + str(e)
+        write_context(step, manage, 0, 0, all_text)
+        asker = manage.context.get('form_data', {}).get('asker', None)
+        post_response_handler.handler(chat_id, chat_record_id, paragraph_list, problem_text,
+                                      all_text, manage, step, padding_problem_text, client_id, reasoning_content='',
+                                      asker=asker)
+        add_access_num(client_id, client_type, manage.context.get('application_id'))
+        yield manage.get_base_to_response().to_stream_chunk_response(chat_id, str(chat_record_id), 'ai-chat-node',
+                                                                     [], all_text,
+                                                                     False,
+                                                                     0, 0, {'node_is_end': False,
+                                                                            'view_type': 'many_view',
+                                                                            'node_type': 'ai-chat-node',
+                                                                            'real_node_id': 'ai-chat-node',
+                                                                            'reasoning_content': ''})
 
 
 class BaseChatStep(IChatStep):
@@ -82,21 +159,29 @@ class BaseChatStep(IChatStep):
                 chat_id,
                 problem_text,
                 post_response_handler: PostResponseHandler,
-                chat_model: BaseChatModel = None,
+                model_id: str = None,
+                user_id: str = None,
                 paragraph_list=None,
-                manage: PiplineManage = None,
+                manage: PipelineManage = None,
                 padding_problem_text: str = None,
                 stream: bool = True,
                 client_id=None, client_type=None,
+                no_references_setting=None,
+                model_params_setting=None,
+                model_setting=None,
                 **kwargs):
+        chat_model = get_model_instance_by_model_user_id(model_id, user_id,
+                                                         **model_params_setting) if model_id is not None else None
         if stream:
             return self.execute_stream(message_list, chat_id, problem_text, post_response_handler, chat_model,
                                        paragraph_list,
-                                       manage, padding_problem_text, client_id, client_type)
+                                       manage, padding_problem_text, client_id, client_type, no_references_setting,
+                                       model_setting)
         else:
             return self.execute_block(message_list, chat_id, problem_text, post_response_handler, chat_model,
                                       paragraph_list,
-                                      manage, padding_problem_text, client_id, client_type)
+                                      manage, padding_problem_text, client_id, client_type, no_references_setting,
+                                      model_setting)
 
     def get_details(self, manage, **kwargs):
         return {
@@ -119,60 +204,131 @@ class BaseChatStep(IChatStep):
         result.append({'role': 'ai', 'content': answer_text})
         return result
 
+    @staticmethod
+    def get_stream_result(message_list: List[BaseMessage],
+                          chat_model: BaseChatModel = None,
+                          paragraph_list=None,
+                          no_references_setting=None,
+                          problem_text=None):
+        if paragraph_list is None:
+            paragraph_list = []
+        directly_return_chunk_list = [AIMessageChunk(content=paragraph.content)
+                                      for paragraph in paragraph_list if (
+                                              paragraph.hit_handling_method == 'directly_return' and paragraph.similarity >= paragraph.directly_return_similarity)]
+        if directly_return_chunk_list is not None and len(directly_return_chunk_list) > 0:
+            return iter(directly_return_chunk_list), False
+        elif len(paragraph_list) == 0 and no_references_setting.get(
+                'status') == 'designated_answer':
+            return iter(
+                [AIMessageChunk(content=no_references_setting.get('value').replace('{question}', problem_text))]), False
+        if chat_model is None:
+            return iter([AIMessageChunk(
+                _('Sorry, the AI model is not configured. Please go to the application to set up the AI model first.'))]), False
+        else:
+            return chat_model.stream(message_list), True
+
     def execute_stream(self, message_list: List[BaseMessage],
                        chat_id,
                        problem_text,
                        post_response_handler: PostResponseHandler,
                        chat_model: BaseChatModel = None,
                        paragraph_list=None,
-                       manage: PiplineManage = None,
+                       manage: PipelineManage = None,
                        padding_problem_text: str = None,
-                       client_id=None, client_type=None):
-        # 调用模型
-        if chat_model is None:
-            chat_result = iter(
-                [BaseMessageChunk(content=paragraph.title + "\n" + paragraph.content) for paragraph in paragraph_list])
-        else:
-            chat_result = chat_model.stream(message_list)
-
+                       client_id=None, client_type=None,
+                       no_references_setting=None,
+                       model_setting=None):
+        chat_result, is_ai_chat = self.get_stream_result(message_list, chat_model, paragraph_list,
+                                                         no_references_setting, problem_text)
         chat_record_id = uuid.uuid1()
         r = StreamingHttpResponse(
             streaming_content=event_content(chat_result, chat_id, chat_record_id, paragraph_list,
                                             post_response_handler, manage, self, chat_model, message_list, problem_text,
-                                            padding_problem_text, client_id, client_type),
+                                            padding_problem_text, client_id, client_type, is_ai_chat, model_setting),
             content_type='text/event-stream;charset=utf-8')
 
         r['Cache-Control'] = 'no-cache'
         return r
 
+    @staticmethod
+    def get_block_result(message_list: List[BaseMessage],
+                         chat_model: BaseChatModel = None,
+                         paragraph_list=None,
+                         no_references_setting=None,
+                         problem_text=None):
+        if paragraph_list is None:
+            paragraph_list = []
+        directly_return_chunk_list = [AIMessageChunk(content=paragraph.content)
+                                      for paragraph in paragraph_list if (
+                                              paragraph.hit_handling_method == 'directly_return' and paragraph.similarity >= paragraph.directly_return_similarity)]
+        if directly_return_chunk_list is not None and len(directly_return_chunk_list) > 0:
+            return directly_return_chunk_list[0], False
+        elif len(paragraph_list) == 0 and no_references_setting.get(
+                'status') == 'designated_answer':
+            return AIMessage(no_references_setting.get('value').replace('{question}', problem_text)), False
+        if chat_model is None:
+            return AIMessage(
+                _('Sorry, the AI model is not configured. Please go to the application to set up the AI model first.')), False
+        else:
+            return chat_model.invoke(message_list), True
+
     def execute_block(self, message_list: List[BaseMessage],
                       chat_id,
                       problem_text,
                       post_response_handler: PostResponseHandler,
                       chat_model: BaseChatModel = None,
                       paragraph_list=None,
-                      manage: PiplineManage = None,
+                      manage: PipelineManage = None,
                       padding_problem_text: str = None,
-                      client_id=None, client_type=None):
-        # 调用模型
-        if chat_model is None:
-            chat_result = AIMessage(
-                content="\n\n".join([paragraph.title + "\n" + paragraph.content for paragraph in paragraph_list]))
-        else:
-            chat_result = chat_model.invoke(message_list)
+                      client_id=None, client_type=None, no_references_setting=None,
+                      model_setting=None):
+        reasoning_content_enable = model_setting.get('reasoning_content_enable', False)
+        reasoning_content_start = model_setting.get('reasoning_content_start', '<think>')
+        reasoning_content_end = model_setting.get('reasoning_content_end', '</think>')
+        reasoning = Reasoning(reasoning_content_start,
+                              reasoning_content_end)
         chat_record_id = uuid.uuid1()
-        request_token = chat_model.get_num_tokens_from_messages(message_list)
-        response_token = chat_model.get_num_tokens(chat_result.content)
-        self.context['message_tokens'] = request_token
-        self.context['answer_tokens'] = response_token
-        current_time = time.time()
-        self.context['answer_text'] = chat_result.content
-        self.context['run_time'] = current_time - self.context['start_time']
-        manage.context['run_time'] = current_time - manage.context['start_time']
-        manage.context['message_tokens'] = manage.context['message_tokens'] + request_token
-        manage.context['answer_tokens'] = manage.context['answer_tokens'] + response_token
-        post_response_handler.handler(chat_id, chat_record_id, paragraph_list, problem_text,
-                                      chat_result.content, manage, self, padding_problem_text)
-        add_access_num(client_id, client_type)
-        return result.success({'chat_id': str(chat_id), 'id': str(chat_record_id), 'operate': True,
-                               'content': chat_result.content, 'is_end': True})
+        # 调用模型
+        try:
+            chat_result, is_ai_chat = self.get_block_result(message_list, chat_model, paragraph_list,
+                                                            no_references_setting, problem_text)
+            if is_ai_chat:
+                request_token = chat_model.get_num_tokens_from_messages(message_list)
+                response_token = chat_model.get_num_tokens(chat_result.content)
+            else:
+                request_token = 0
+                response_token = 0
+            write_context(self, manage, request_token, response_token, chat_result.content)
+            reasoning_result = reasoning.get_reasoning_content(chat_result)
+            reasoning_result_end = reasoning.get_end_reasoning_content()
+            content = reasoning_result.get('content') + reasoning_result_end.get('content')
+            if 'reasoning_content' in chat_result.response_metadata:
+                reasoning_content = chat_result.response_metadata.get('reasoning_content', '')
+            else:
+                reasoning_content = reasoning_result.get('reasoning_content') + reasoning_result_end.get(
+                    'reasoning_content')
+            asker = manage.context.get('form_data', {}).get('asker', None)
+            post_response_handler.handler(chat_id, chat_record_id, paragraph_list, problem_text,
+                                          content, manage, self, padding_problem_text, client_id,
+                                          reasoning_content=reasoning_content if reasoning_content_enable else '',
+                                          asker=asker)
+            add_access_num(client_id, client_type, manage.context.get('application_id'))
+            return manage.get_base_to_response().to_block_response(str(chat_id), str(chat_record_id),
+                                                                   content, True,
+                                                                   request_token, response_token,
+                                                                   {
+                                                                       'reasoning_content': reasoning_content if reasoning_content_enable else '',
+                                                                       'answer_list': [{
+                                                                           'content': content,
+                                                                           'reasoning_content': reasoning_content if reasoning_content_enable else ''
+                                                                       }]})
+        except Exception as e:
+            all_text = 'Exception:' + str(e)
+            write_context(self, manage, 0, 0, all_text)
+            asker = manage.context.get('form_data', {}).get('asker', None)
+            post_response_handler.handler(chat_id, chat_record_id, paragraph_list, problem_text,
+                                          all_text, manage, self, padding_problem_text, client_id, reasoning_content='',
+                                          asker=asker)
+            add_access_num(client_id, client_type, manage.context.get('application_id'))
+            return manage.get_base_to_response().to_block_response(str(chat_id), str(chat_record_id), all_text, True, 0,
+                                                                   0, _status=status.HTTP_500_INTERNAL_SERVER_ERROR)

apps/application/chat_pipeline/step/generate_human_message_step/i_generate_human_message_step.py

@@ -9,41 +9,46 @@
 from abc import abstractmethod
 from typing import Type, List
 
+from django.utils.translation import gettext_lazy as _
 from langchain.schema import BaseMessage
 from rest_framework import serializers
 
 from application.chat_pipeline.I_base_chat_pipeline import IBaseChatPipelineStep, ParagraphPipelineModel
-from application.chat_pipeline.pipeline_manage import PiplineManage
+from application.chat_pipeline.pipeline_manage import PipelineManage
 from application.models import ChatRecord
+from application.serializers.application_serializers import NoReferencesSetting
 from common.field.common import InstanceField
 from common.util.field_message import ErrMessage
-from dataset.models import Paragraph
 
 
 class IGenerateHumanMessageStep(IBaseChatPipelineStep):
     class InstanceSerializer(serializers.Serializer):
         # 问题
-        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.char("问题"))
+        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.char(_("question")))
         # 段落列表
         paragraph_list = serializers.ListField(child=InstanceField(model_type=ParagraphPipelineModel, required=True),
-                                               error_messages=ErrMessage.list("段落列表"))
+                                               error_messages=ErrMessage.list(_("Paragraph List")))
         # 历史对答
         history_chat_record = serializers.ListField(child=InstanceField(model_type=ChatRecord, required=True),
-                                                    error_messages=ErrMessage.list("历史对答"))
+                                                    error_messages=ErrMessage.list(_("History Questions")))
         # 多轮对话数量
-        dialogue_number = serializers.IntegerField(required=True, error_messages=ErrMessage.integer("多轮对话数量"))
+        dialogue_number = serializers.IntegerField(required=True, error_messages=ErrMessage.integer(_("Number of multi-round conversations")))
         # 最大携带知识库段落长度
         max_paragraph_char_number = serializers.IntegerField(required=True, error_messages=ErrMessage.integer(
-            "最大携带知识库段落长度"))
+            _("Maximum length of the knowledge base paragraph")))
         # 模板
-        prompt = serializers.CharField(required=True, error_messages=ErrMessage.char("提示词"))
+        prompt = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Prompt word")))
+        system = serializers.CharField(required=False, allow_null=True, allow_blank=True,
+                                       error_messages=ErrMessage.char(_("System prompt words (role)")))
         # 补齐问题
-        padding_problem_text = serializers.CharField(required=False, error_messages=ErrMessage.char("补齐问题"))
+        padding_problem_text = serializers.CharField(required=False, error_messages=ErrMessage.char(_("Completion problem")))
+        # 未查询到引用分段
+        no_references_setting = NoReferencesSetting(required=True, error_messages=ErrMessage.base(_("No reference segment settings")))
 
-    def get_step_serializer(self, manage: PiplineManage) -> Type[serializers.Serializer]:
+    def get_step_serializer(self, manage: PipelineManage) -> Type[serializers.Serializer]:
         return self.InstanceSerializer
 
-    def _run(self, manage: PiplineManage):
+    def _run(self, manage: PipelineManage):
         message_list = self.execute(**self.context['step_args'])
         manage.context['message_list'] = message_list
 
@@ -56,6 +61,8 @@ class IGenerateHumanMessageStep(IBaseChatPipelineStep):
                 max_paragraph_char_number: int,
                 prompt: str,
                 padding_problem_text: str = None,
+                no_references_setting=None,
+                system=None,
                 **kwargs) -> List[BaseMessage]:
         """
 
@@ -67,6 +74,8 @@ class IGenerateHumanMessageStep(IBaseChatPipelineStep):
         :param prompt:                     模板
         :param padding_problem_text        用户修改文本
         :param kwargs:                     其他参数
+        :param no_references_setting:     无引用分段设置
+        :param system                     系统提示称
         :return:
         """
         pass

apps/application/chat_pipeline/step/generate_human_message_step/impl/base_generate_human_message_step.py

@@ -6,9 +6,10 @@
     @date：2024/1/10 17:50
     @desc:
 """
-from typing import List
+from typing import List, Dict
 
 from langchain.schema import BaseMessage, HumanMessage
+from langchain_core.messages import SystemMessage
 
 from application.chat_pipeline.I_base_chat_pipeline import ParagraphPipelineModel
 from application.chat_pipeline.step.generate_human_message_step.i_generate_human_message_step import \
@@ -26,22 +27,37 @@ class BaseGenerateHumanMessageStep(IGenerateHumanMessageStep):
                 max_paragraph_char_number: int,
                 prompt: str,
                 padding_problem_text: str = None,
+                no_references_setting=None,
+                system=None,
                 **kwargs) -> List[BaseMessage]:
+        prompt = prompt if (paragraph_list is not None and len(paragraph_list) > 0) else no_references_setting.get(
+            'value')
         exec_problem_text = padding_problem_text if padding_problem_text is not None else problem_text
         start_index = len(history_chat_record) - dialogue_number
         history_message = [[history_chat_record[index].get_human_message(), history_chat_record[index].get_ai_message()]
                            for index in
                            range(start_index if start_index > 0 else 0, len(history_chat_record))]
+        if system is not None and len(system) > 0:
+            return [SystemMessage(system), *flat_map(history_message),
+                    self.to_human_message(prompt, exec_problem_text, max_paragraph_char_number, paragraph_list,
+                                          no_references_setting)]
+
         return [*flat_map(history_message),
-                self.to_human_message(prompt, exec_problem_text, max_paragraph_char_number, paragraph_list)]
+                self.to_human_message(prompt, exec_problem_text, max_paragraph_char_number, paragraph_list,
+                                      no_references_setting)]
 
     @staticmethod
     def to_human_message(prompt: str,
                          problem: str,
                          max_paragraph_char_number: int,
-                         paragraph_list: List[ParagraphPipelineModel]):
+                         paragraph_list: List[ParagraphPipelineModel],
+                         no_references_setting: Dict):
         if paragraph_list is None or len(paragraph_list) == 0:
-            return HumanMessage(content=problem)
+            if no_references_setting.get('status') == 'ai_questioning':
+                return HumanMessage(
+                    content=no_references_setting.get('value').replace('{question}', problem))
+            else:
+                return HumanMessage(content=prompt.replace('{data}', "").replace('{question}', problem))
         temp_data = ""
         data_list = []
         for p in paragraph_list:
@@ -54,4 +70,4 @@ class BaseGenerateHumanMessageStep(IGenerateHumanMessageStep):
             else:
                 data_list.append(f"<data>{content}</data>")
         data = "\n".join(data_list)
-        return HumanMessage(content=prompt.format(**{'data': data, 'question': problem}))
+        return HumanMessage(content=prompt.replace('{data}', data).replace('{question}', problem))

apps/application/chat_pipeline/step/reset_problem_step/i_reset_problem_step.py

@@ -9,12 +9,11 @@
 from abc import abstractmethod
 from typing import Type, List
 
-from langchain.chat_models.base import BaseChatModel
+from django.utils.translation import gettext_lazy as _
 from rest_framework import serializers
 
 from application.chat_pipeline.I_base_chat_pipeline import IBaseChatPipelineStep
-from application.chat_pipeline.pipeline_manage import PiplineManage
-from application.chat_pipeline.step.chat_step.i_chat_step import ModelField
+from application.chat_pipeline.pipeline_manage import PipelineManage
 from application.models import ChatRecord
 from common.field.common import InstanceField
 from common.util.field_message import ErrMessage
@@ -23,17 +22,21 @@ from common.util.field_message import ErrMessage
 class IResetProblemStep(IBaseChatPipelineStep):
     class InstanceSerializer(serializers.Serializer):
         # 问题文本
-        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.float("问题文本"))
+        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.float(_("question")))
         # 历史对答
         history_chat_record = serializers.ListField(child=InstanceField(model_type=ChatRecord, required=True),
-                                                    error_messages=ErrMessage.list("历史对答"))
+                                                    error_messages=ErrMessage.list(_("History Questions")))
         # 大语言模型
-        chat_model = ModelField(error_messages=ErrMessage.base("大语言模型"))
+        model_id = serializers.UUIDField(required=False, allow_null=True, error_messages=ErrMessage.uuid(_("Model id")))
+        user_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid(_("User ID")))
+        problem_optimization_prompt = serializers.CharField(required=False, max_length=102400,
+                                                            error_messages=ErrMessage.char(
+                                                                _("Question completion prompt")))
 
-    def get_step_serializer(self, manage: PiplineManage) -> Type[serializers.Serializer]:
+    def get_step_serializer(self, manage: PipelineManage) -> Type[serializers.Serializer]:
         return self.InstanceSerializer
 
-    def _run(self, manage: PiplineManage):
+    def _run(self, manage: PipelineManage):
         padding_problem = self.execute(**self.context.get('step_args'))
         # 用户输入问题
         source_problem_text = self.context.get('step_args').get('problem_text')
@@ -42,10 +45,13 @@ class IResetProblemStep(IBaseChatPipelineStep):
         manage.context['problem_text'] = source_problem_text
         manage.context['padding_problem_text'] = padding_problem
         # 累加tokens
-        manage.context['message_tokens'] = manage.context['message_tokens'] + self.context.get('message_tokens')
-        manage.context['answer_tokens'] = manage.context['answer_tokens'] + self.context.get('answer_tokens')
+        manage.context['message_tokens'] = manage.context.get('message_tokens', 0) + self.context.get('message_tokens',
+                                                                                                      0)
+        manage.context['answer_tokens'] = manage.context.get('answer_tokens', 0) + self.context.get('answer_tokens', 0)
 
     @abstractmethod
-    def execute(self, problem_text: str, history_chat_record: List[ChatRecord] = None, chat_model: BaseChatModel = None,
+    def execute(self, problem_text: str, history_chat_record: List[ChatRecord] = None, model_id: str = None,
+                problem_optimization_prompt=None,
+                user_id=None,
                 **kwargs):
         pass

apps/application/chat_pipeline/step/reset_problem_step/impl/base_reset_problem_step.py

@@ -8,30 +8,51 @@
 """
 from typing import List
 
-from langchain.chat_models.base import BaseChatModel
+from django.utils.translation import gettext as _
 from langchain.schema import HumanMessage
 
 from application.chat_pipeline.step.reset_problem_step.i_reset_problem_step import IResetProblemStep
 from application.models import ChatRecord
 from common.util.split_model import flat_map
+from setting.models_provider.tools import get_model_instance_by_model_user_id
 
-prompt = (
-    '()里面是用户问题,根据上下文回答揣测用户问题({question}) 要求: 输出一个补全问题,并且放在<data></data>标签中')
+prompt = _(
+    "() contains the user's question. Answer the guessed user's question based on the context ({question}) Requirement: Output a complete question and put it in the <data></data> tag")
 
 
 class BaseResetProblemStep(IResetProblemStep):
-    def execute(self, problem_text: str, history_chat_record: List[ChatRecord] = None, chat_model: BaseChatModel = None,
+    def execute(self, problem_text: str, history_chat_record: List[ChatRecord] = None, model_id: str = None,
+                problem_optimization_prompt=None,
+                user_id=None,
                 **kwargs) -> str:
+        chat_model = get_model_instance_by_model_user_id(model_id, user_id) if model_id is not None else None
+        if chat_model is None:
+            return problem_text
         start_index = len(history_chat_record) - 3
         history_message = [[history_chat_record[index].get_human_message(), history_chat_record[index].get_ai_message()]
                            for index in
                            range(start_index if start_index > 0 else 0, len(history_chat_record))]
+        reset_prompt = problem_optimization_prompt if problem_optimization_prompt else prompt
         message_list = [*flat_map(history_message),
-                        HumanMessage(content=prompt.format(**{'question': problem_text}))]
-        response = chat_model(message_list)
-        padding_problem = response.content[response.content.index('<data>') + 6:response.content.index('</data>')]
-        self.context['message_tokens'] = chat_model.get_num_tokens_from_messages(message_list)
-        self.context['answer_tokens'] = chat_model.get_num_tokens(padding_problem)
+                        HumanMessage(content=reset_prompt.replace('{question}', problem_text))]
+        response = chat_model.invoke(message_list)
+        padding_problem = problem_text
+        if response.content.__contains__("<data>") and response.content.__contains__('</data>'):
+            padding_problem_data = response.content[
+                                   response.content.index('<data>') + 6:response.content.index('</data>')]
+            if padding_problem_data is not None and len(padding_problem_data.strip()) > 0:
+                padding_problem = padding_problem_data
+        elif len(response.content) > 0:
+            padding_problem = response.content
+
+        try:
+            request_token = chat_model.get_num_tokens_from_messages(message_list)
+            response_token = chat_model.get_num_tokens(padding_problem)
+        except Exception as e:
+            request_token = 0
+            response_token = 0
+        self.context['message_tokens'] = request_token
+        self.context['answer_tokens'] = response_token
         return padding_problem
 
     def get_details(self, manage, **kwargs):
@@ -39,8 +60,8 @@ class BaseResetProblemStep(IResetProblemStep):
             'step_type': 'problem_padding',
             'run_time': self.context['run_time'],
             'model_id': str(manage.context['model_id']) if 'model_id' in manage.context else None,
-            'message_tokens': self.context['message_tokens'],
-            'answer_tokens': self.context['answer_tokens'],
+            'message_tokens': self.context.get('message_tokens', 0),
+            'answer_tokens': self.context.get('answer_tokens', 0),
             'cost': 0,
             'padding_problem_text': self.context.get('padding_problem_text'),
             'problem_text': self.context.get("step_args").get('problem_text'),

apps/application/chat_pipeline/step/search_dataset_step/i_search_dataset_step.py

@@ -6,43 +6,51 @@
     @date：2024/1/9 18:10
     @desc: 检索知识库
 """
+import re
 from abc import abstractmethod
 from typing import List, Type
 
+from django.core import validators
+from django.utils.translation import gettext_lazy as _
 from rest_framework import serializers
 
 from application.chat_pipeline.I_base_chat_pipeline import IBaseChatPipelineStep, ParagraphPipelineModel
-from application.chat_pipeline.pipeline_manage import PiplineManage
+from application.chat_pipeline.pipeline_manage import PipelineManage
 from common.util.field_message import ErrMessage
-from dataset.models import Paragraph
 
 
 class ISearchDatasetStep(IBaseChatPipelineStep):
     class InstanceSerializer(serializers.Serializer):
         # 原始问题文本
-        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.char("问题"))
+        problem_text = serializers.CharField(required=True, error_messages=ErrMessage.char(_("question")))
         # 系统补全问题文本
-        padding_problem_text = serializers.CharField(required=False, error_messages=ErrMessage.char("系统补全问题文本"))
+        padding_problem_text = serializers.CharField(required=False,
+                                                     error_messages=ErrMessage.char(_("System completes question text")))
         # 需要查询的数据集id列表
         dataset_id_list = serializers.ListField(required=True, child=serializers.UUIDField(required=True),
-                                                error_messages=ErrMessage.list("数据集id列表"))
+                                                error_messages=ErrMessage.list(_("Dataset id list")))
         # 需要排除的文档id
         exclude_document_id_list = serializers.ListField(required=True, child=serializers.UUIDField(required=True),
-                                                         error_messages=ErrMessage.list("排除的文档id列表"))
+                                                         error_messages=ErrMessage.list(_("List of document ids to exclude")))
         # 需要排除向量id
         exclude_paragraph_id_list = serializers.ListField(required=True, child=serializers.UUIDField(required=True),
-                                                          error_messages=ErrMessage.list("排除向量id列表"))
+                                                          error_messages=ErrMessage.list(_("List of exclusion vector ids")))
         # 需要查询的条数
         top_n = serializers.IntegerField(required=True,
-                                         error_messages=ErrMessage.integer("引用分段数"))
+                                         error_messages=ErrMessage.integer(_("Reference segment number")))
         # 相似度 0-1之间
         similarity = serializers.FloatField(required=True, max_value=1, min_value=0,
-                                            error_messages=ErrMessage.float("引用分段数"))
+                                            error_messages=ErrMessage.float(_("Similarity")))
+        search_mode = serializers.CharField(required=True, validators=[
+            validators.RegexValidator(regex=re.compile("^embedding|keywords|blend$"),
+                                      message=_("The type only supports embedding|keywords|blend"), code=500)
+        ], error_messages=ErrMessage.char(_("Retrieval Mode")))
+        user_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid(_("User ID")))
 
-    def get_step_serializer(self, manage: PiplineManage) -> Type[InstanceSerializer]:
+    def get_step_serializer(self, manage: PipelineManage) -> Type[InstanceSerializer]:
         return self.InstanceSerializer
 
-    def _run(self, manage: PiplineManage):
+    def _run(self, manage: PipelineManage):
         paragraph_list = self.execute(**self.context['step_args'])
         manage.context['paragraph_list'] = paragraph_list
         self.context['paragraph_list'] = paragraph_list
@@ -50,6 +58,8 @@ class ISearchDatasetStep(IBaseChatPipelineStep):
     @abstractmethod
     def execute(self, problem_text: str, dataset_id_list: list[str], exclude_document_id_list: list[str],
                 exclude_paragraph_id_list: list[str], top_n: int, similarity: float, padding_problem_text: str = None,
+                search_mode: str = None,
+                user_id=None,
                 **kwargs) -> List[ParagraphPipelineModel]:
         """
         关于 用户和补全问题 说明: 补全问题如果有就使用补全问题去查询 反之就用用户原始问题查询
@@ -60,6 +70,8 @@ class ISearchDatasetStep(IBaseChatPipelineStep):
         :param exclude_document_id_list:           需要排除的文档id
         :param exclude_paragraph_id_list:          需要排除段落id
         :param padding_problem_text                补全问题
+        :param search_mode                         检索模式
+        :param user_id                             用户id
         :return: 段落列表
         """
         pass

apps/application/chat_pipeline/step/search_dataset_step/impl/base_search_dataset_step.py

@@ -10,31 +10,63 @@ import os
 from typing import List, Dict
 
 from django.db.models import QuerySet
+from django.utils.translation import gettext_lazy as _
+from rest_framework.utils.formatting import lazy_format
 
 from application.chat_pipeline.I_base_chat_pipeline import ParagraphPipelineModel
 from application.chat_pipeline.step.search_dataset_step.i_search_dataset_step import ISearchDatasetStep
-from common.config.embedding_config import VectorStore, EmbeddingModel
+from common.config.embedding_config import VectorStore, ModelManage
 from common.db.search import native_search
 from common.util.file_util import get_file_content
-from dataset.models import Paragraph
+from dataset.models import Paragraph, DataSet
+from embedding.models import SearchMode
+from setting.models import Model
+from setting.models_provider import get_model
 from smartdoc.conf import PROJECT_DIR
 
 
+def get_model_by_id(_id, user_id):
+    model = QuerySet(Model).filter(id=_id).first()
+    if model is None:
+        raise Exception(_("Model does not exist"))
+    if model.permission_type == 'PRIVATE' and str(model.user_id) != str(user_id):
+        message = lazy_format(_('No permission to use this model {model_name}'), model_name=model.name)
+        raise Exception(message)
+    return model
+
+
+def get_embedding_id(dataset_id_list):
+    dataset_list = QuerySet(DataSet).filter(id__in=dataset_id_list)
+    if len(set([dataset.embedding_mode_id for dataset in dataset_list])) > 1:
+        raise Exception(_("The vector model of the associated knowledge base is inconsistent and the segmentation cannot be recalled."))
+    if len(dataset_list) == 0:
+        raise Exception(_("The knowledge base setting is wrong, please reset the knowledge base"))
+    return dataset_list[0].embedding_mode_id
+
+
 class BaseSearchDatasetStep(ISearchDatasetStep):
 
     def execute(self, problem_text: str, dataset_id_list: list[str], exclude_document_id_list: list[str],
                 exclude_paragraph_id_list: list[str], top_n: int, similarity: float, padding_problem_text: str = None,
+                search_mode: str = None,
+                user_id=None,
                 **kwargs) -> List[ParagraphPipelineModel]:
+        if len(dataset_id_list) == 0:
+            return []
         exec_problem_text = padding_problem_text if padding_problem_text is not None else problem_text
-        embedding_model = EmbeddingModel.get_embedding_model()
+        model_id = get_embedding_id(dataset_id_list)
+        model = get_model_by_id(model_id, user_id)
+        self.context['model_name'] = model.name
+        embedding_model = ModelManage.get_model(model_id, lambda _id: get_model(model))
         embedding_value = embedding_model.embed_query(exec_problem_text)
         vector = VectorStore.get_embedding_vector()
-        embedding_list = vector.query(embedding_value, dataset_id_list, exclude_document_id_list,
-                                      exclude_paragraph_id_list, True, top_n, similarity)
+        embedding_list = vector.query(exec_problem_text, embedding_value, dataset_id_list, exclude_document_id_list,
+                                      exclude_paragraph_id_list, True, top_n, similarity, SearchMode(search_mode))
         if embedding_list is None:
             return []
-        paragraph_list = self.list_paragraph([row.get('paragraph_id') for row in embedding_list], vector)
-        return [self.reset_paragraph(paragraph, embedding_list) for paragraph in paragraph_list]
+        paragraph_list = self.list_paragraph(embedding_list, vector)
+        result = [self.reset_paragraph(paragraph, embedding_list) for paragraph in paragraph_list]
+        return result
 
     @staticmethod
     def reset_paragraph(paragraph: Dict, embedding_list: List) -> ParagraphPipelineModel:
@@ -48,10 +80,23 @@ class BaseSearchDatasetStep(ISearchDatasetStep):
                     .add_comprehensive_score(find_embedding.get('comprehensive_score'))
                     .add_dataset_name(paragraph.get('dataset_name'))
                     .add_document_name(paragraph.get('document_name'))
+                    .add_hit_handling_method(paragraph.get('hit_handling_method'))
+                    .add_directly_return_similarity(paragraph.get('directly_return_similarity'))
+                    .add_meta(paragraph.get('meta'))
                     .build())
 
     @staticmethod
-    def list_paragraph(paragraph_id_list: List, vector):
+    def get_similarity(paragraph, embedding_list: List):
+        filter_embedding_list = [embedding for embedding in embedding_list if
+                                 str(embedding.get('paragraph_id')) == str(paragraph.get('id'))]
+        if filter_embedding_list is not None and len(filter_embedding_list) > 0:
+            find_embedding = filter_embedding_list[-1]
+            return find_embedding.get('comprehensive_score')
+        return 0
+
+    @staticmethod
+    def list_paragraph(embedding_list: List, vector):
+        paragraph_id_list = [row.get('paragraph_id') for row in embedding_list]
         if paragraph_id_list is None or len(paragraph_id_list) == 0:
             return []
         paragraph_list = native_search(QuerySet(Paragraph).filter(id__in=paragraph_id_list),
@@ -65,6 +110,16 @@ class BaseSearchDatasetStep(ISearchDatasetStep):
             for paragraph_id in paragraph_id_list:
                 if not exist_paragraph_list.__contains__(paragraph_id):
                     vector.delete_by_paragraph_id(paragraph_id)
+        # 如果存在直接返回的则取直接返回段落
+        hit_handling_method_paragraph = [paragraph for paragraph in paragraph_list if
+                                         (paragraph.get(
+                                             'hit_handling_method') == 'directly_return' and BaseSearchDatasetStep.get_similarity(
+                                             paragraph, embedding_list) >= paragraph.get(
+                                             'directly_return_similarity'))]
+        if len(hit_handling_method_paragraph) > 0:
+            # 找到评分最高的
+            return [sorted(hit_handling_method_paragraph,
+                           key=lambda p: BaseSearchDatasetStep.get_similarity(p, embedding_list))[-1]]
         return paragraph_list
 
     def get_details(self, manage, **kwargs):
@@ -76,7 +131,7 @@ class BaseSearchDatasetStep(ISearchDatasetStep):
             'run_time': self.context['run_time'],
             'problem_text': step_args.get(
                 'padding_problem_text') if 'padding_problem_text' in step_args else step_args.get('problem_text'),
-            'model_name': EmbeddingModel.get_embedding_model().model_name,
+            'model_name': self.context.get('model_name'),
             'message_tokens': 0,
             'answer_tokens': 0,
             'cost': 0

apps/application/flow/__init__.py

@@ -0,0 +1,8 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/6/7 14:43
+    @desc:
+"""

apps/application/flow/common.py

@@ -0,0 +1,44 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： common.py
+    @date：2024/12/11 17:57
+    @desc:
+"""
+
+
+class Answer:
+    def __init__(self, content, view_type, runtime_node_id, chat_record_id, child_node, real_node_id,
+                 reasoning_content):
+        self.view_type = view_type
+        self.content = content
+        self.reasoning_content = reasoning_content
+        self.runtime_node_id = runtime_node_id
+        self.chat_record_id = chat_record_id
+        self.child_node = child_node
+        self.real_node_id = real_node_id
+
+    def to_dict(self):
+        return {'view_type': self.view_type, 'content': self.content, 'runtime_node_id': self.runtime_node_id,
+                'chat_record_id': self.chat_record_id,
+                'child_node': self.child_node,
+                'reasoning_content': self.reasoning_content,
+                'real_node_id': self.real_node_id}
+
+
+class NodeChunk:
+    def __init__(self):
+        self.status = 0
+        self.chunk_list = []
+
+    def add_chunk(self, chunk):
+        self.chunk_list.append(chunk)
+
+    def end(self, chunk=None):
+        if chunk is not None:
+            self.add_chunk(chunk)
+        self.status = 200
+
+    def is_end(self):
+        return self.status == 200

apps/application/flow/default_workflow.json

@@ -0,0 +1,451 @@
+{
+    "nodes": [
+        {
+            "id": "base-node",
+            "type": "base-node",
+            "x": 360,
+            "y": 2810,
+            "properties": {
+                "config": {
+
+                },
+                "height": 825.6,
+                "stepName": "基本信息",
+                "node_data": {
+                    "desc": "",
+                    "name": "maxkbapplication",
+                    "prologue": "您好，我是 MaxKB 小助手，您可以向我提出 MaxKB 使用问题。\n- MaxKB 主要功能有什么？\n- MaxKB 支持哪些大语言模型？\n- MaxKB 支持哪些文档类型？"
+                },
+                "input_field_list": [
+
+                ]
+            }
+        },
+        {
+            "id": "start-node",
+            "type": "start-node",
+            "x": 430,
+            "y": 3660,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "用户问题",
+                            "value": "question"
+                        }
+                    ],
+                    "globalFields": [
+                        {
+                            "label": "当前时间",
+                            "value": "time"
+                        }
+                    ]
+                },
+                "fields": [
+                    {
+                        "label": "用户问题",
+                        "value": "question"
+                    }
+                ],
+                "height": 276,
+                "stepName": "开始",
+                "globalFields": [
+                    {
+                        "label": "当前时间",
+                        "value": "time"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "type": "search-dataset-node",
+            "x": 840,
+            "y": 3210,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "检索结果的分段列表",
+                            "value": "paragraph_list"
+                        },
+                        {
+                            "label": "满足直接回答的分段列表",
+                            "value": "is_hit_handling_method_list"
+                        },
+                        {
+                            "label": "检索结果",
+                            "value": "data"
+                        },
+                        {
+                            "label": "满足直接回答的分段内容",
+                            "value": "directly_return"
+                        }
+                    ]
+                },
+                "height": 794,
+                "stepName": "知识库检索",
+                "node_data": {
+                    "dataset_id_list": [
+
+                    ],
+                    "dataset_setting": {
+                        "top_n": 3,
+                        "similarity": 0.6,
+                        "search_mode": "embedding",
+                        "max_paragraph_char_number": 5000
+                    },
+                    "question_reference_address": [
+                        "start-node",
+                        "question"
+                    ],
+                    "source_dataset_id_list": [
+
+                    ]
+                }
+            }
+        },
+        {
+            "id": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "type": "condition-node",
+            "x": 1490,
+            "y": 3210,
+            "properties": {
+                "width": 600,
+                "config": {
+                    "fields": [
+                        {
+                            "label": "分支名称",
+                            "value": "branch_name"
+                        }
+                    ]
+                },
+                "height": 543.675,
+                "stepName": "判断器",
+                "node_data": {
+                    "branch": [
+                        {
+                            "id": "1009",
+                            "type": "IF",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "is_hit_handling_method_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "4908",
+                            "type": "ELSE IF 1",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "paragraph_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "161",
+                            "type": "ELSE",
+                            "condition": "and",
+                            "conditions": [
+
+                            ]
+                        }
+                    ]
+                },
+                "branch_condition_list": [
+                    {
+                        "index": 0,
+                        "height": 121.225,
+                        "id": "1009"
+                    },
+                    {
+                        "index": 1,
+                        "height": 121.225,
+                        "id": "4908"
+                    },
+                    {
+                        "index": 2,
+                        "height": 44,
+                        "id": "161"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "type": "reply-node",
+            "x": 2170,
+            "y": 2480,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 378,
+                "stepName": "指定回复",
+                "node_data": {
+                    "fields": [
+                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                        "directly_return"
+                    ],
+                    "content": "",
+                    "reply_type": "referencing",
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3200,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 对话",
+                "node_data": {
+                    "prompt": "已知信息：\n{{知识库检索.data}}\n问题：\n{{开始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3970,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 对话1",
+                "node_data": {
+                    "prompt": "{{开始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        }
+    ],
+    "edges": [
+        {
+            "id": "7d0f166f-c472-41b2-b9a2-c294f4c83d73",
+            "type": "app-edge",
+            "sourceNodeId": "start-node",
+            "targetNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "startPoint": {
+                "x": 590,
+                "y": 3660
+            },
+            "endPoint": {
+                "x": 680,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 590,
+                    "y": 3660
+                },
+                {
+                    "x": 700,
+                    "y": 3660
+                },
+                {
+                    "x": 570,
+                    "y": 3210
+                },
+                {
+                    "x": 680,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "start-node_right",
+            "targetAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_left"
+        },
+        {
+            "id": "35cb86dd-f328-429e-a973-12fd7218b696",
+            "type": "app-edge",
+            "sourceNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "targetNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "startPoint": {
+                "x": 1000,
+                "y": 3210
+            },
+            "endPoint": {
+                "x": 1200,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1000,
+                    "y": 3210
+                },
+                {
+                    "x": 1110,
+                    "y": 3210
+                },
+                {
+                    "x": 1090,
+                    "y": 3210
+                },
+                {
+                    "x": 1200,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_right",
+            "targetAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_left"
+        },
+        {
+            "id": "e8f6cfe6-7e48-41cd-abd3-abfb5304d0d8",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "startPoint": {
+                "x": 1780,
+                "y": 3073.775
+            },
+            "endPoint": {
+                "x": 2010,
+                "y": 2480
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1890,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1900,
+                    "y": 2480
+                },
+                {
+                    "x": 2010,
+                    "y": 2480
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_1009_right",
+            "targetAnchorId": "4ffe1086-25df-4c85-b168-979b5bbf0a26_left"
+        },
+        {
+            "id": "994ff325-6f7a-4ebc-b61b-10e15519d6d2",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "startPoint": {
+                "x": 1780,
+                "y": 3203
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3200
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3200
+                },
+                {
+                    "x": 2000,
+                    "y": 3200
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_4908_right",
+            "targetAnchorId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb_left"
+        },
+        {
+            "id": "19270caf-bb9f-4ba7-9bf8-200aa70fecd5",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "startPoint": {
+                "x": 1780,
+                "y": 3293.6124999999997
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3970
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3970
+                },
+                {
+                    "x": 2000,
+                    "y": 3970
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_161_right",
+            "targetAnchorId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7_left"
+        }
+    ]
+}

apps/application/flow/default_workflow_en.json

@@ -0,0 +1,451 @@
+{
+    "nodes": [
+        {
+            "id": "base-node",
+            "type": "base-node",
+            "x": 360,
+            "y": 2810,
+            "properties": {
+                "config": {
+
+                },
+                "height": 825.6,
+                "stepName": "Base",
+                "node_data": {
+                    "desc": "",
+                    "name": "maxkbapplication",
+                    "prologue": "Hello, I am the MaxKB assistant. You can ask me about MaxKB usage issues.\n-What are the main functions of MaxKB?\n-What major language models does MaxKB support?\n-What document types does MaxKB support?"
+                },
+                "input_field_list": [
+
+                ]
+            }
+        },
+        {
+            "id": "start-node",
+            "type": "start-node",
+            "x": 430,
+            "y": 3660,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "用户问题",
+                            "value": "question"
+                        }
+                    ],
+                    "globalFields": [
+                        {
+                            "label": "当前时间",
+                            "value": "time"
+                        }
+                    ]
+                },
+                "fields": [
+                    {
+                        "label": "用户问题",
+                        "value": "question"
+                    }
+                ],
+                "height": 276,
+                "stepName": "Start",
+                "globalFields": [
+                    {
+                        "label": "当前时间",
+                        "value": "time"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "type": "search-dataset-node",
+            "x": 840,
+            "y": 3210,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "检索结果的分段列表",
+                            "value": "paragraph_list"
+                        },
+                        {
+                            "label": "满足直接回答的分段列表",
+                            "value": "is_hit_handling_method_list"
+                        },
+                        {
+                            "label": "检索结果",
+                            "value": "data"
+                        },
+                        {
+                            "label": "满足直接回答的分段内容",
+                            "value": "directly_return"
+                        }
+                    ]
+                },
+                "height": 794,
+                "stepName": "Knowledge Search",
+                "node_data": {
+                    "dataset_id_list": [
+
+                    ],
+                    "dataset_setting": {
+                        "top_n": 3,
+                        "similarity": 0.6,
+                        "search_mode": "embedding",
+                        "max_paragraph_char_number": 5000
+                    },
+                    "question_reference_address": [
+                        "start-node",
+                        "question"
+                    ],
+                    "source_dataset_id_list": [
+
+                    ]
+                }
+            }
+        },
+        {
+            "id": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "type": "condition-node",
+            "x": 1490,
+            "y": 3210,
+            "properties": {
+                "width": 600,
+                "config": {
+                    "fields": [
+                        {
+                            "label": "分支名称",
+                            "value": "branch_name"
+                        }
+                    ]
+                },
+                "height": 543.675,
+                "stepName": "Conditional Branch",
+                "node_data": {
+                    "branch": [
+                        {
+                            "id": "1009",
+                            "type": "IF",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "is_hit_handling_method_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "4908",
+                            "type": "ELSE IF 1",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "paragraph_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "161",
+                            "type": "ELSE",
+                            "condition": "and",
+                            "conditions": [
+
+                            ]
+                        }
+                    ]
+                },
+                "branch_condition_list": [
+                    {
+                        "index": 0,
+                        "height": 121.225,
+                        "id": "1009"
+                    },
+                    {
+                        "index": 1,
+                        "height": 121.225,
+                        "id": "4908"
+                    },
+                    {
+                        "index": 2,
+                        "height": 44,
+                        "id": "161"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "type": "reply-node",
+            "x": 2170,
+            "y": 2480,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 378,
+                "stepName": "Specified Reply",
+                "node_data": {
+                    "fields": [
+                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                        "directly_return"
+                    ],
+                    "content": "",
+                    "reply_type": "referencing",
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3200,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI Chat",
+                "node_data": {
+                    "prompt": "Known information:\n{{Knowledge Search.data}}\nQuestion:\n{{Start.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3970,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI Chat1",
+                "node_data": {
+                    "prompt": "{{Start.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        }
+    ],
+    "edges": [
+        {
+            "id": "7d0f166f-c472-41b2-b9a2-c294f4c83d73",
+            "type": "app-edge",
+            "sourceNodeId": "start-node",
+            "targetNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "startPoint": {
+                "x": 590,
+                "y": 3660
+            },
+            "endPoint": {
+                "x": 680,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 590,
+                    "y": 3660
+                },
+                {
+                    "x": 700,
+                    "y": 3660
+                },
+                {
+                    "x": 570,
+                    "y": 3210
+                },
+                {
+                    "x": 680,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "start-node_right",
+            "targetAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_left"
+        },
+        {
+            "id": "35cb86dd-f328-429e-a973-12fd7218b696",
+            "type": "app-edge",
+            "sourceNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "targetNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "startPoint": {
+                "x": 1000,
+                "y": 3210
+            },
+            "endPoint": {
+                "x": 1200,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1000,
+                    "y": 3210
+                },
+                {
+                    "x": 1110,
+                    "y": 3210
+                },
+                {
+                    "x": 1090,
+                    "y": 3210
+                },
+                {
+                    "x": 1200,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_right",
+            "targetAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_left"
+        },
+        {
+            "id": "e8f6cfe6-7e48-41cd-abd3-abfb5304d0d8",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "startPoint": {
+                "x": 1780,
+                "y": 3073.775
+            },
+            "endPoint": {
+                "x": 2010,
+                "y": 2480
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1890,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1900,
+                    "y": 2480
+                },
+                {
+                    "x": 2010,
+                    "y": 2480
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_1009_right",
+            "targetAnchorId": "4ffe1086-25df-4c85-b168-979b5bbf0a26_left"
+        },
+        {
+            "id": "994ff325-6f7a-4ebc-b61b-10e15519d6d2",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "startPoint": {
+                "x": 1780,
+                "y": 3203
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3200
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3200
+                },
+                {
+                    "x": 2000,
+                    "y": 3200
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_4908_right",
+            "targetAnchorId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb_left"
+        },
+        {
+            "id": "19270caf-bb9f-4ba7-9bf8-200aa70fecd5",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "startPoint": {
+                "x": 1780,
+                "y": 3293.6124999999997
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3970
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3970
+                },
+                {
+                    "x": 2000,
+                    "y": 3970
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_161_right",
+            "targetAnchorId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7_left"
+        }
+    ]
+}

apps/application/flow/default_workflow_zh.json

@@ -0,0 +1,451 @@
+{
+    "nodes": [
+        {
+            "id": "base-node",
+            "type": "base-node",
+            "x": 360,
+            "y": 2810,
+            "properties": {
+                "config": {
+
+                },
+                "height": 825.6,
+                "stepName": "基本信息",
+                "node_data": {
+                    "desc": "",
+                    "name": "maxkbapplication",
+                    "prologue": "您好，我是 MaxKB 小助手，您可以向我提出 MaxKB 使用问题。\n- MaxKB 主要功能有什么？\n- MaxKB 支持哪些大语言模型？\n- MaxKB 支持哪些文档类型？"
+                },
+                "input_field_list": [
+
+                ]
+            }
+        },
+        {
+            "id": "start-node",
+            "type": "start-node",
+            "x": 430,
+            "y": 3660,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "用户问题",
+                            "value": "question"
+                        }
+                    ],
+                    "globalFields": [
+                        {
+                            "label": "当前时间",
+                            "value": "time"
+                        }
+                    ]
+                },
+                "fields": [
+                    {
+                        "label": "用户问题",
+                        "value": "question"
+                    }
+                ],
+                "height": 276,
+                "stepName": "开始",
+                "globalFields": [
+                    {
+                        "label": "当前时间",
+                        "value": "time"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "type": "search-dataset-node",
+            "x": 840,
+            "y": 3210,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "检索结果的分段列表",
+                            "value": "paragraph_list"
+                        },
+                        {
+                            "label": "满足直接回答的分段列表",
+                            "value": "is_hit_handling_method_list"
+                        },
+                        {
+                            "label": "检索结果",
+                            "value": "data"
+                        },
+                        {
+                            "label": "满足直接回答的分段内容",
+                            "value": "directly_return"
+                        }
+                    ]
+                },
+                "height": 794,
+                "stepName": "知识库检索",
+                "node_data": {
+                    "dataset_id_list": [
+
+                    ],
+                    "dataset_setting": {
+                        "top_n": 3,
+                        "similarity": 0.6,
+                        "search_mode": "embedding",
+                        "max_paragraph_char_number": 5000
+                    },
+                    "question_reference_address": [
+                        "start-node",
+                        "question"
+                    ],
+                    "source_dataset_id_list": [
+
+                    ]
+                }
+            }
+        },
+        {
+            "id": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "type": "condition-node",
+            "x": 1490,
+            "y": 3210,
+            "properties": {
+                "width": 600,
+                "config": {
+                    "fields": [
+                        {
+                            "label": "分支名称",
+                            "value": "branch_name"
+                        }
+                    ]
+                },
+                "height": 543.675,
+                "stepName": "判断器",
+                "node_data": {
+                    "branch": [
+                        {
+                            "id": "1009",
+                            "type": "IF",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "is_hit_handling_method_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "4908",
+                            "type": "ELSE IF 1",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "paragraph_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "161",
+                            "type": "ELSE",
+                            "condition": "and",
+                            "conditions": [
+
+                            ]
+                        }
+                    ]
+                },
+                "branch_condition_list": [
+                    {
+                        "index": 0,
+                        "height": 121.225,
+                        "id": "1009"
+                    },
+                    {
+                        "index": 1,
+                        "height": 121.225,
+                        "id": "4908"
+                    },
+                    {
+                        "index": 2,
+                        "height": 44,
+                        "id": "161"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "type": "reply-node",
+            "x": 2170,
+            "y": 2480,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 378,
+                "stepName": "指定回复",
+                "node_data": {
+                    "fields": [
+                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                        "directly_return"
+                    ],
+                    "content": "",
+                    "reply_type": "referencing",
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3200,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 对话",
+                "node_data": {
+                    "prompt": "已知信息：\n{{知识库检索.data}}\n问题：\n{{开始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3970,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 对话1",
+                "node_data": {
+                    "prompt": "{{开始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        }
+    ],
+    "edges": [
+        {
+            "id": "7d0f166f-c472-41b2-b9a2-c294f4c83d73",
+            "type": "app-edge",
+            "sourceNodeId": "start-node",
+            "targetNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "startPoint": {
+                "x": 590,
+                "y": 3660
+            },
+            "endPoint": {
+                "x": 680,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 590,
+                    "y": 3660
+                },
+                {
+                    "x": 700,
+                    "y": 3660
+                },
+                {
+                    "x": 570,
+                    "y": 3210
+                },
+                {
+                    "x": 680,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "start-node_right",
+            "targetAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_left"
+        },
+        {
+            "id": "35cb86dd-f328-429e-a973-12fd7218b696",
+            "type": "app-edge",
+            "sourceNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "targetNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "startPoint": {
+                "x": 1000,
+                "y": 3210
+            },
+            "endPoint": {
+                "x": 1200,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1000,
+                    "y": 3210
+                },
+                {
+                    "x": 1110,
+                    "y": 3210
+                },
+                {
+                    "x": 1090,
+                    "y": 3210
+                },
+                {
+                    "x": 1200,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_right",
+            "targetAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_left"
+        },
+        {
+            "id": "e8f6cfe6-7e48-41cd-abd3-abfb5304d0d8",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "startPoint": {
+                "x": 1780,
+                "y": 3073.775
+            },
+            "endPoint": {
+                "x": 2010,
+                "y": 2480
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1890,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1900,
+                    "y": 2480
+                },
+                {
+                    "x": 2010,
+                    "y": 2480
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_1009_right",
+            "targetAnchorId": "4ffe1086-25df-4c85-b168-979b5bbf0a26_left"
+        },
+        {
+            "id": "994ff325-6f7a-4ebc-b61b-10e15519d6d2",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "startPoint": {
+                "x": 1780,
+                "y": 3203
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3200
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3200
+                },
+                {
+                    "x": 2000,
+                    "y": 3200
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_4908_right",
+            "targetAnchorId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb_left"
+        },
+        {
+            "id": "19270caf-bb9f-4ba7-9bf8-200aa70fecd5",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "startPoint": {
+                "x": 1780,
+                "y": 3293.6124999999997
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3970
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3970
+                },
+                {
+                    "x": 2000,
+                    "y": 3970
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_161_right",
+            "targetAnchorId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7_left"
+        }
+    ]
+}

apps/application/flow/default_workflow_zh_Hant.json

@@ -0,0 +1,451 @@
+{
+    "nodes": [
+        {
+            "id": "base-node",
+            "type": "base-node",
+            "x": 360,
+            "y": 2810,
+            "properties": {
+                "config": {
+
+                },
+                "height": 825.6,
+                "stepName": "基本資訊",
+                "node_data": {
+                    "desc": "",
+                    "name": "maxkbapplication",
+                    "prologue": "您好，我是MaxKB小助手，您可以向我提出MaxKB使用問題。\n- MaxKB主要功能有什麼？\n- MaxKB支持哪些大語言模型？\n- MaxKB支持哪些文檔類型？"
+                },
+                "input_field_list": [
+
+                ]
+            }
+        },
+        {
+            "id": "start-node",
+            "type": "start-node",
+            "x": 430,
+            "y": 3660,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "用户问题",
+                            "value": "question"
+                        }
+                    ],
+                    "globalFields": [
+                        {
+                            "label": "当前时间",
+                            "value": "time"
+                        }
+                    ]
+                },
+                "fields": [
+                    {
+                        "label": "用户问题",
+                        "value": "question"
+                    }
+                ],
+                "height": 276,
+                "stepName": "開始",
+                "globalFields": [
+                    {
+                        "label": "当前时间",
+                        "value": "time"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "type": "search-dataset-node",
+            "x": 840,
+            "y": 3210,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "检索结果的分段列表",
+                            "value": "paragraph_list"
+                        },
+                        {
+                            "label": "满足直接回答的分段列表",
+                            "value": "is_hit_handling_method_list"
+                        },
+                        {
+                            "label": "检索结果",
+                            "value": "data"
+                        },
+                        {
+                            "label": "满足直接回答的分段内容",
+                            "value": "directly_return"
+                        }
+                    ]
+                },
+                "height": 794,
+                "stepName": "知識庫檢索",
+                "node_data": {
+                    "dataset_id_list": [
+
+                    ],
+                    "dataset_setting": {
+                        "top_n": 3,
+                        "similarity": 0.6,
+                        "search_mode": "embedding",
+                        "max_paragraph_char_number": 5000
+                    },
+                    "question_reference_address": [
+                        "start-node",
+                        "question"
+                    ],
+                    "source_dataset_id_list": [
+
+                    ]
+                }
+            }
+        },
+        {
+            "id": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "type": "condition-node",
+            "x": 1490,
+            "y": 3210,
+            "properties": {
+                "width": 600,
+                "config": {
+                    "fields": [
+                        {
+                            "label": "分支名称",
+                            "value": "branch_name"
+                        }
+                    ]
+                },
+                "height": 543.675,
+                "stepName": "判斷器",
+                "node_data": {
+                    "branch": [
+                        {
+                            "id": "1009",
+                            "type": "IF",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "is_hit_handling_method_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "4908",
+                            "type": "ELSE IF 1",
+                            "condition": "and",
+                            "conditions": [
+                                {
+                                    "field": [
+                                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                                        "paragraph_list"
+                                    ],
+                                    "value": "1",
+                                    "compare": "len_ge"
+                                }
+                            ]
+                        },
+                        {
+                            "id": "161",
+                            "type": "ELSE",
+                            "condition": "and",
+                            "conditions": [
+
+                            ]
+                        }
+                    ]
+                },
+                "branch_condition_list": [
+                    {
+                        "index": 0,
+                        "height": 121.225,
+                        "id": "1009"
+                    },
+                    {
+                        "index": 1,
+                        "height": 121.225,
+                        "id": "4908"
+                    },
+                    {
+                        "index": 2,
+                        "height": 44,
+                        "id": "161"
+                    }
+                ]
+            }
+        },
+        {
+            "id": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "type": "reply-node",
+            "x": 2170,
+            "y": 2480,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 378,
+                "stepName": "指定回覆",
+                "node_data": {
+                    "fields": [
+                        "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+                        "directly_return"
+                    ],
+                    "content": "",
+                    "reply_type": "referencing",
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3200,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 對話",
+                "node_data": {
+                    "prompt": "已知資訊：\n{{知識庫檢索.data}}\n問題：\n{{開始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        },
+        {
+            "id": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "type": "ai-chat-node",
+            "x": 2160,
+            "y": 3970,
+            "properties": {
+                "config": {
+                    "fields": [
+                        {
+                            "label": "AI 回答内容",
+                            "value": "answer"
+                        }
+                    ]
+                },
+                "height": 763,
+                "stepName": "AI 對話1",
+                "node_data": {
+                    "prompt": "{{開始.question}}",
+                    "system": "",
+                    "model_id": "",
+                    "dialogue_number": 0,
+                    "is_result": true
+                }
+            }
+        }
+    ],
+    "edges": [
+        {
+            "id": "7d0f166f-c472-41b2-b9a2-c294f4c83d73",
+            "type": "app-edge",
+            "sourceNodeId": "start-node",
+            "targetNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "startPoint": {
+                "x": 590,
+                "y": 3660
+            },
+            "endPoint": {
+                "x": 680,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 590,
+                    "y": 3660
+                },
+                {
+                    "x": 700,
+                    "y": 3660
+                },
+                {
+                    "x": 570,
+                    "y": 3210
+                },
+                {
+                    "x": 680,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "start-node_right",
+            "targetAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_left"
+        },
+        {
+            "id": "35cb86dd-f328-429e-a973-12fd7218b696",
+            "type": "app-edge",
+            "sourceNodeId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5",
+            "targetNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "startPoint": {
+                "x": 1000,
+                "y": 3210
+            },
+            "endPoint": {
+                "x": 1200,
+                "y": 3210
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1000,
+                    "y": 3210
+                },
+                {
+                    "x": 1110,
+                    "y": 3210
+                },
+                {
+                    "x": 1090,
+                    "y": 3210
+                },
+                {
+                    "x": 1200,
+                    "y": 3210
+                }
+            ],
+            "sourceAnchorId": "b931efe5-5b66-46e0-ae3b-0160cb18eeb5_right",
+            "targetAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_left"
+        },
+        {
+            "id": "e8f6cfe6-7e48-41cd-abd3-abfb5304d0d8",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "4ffe1086-25df-4c85-b168-979b5bbf0a26",
+            "startPoint": {
+                "x": 1780,
+                "y": 3073.775
+            },
+            "endPoint": {
+                "x": 2010,
+                "y": 2480
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1890,
+                    "y": 3073.775
+                },
+                {
+                    "x": 1900,
+                    "y": 2480
+                },
+                {
+                    "x": 2010,
+                    "y": 2480
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_1009_right",
+            "targetAnchorId": "4ffe1086-25df-4c85-b168-979b5bbf0a26_left"
+        },
+        {
+            "id": "994ff325-6f7a-4ebc-b61b-10e15519d6d2",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb",
+            "startPoint": {
+                "x": 1780,
+                "y": 3203
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3200
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3203
+                },
+                {
+                    "x": 1890,
+                    "y": 3200
+                },
+                {
+                    "x": 2000,
+                    "y": 3200
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_4908_right",
+            "targetAnchorId": "f1f1ee18-5a02-46f6-b4e6-226253cdffbb_left"
+        },
+        {
+            "id": "19270caf-bb9f-4ba7-9bf8-200aa70fecd5",
+            "type": "app-edge",
+            "sourceNodeId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b",
+            "targetNodeId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7",
+            "startPoint": {
+                "x": 1780,
+                "y": 3293.6124999999997
+            },
+            "endPoint": {
+                "x": 2000,
+                "y": 3970
+            },
+            "properties": {
+
+            },
+            "pointsList": [
+                {
+                    "x": 1780,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3293.6124999999997
+                },
+                {
+                    "x": 1890,
+                    "y": 3970
+                },
+                {
+                    "x": 2000,
+                    "y": 3970
+                }
+            ],
+            "sourceAnchorId": "fc60863a-dec2-4854-9e5a-7a44b7187a2b_161_right",
+            "targetAnchorId": "309d0eef-c597-46b5-8d51-b9a28aaef4c7_left"
+        }
+    ]
+}

apps/application/flow/i_step_node.py

@@ -0,0 +1,256 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： i_step_node.py
+    @date：2024/6/3 14:57
+    @desc:
+"""
+import time
+import uuid
+from abc import abstractmethod
+from hashlib import sha1
+from typing import Type, Dict, List
+
+from django.core import cache
+from django.db.models import QuerySet
+from rest_framework import serializers
+from rest_framework.exceptions import ValidationError, ErrorDetail
+
+from application.flow.common import Answer, NodeChunk
+from application.models import ChatRecord
+from application.models.api_key_model import ApplicationPublicAccessClient
+from common.constants.authentication_type import AuthenticationType
+from common.field.common import InstanceField
+from common.util.field_message import ErrMessage
+
+chat_cache = cache.caches['chat_cache']
+
+
+def write_context(step_variable: Dict, global_variable: Dict, node, workflow):
+    if step_variable is not None:
+        for key in step_variable:
+            node.context[key] = step_variable[key]
+        if workflow.is_result(node, NodeResult(step_variable, global_variable)) and 'answer' in step_variable:
+            answer = step_variable['answer']
+            yield answer
+            node.answer_text = answer
+    if global_variable is not None:
+        for key in global_variable:
+            workflow.context[key] = global_variable[key]
+    node.context['run_time'] = time.time() - node.context['start_time']
+
+
+def is_interrupt(node, step_variable: Dict, global_variable: Dict):
+    return node.type == 'form-node' and not node.context.get('is_submit', False)
+
+
+class WorkFlowPostHandler:
+    def __init__(self, chat_info, client_id, client_type):
+        self.chat_info = chat_info
+        self.client_id = client_id
+        self.client_type = client_type
+
+    def handler(self, chat_id,
+                chat_record_id,
+                answer,
+                workflow):
+        question = workflow.params['question']
+        details = workflow.get_runtime_details()
+        message_tokens = sum([row.get('message_tokens') for row in details.values() if
+                              'message_tokens' in row and row.get('message_tokens') is not None])
+        answer_tokens = sum([row.get('answer_tokens') for row in details.values() if
+                             'answer_tokens' in row and row.get('answer_tokens') is not None])
+        answer_text_list = workflow.get_answer_text_list()
+        answer_text = '\n\n'.join(
+            '\n\n'.join([a.get('content') for a in answer]) for answer in
+            answer_text_list)
+        if workflow.chat_record is not None:
+            chat_record = workflow.chat_record
+            chat_record.answer_text = answer_text
+            chat_record.details = details
+            chat_record.message_tokens = message_tokens
+            chat_record.answer_tokens = answer_tokens
+            chat_record.answer_text_list = answer_text_list
+            chat_record.run_time = time.time() - workflow.context['start_time']
+        else:
+            chat_record = ChatRecord(id=chat_record_id,
+                                     chat_id=chat_id,
+                                     problem_text=question,
+                                     answer_text=answer_text,
+                                     details=details,
+                                     message_tokens=message_tokens,
+                                     answer_tokens=answer_tokens,
+                                     answer_text_list=answer_text_list,
+                                     run_time=time.time() - workflow.context['start_time'],
+                                     index=0)
+        asker = workflow.context.get('asker', None)
+        self.chat_info.append_chat_record(chat_record, self.client_id, asker)
+        # 重新设置缓存
+        chat_cache.set(chat_id,
+                       self.chat_info, timeout=60 * 30)
+        if self.client_type == AuthenticationType.APPLICATION_ACCESS_TOKEN.value:
+            application_public_access_client = (QuerySet(ApplicationPublicAccessClient)
+                                                .filter(client_id=self.client_id,
+                                                        application_id=self.chat_info.application.id).first())
+            if application_public_access_client is not None:
+                application_public_access_client.access_num = application_public_access_client.access_num + 1
+                application_public_access_client.intraday_access_num = application_public_access_client.intraday_access_num + 1
+                application_public_access_client.save()
+
+
+class NodeResult:
+    def __init__(self, node_variable: Dict, workflow_variable: Dict,
+                 _write_context=write_context, _is_interrupt=is_interrupt):
+        self._write_context = _write_context
+        self.node_variable = node_variable
+        self.workflow_variable = workflow_variable
+        self._is_interrupt = _is_interrupt
+
+    def write_context(self, node, workflow):
+        return self._write_context(self.node_variable, self.workflow_variable, node, workflow)
+
+    def is_assertion_result(self):
+        return 'branch_id' in self.node_variable
+
+    def is_interrupt_exec(self, current_node):
+        """
+        是否中断执行
+        @param current_node:
+        @return:
+        """
+        return self._is_interrupt(current_node, self.node_variable, self.workflow_variable)
+
+
+class ReferenceAddressSerializer(serializers.Serializer):
+    node_id = serializers.CharField(required=True, error_messages=ErrMessage.char("节点id"))
+    fields = serializers.ListField(
+        child=serializers.CharField(required=True, error_messages=ErrMessage.char("节点字段")), required=True,
+        error_messages=ErrMessage.list("节点字段数组"))
+
+
+class FlowParamsSerializer(serializers.Serializer):
+    # 历史对答
+    history_chat_record = serializers.ListField(child=InstanceField(model_type=ChatRecord, required=True),
+                                                error_messages=ErrMessage.list("历史对答"))
+
+    question = serializers.CharField(required=True, error_messages=ErrMessage.list("用户问题"))
+
+    chat_id = serializers.CharField(required=True, error_messages=ErrMessage.list("对话id"))
+
+    chat_record_id = serializers.CharField(required=True, error_messages=ErrMessage.char("对话记录id"))
+
+    stream = serializers.BooleanField(required=True, error_messages=ErrMessage.boolean("流式输出"))
+
+    client_id = serializers.CharField(required=False, error_messages=ErrMessage.char("客户端id"))
+
+    client_type = serializers.CharField(required=False, error_messages=ErrMessage.char("客户端类型"))
+
+    user_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid("用户id"))
+    re_chat = serializers.BooleanField(required=True, error_messages=ErrMessage.boolean("换个答案"))
+
+
+class INode:
+    view_type = 'many_view'
+
+    @abstractmethod
+    def save_context(self, details, workflow_manage):
+        pass
+
+    def get_answer_list(self) -> List[Answer] | None:
+        if self.answer_text is None:
+            return None
+        reasoning_content_enable = self.context.get('model_setting', {}).get('reasoning_content_enable', False)
+        return [
+            Answer(self.answer_text, self.view_type, self.runtime_node_id, self.workflow_params['chat_record_id'], {},
+                   self.runtime_node_id, self.context.get('reasoning_content', '') if reasoning_content_enable else '')]
+
+    def __init__(self, node, workflow_params, workflow_manage, up_node_id_list=None,
+                 get_node_params=lambda node: node.properties.get('node_data')):
+        # 当前步骤上下文,用于存储当前步骤信息
+        self.status = 200
+        self.err_message = ''
+        self.node = node
+        self.node_params = get_node_params(node)
+        self.workflow_params = workflow_params
+        self.workflow_manage = workflow_manage
+        self.node_params_serializer = None
+        self.flow_params_serializer = None
+        self.context = {}
+        self.answer_text = None
+        self.id = node.id
+        if up_node_id_list is None:
+            up_node_id_list = []
+        self.up_node_id_list = up_node_id_list
+        self.node_chunk = NodeChunk()
+        self.runtime_node_id = sha1(uuid.NAMESPACE_DNS.bytes + bytes(str(uuid.uuid5(uuid.NAMESPACE_DNS,
+                                                                                    "".join([*sorted(up_node_id_list),
+                                                                                             node.id]))),
+                                                                     "utf-8")).hexdigest()
+
+    def valid_args(self, node_params, flow_params):
+        flow_params_serializer_class = self.get_flow_params_serializer_class()
+        node_params_serializer_class = self.get_node_params_serializer_class()
+        if flow_params_serializer_class is not None and flow_params is not None:
+            self.flow_params_serializer = flow_params_serializer_class(data=flow_params)
+            self.flow_params_serializer.is_valid(raise_exception=True)
+        if node_params_serializer_class is not None:
+            self.node_params_serializer = node_params_serializer_class(data=node_params)
+            self.node_params_serializer.is_valid(raise_exception=True)
+        if self.node.properties.get('status', 200) != 200:
+            raise ValidationError(ErrorDetail(f'节点{self.node.properties.get("stepName")} 不可用'))
+
+    def get_reference_field(self, fields: List[str]):
+        return self.get_field(self.context, fields)
+
+    @staticmethod
+    def get_field(obj, fields: List[str]):
+        for field in fields:
+            value = obj.get(field)
+            if value is None:
+                return None
+            else:
+                obj = value
+        return obj
+
+    @abstractmethod
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        pass
+
+    def get_flow_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return FlowParamsSerializer
+
+    def get_write_error_context(self, e):
+        self.status = 500
+        self.answer_text = str(e)
+        self.err_message = str(e)
+        self.context['run_time'] = time.time() - self.context['start_time']
+
+        def write_error_context(answer, status=200):
+            pass
+
+        return write_error_context
+
+    def run(self) -> NodeResult:
+        """
+        :return: 执行结果
+        """
+        start_time = time.time()
+        self.context['start_time'] = start_time
+        result = self._run()
+        self.context['run_time'] = time.time() - start_time
+        return result
+
+    def _run(self):
+        result = self.execute()
+        return result
+
+    def execute(self, **kwargs) -> NodeResult:
+        pass
+
+    def get_details(self, index: int, **kwargs):
+        """
+        运行详情
+        :return: 步骤详情
+        """
+        return {}

apps/application/flow/step_node/__init__.py

@@ -0,0 +1,42 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/6/7 14:43
+    @desc:
+"""
+from .ai_chat_step_node import *
+from .application_node import BaseApplicationNode
+from .condition_node import *
+from .direct_reply_node import *
+from .form_node import *
+from .function_lib_node import *
+from .function_node import *
+from .question_node import *
+from .reranker_node import *
+
+from .document_extract_node import *
+from .image_understand_step_node import *
+from .image_generate_step_node import *
+
+from .search_dataset_node import *
+from .speech_to_text_step_node import BaseSpeechToTextNode
+from .start_node import *
+from .text_to_speech_step_node.impl.base_text_to_speech_node import BaseTextToSpeechNode
+from .variable_assign_node import BaseVariableAssignNode
+from .mcp_node import BaseMcpNode
+
+node_list = [BaseStartStepNode, BaseChatNode, BaseSearchDatasetNode, BaseQuestionNode,
+             BaseConditionNode, BaseReplyNode,
+             BaseFunctionNodeNode, BaseFunctionLibNodeNode, BaseRerankerNode, BaseApplicationNode,
+             BaseDocumentExtractNode,
+             BaseImageUnderstandNode, BaseFormNode, BaseSpeechToTextNode, BaseTextToSpeechNode,
+             BaseImageGenerateNode, BaseVariableAssignNode, BaseMcpNode]
+
+
+def get_node(node_type):
+    find_list = [node for node in node_list if node.type == node_type]
+    if len(find_list) > 0:
+        return find_list[0]
+    return None

apps/application/flow/step_node/ai_chat_step_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py
+    @date：2024/6/11 15:29
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/ai_chat_step_node/i_chat_node.py

@@ -0,0 +1,58 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： i_chat_node.py
+    @date：2024/6/4 13:58
+    @desc:
+"""
+from typing import Type
+
+from django.utils.translation import gettext_lazy as _
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+
+
+class ChatNodeSerializer(serializers.Serializer):
+    model_id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Model id")))
+    system = serializers.CharField(required=False, allow_blank=True, allow_null=True,
+                                   error_messages=ErrMessage.char(_("Role Setting")))
+    prompt = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Prompt word")))
+    # 多轮对话数量
+    dialogue_number = serializers.IntegerField(required=True, error_messages=ErrMessage.integer(
+        _("Number of multi-round conversations")))
+
+    is_result = serializers.BooleanField(required=False,
+                                         error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    model_params_setting = serializers.DictField(required=False,
+                                                 error_messages=ErrMessage.dict(_("Model parameter settings")))
+    model_setting = serializers.DictField(required=False,
+                                          error_messages=ErrMessage.dict('Model settings'))
+    dialogue_type = serializers.CharField(required=False, allow_blank=True, allow_null=True,
+                                          error_messages=ErrMessage.char(_("Context Type")))
+    mcp_enable = serializers.BooleanField(required=False,
+                                         error_messages=ErrMessage.boolean(_("Whether to enable MCP")))
+    mcp_servers = serializers.JSONField(required=False, error_messages=ErrMessage.list(_("MCP Server")))
+
+
+class IChatNode(INode):
+    type = 'ai-chat-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ChatNodeSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, model_id, system, prompt, dialogue_number, history_chat_record, stream, chat_id,
+                chat_record_id,
+                model_params_setting=None,
+                dialogue_type=None,
+                model_setting=None,
+                mcp_enable=False,
+                mcp_servers=None,
+                **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/ai_chat_step_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py
+    @date：2024/6/11 15:34
+    @desc:
+"""
+from .base_chat_node import BaseChatNode

apps/application/flow/step_node/ai_chat_step_node/impl/base_chat_node.py

@@ -0,0 +1,288 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： base_question_node.py
+    @date：2024/6/4 14:30
+    @desc:
+"""
+import asyncio
+import json
+import re
+import time
+from functools import reduce
+from types import AsyncGeneratorType
+from typing import List, Dict
+
+from django.db.models import QuerySet
+from langchain.schema import HumanMessage, SystemMessage
+from langchain_core.messages import BaseMessage, AIMessage, AIMessageChunk, ToolMessage
+from langchain_mcp_adapters.client import MultiServerMCPClient
+from langgraph.prebuilt import create_react_agent
+
+from application.flow.i_step_node import NodeResult, INode
+from application.flow.step_node.ai_chat_step_node.i_chat_node import IChatNode
+from application.flow.tools import Reasoning
+from setting.models import Model
+from setting.models_provider import get_model_credential
+from setting.models_provider.tools import get_model_instance_by_model_user_id
+
+tool_message_template = """
+<details>
+    <summary>
+        <strong>Called MCP Tool: <em>%s</em></strong>
+    </summary>
+
+```json
+%s
+```
+</details>
+
+"""
+
+
+def _write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow, answer: str,
+                   reasoning_content: str):
+    chat_model = node_variable.get('chat_model')
+    message_tokens = chat_model.get_num_tokens_from_messages(node_variable.get('message_list'))
+    answer_tokens = chat_model.get_num_tokens(answer)
+    node.context['message_tokens'] = message_tokens
+    node.context['answer_tokens'] = answer_tokens
+    node.context['answer'] = answer
+    node.context['history_message'] = node_variable['history_message']
+    node.context['question'] = node_variable['question']
+    node.context['run_time'] = time.time() - node.context['start_time']
+    node.context['reasoning_content'] = reasoning_content
+    if workflow.is_result(node, NodeResult(node_variable, workflow_variable)):
+        node.answer_text = answer
+
+
+def write_context_stream(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据 (流式)
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result')
+    answer = ''
+    reasoning_content = ''
+    model_setting = node.context.get('model_setting',
+                                     {'reasoning_content_enable': False, 'reasoning_content_end': '</think>',
+                                      'reasoning_content_start': '<think>'})
+    reasoning = Reasoning(model_setting.get('reasoning_content_start', '<think>'),
+                          model_setting.get('reasoning_content_end', '</think>'))
+    response_reasoning_content = False
+
+    for chunk in response:
+        reasoning_chunk = reasoning.get_reasoning_content(chunk)
+        content_chunk = reasoning_chunk.get('content')
+        if 'reasoning_content' in chunk.additional_kwargs:
+            response_reasoning_content = True
+            reasoning_content_chunk = chunk.additional_kwargs.get('reasoning_content', '')
+        else:
+            reasoning_content_chunk = reasoning_chunk.get('reasoning_content')
+        answer += content_chunk
+        if reasoning_content_chunk is None:
+            reasoning_content_chunk = ''
+        reasoning_content += reasoning_content_chunk
+        yield {'content': content_chunk,
+               'reasoning_content': reasoning_content_chunk if model_setting.get('reasoning_content_enable',
+                                                                                 False) else ''}
+
+    reasoning_chunk = reasoning.get_end_reasoning_content()
+    answer += reasoning_chunk.get('content')
+    reasoning_content_chunk = ""
+    if not response_reasoning_content:
+        reasoning_content_chunk = reasoning_chunk.get(
+            'reasoning_content')
+    yield {'content': reasoning_chunk.get('content'),
+           'reasoning_content': reasoning_content_chunk if model_setting.get('reasoning_content_enable',
+                                                                             False) else ''}
+    _write_context(node_variable, workflow_variable, node, workflow, answer, reasoning_content)
+
+
+async def _yield_mcp_response(chat_model, message_list, mcp_servers):
+    async with MultiServerMCPClient(json.loads(mcp_servers)) as client:
+        agent = create_react_agent(chat_model, client.get_tools())
+        response = agent.astream({"messages": message_list}, stream_mode='messages')
+        async for chunk in response:
+            if isinstance(chunk[0], ToolMessage):
+                content = tool_message_template % (chunk[0].name, chunk[0].content)
+                chunk[0].content = content
+                yield chunk[0]
+            if isinstance(chunk[0], AIMessageChunk):
+                yield chunk[0]
+
+
+def mcp_response_generator(chat_model, message_list, mcp_servers):
+    loop = asyncio.new_event_loop()
+    try:
+        async_gen = _yield_mcp_response(chat_model, message_list, mcp_servers)
+        while True:
+            try:
+                chunk = loop.run_until_complete(anext_async(async_gen))
+                yield chunk
+            except StopAsyncIteration:
+                break
+    except Exception as e:
+        print(f'exception: {e}')
+    finally:
+        loop.close()
+
+
+async def anext_async(agen):
+    return await agen.__anext__()
+
+
+def write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点实例对象
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result')
+    model_setting = node.context.get('model_setting',
+                                     {'reasoning_content_enable': False, 'reasoning_content_end': '</think>',
+                                      'reasoning_content_start': '<think>'})
+    reasoning = Reasoning(model_setting.get('reasoning_content_start'), model_setting.get('reasoning_content_end'))
+    reasoning_result = reasoning.get_reasoning_content(response)
+    reasoning_result_end = reasoning.get_end_reasoning_content()
+    content = reasoning_result.get('content') + reasoning_result_end.get('content')
+    if 'reasoning_content' in response.response_metadata:
+        reasoning_content = response.response_metadata.get('reasoning_content', '')
+    else:
+        reasoning_content = reasoning_result.get('reasoning_content') + reasoning_result_end.get('reasoning_content')
+    _write_context(node_variable, workflow_variable, node, workflow, content, reasoning_content)
+
+
+def get_default_model_params_setting(model_id):
+    model = QuerySet(Model).filter(id=model_id).first()
+    credential = get_model_credential(model.provider, model.model_type, model.model_name)
+    model_params_setting = credential.get_model_params_setting_form(
+        model.model_name).get_default_form_data()
+    return model_params_setting
+
+
+def get_node_message(chat_record, runtime_node_id):
+    node_details = chat_record.get_node_details_runtime_node_id(runtime_node_id)
+    if node_details is None:
+        return []
+    return [HumanMessage(node_details.get('question')), AIMessage(node_details.get('answer'))]
+
+
+def get_workflow_message(chat_record):
+    return [chat_record.get_human_message(), chat_record.get_ai_message()]
+
+
+def get_message(chat_record, dialogue_type, runtime_node_id):
+    return get_node_message(chat_record, runtime_node_id) if dialogue_type == 'NODE' else get_workflow_message(
+        chat_record)
+
+
+class BaseChatNode(IChatNode):
+    def save_context(self, details, workflow_manage):
+        self.context['answer'] = details.get('answer')
+        self.context['question'] = details.get('question')
+        self.context['reasoning_content'] = details.get('reasoning_content')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('answer')
+
+    def execute(self, model_id, system, prompt, dialogue_number, history_chat_record, stream, chat_id, chat_record_id,
+                model_params_setting=None,
+                dialogue_type=None,
+                model_setting=None,
+                mcp_enable=False,
+                mcp_servers=None,
+                **kwargs) -> NodeResult:
+        if dialogue_type is None:
+            dialogue_type = 'WORKFLOW'
+
+        if model_params_setting is None:
+            model_params_setting = get_default_model_params_setting(model_id)
+        if model_setting is None:
+            model_setting = {'reasoning_content_enable': False, 'reasoning_content_end': '</think>',
+                             'reasoning_content_start': '<think>'}
+        self.context['model_setting'] = model_setting
+        chat_model = get_model_instance_by_model_user_id(model_id, self.flow_params_serializer.data.get('user_id'),
+                                                         **model_params_setting)
+        history_message = self.get_history_message(history_chat_record, dialogue_number, dialogue_type,
+                                                   self.runtime_node_id)
+        self.context['history_message'] = history_message
+        question = self.generate_prompt_question(prompt)
+        self.context['question'] = question.content
+        system = self.workflow_manage.generate_prompt(system)
+        self.context['system'] = system
+        message_list = self.generate_message_list(system, prompt, history_message)
+        self.context['message_list'] = message_list
+
+        if mcp_enable and mcp_servers is not None and '"stdio"' not in mcp_servers:
+            r = mcp_response_generator(chat_model, message_list, mcp_servers)
+            return NodeResult(
+                {'result': r, 'chat_model': chat_model, 'message_list': message_list,
+                 'history_message': history_message, 'question': question.content}, {},
+                _write_context=write_context_stream)
+
+        if stream:
+            r = chat_model.stream(message_list)
+            return NodeResult({'result': r, 'chat_model': chat_model, 'message_list': message_list,
+                               'history_message': history_message, 'question': question.content}, {},
+                              _write_context=write_context_stream)
+        else:
+            r = chat_model.invoke(message_list)
+            return NodeResult({'result': r, 'chat_model': chat_model, 'message_list': message_list,
+                               'history_message': history_message, 'question': question.content}, {},
+                              _write_context=write_context)
+
+    @staticmethod
+    def get_history_message(history_chat_record, dialogue_number, dialogue_type, runtime_node_id):
+        start_index = len(history_chat_record) - dialogue_number
+        history_message = reduce(lambda x, y: [*x, *y], [
+            get_message(history_chat_record[index], dialogue_type, runtime_node_id)
+            for index in
+            range(start_index if start_index > 0 else 0, len(history_chat_record))], [])
+        for message in history_message:
+            if isinstance(message.content, str):
+                message.content = re.sub('<form_rander>[\d\D]*?<\/form_rander>', '', message.content)
+        return history_message
+
+    def generate_prompt_question(self, prompt):
+        return HumanMessage(self.workflow_manage.generate_prompt(prompt))
+
+    def generate_message_list(self, system: str, prompt: str, history_message):
+        if system is not None and len(system) > 0:
+            return [SystemMessage(self.workflow_manage.generate_prompt(system)), *history_message,
+                    HumanMessage(self.workflow_manage.generate_prompt(prompt))]
+        else:
+            return [*history_message, HumanMessage(self.workflow_manage.generate_prompt(prompt))]
+
+    @staticmethod
+    def reset_message_list(message_list: List[BaseMessage], answer_text):
+        result = [{'role': 'user' if isinstance(message, HumanMessage) else 'ai', 'content': message.content} for
+                  message
+                  in
+                  message_list]
+        result.append({'role': 'ai', 'content': answer_text})
+        return result
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'system': self.context.get('system'),
+            'history_message': [{'content': message.content, 'role': message.type} for message in
+                                (self.context.get('history_message') if self.context.get(
+                                    'history_message') is not None else [])],
+            'question': self.context.get('question'),
+            'answer': self.context.get('answer'),
+            'reasoning_content': self.context.get('reasoning_content'),
+            'type': self.node.type,
+            'message_tokens': self.context.get('message_tokens'),
+            'answer_tokens': self.context.get('answer_tokens'),
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/application_node/__init__.py

@@ -0,0 +1,2 @@
+# coding=utf-8
+from .impl import *

apps/application/flow/step_node/application_node/i_application_node.py

@@ -0,0 +1,86 @@
+# coding=utf-8
+from typing import Type
+
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+
+from django.utils.translation import gettext_lazy as _
+
+
+class ApplicationNodeSerializer(serializers.Serializer):
+    application_id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Application ID")))
+    question_reference_address = serializers.ListField(required=True,
+                                                       error_messages=ErrMessage.list(_("User Questions")))
+    api_input_field_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("API Input Fields")))
+    user_input_field_list = serializers.ListField(required=False,
+                                                  error_messages=ErrMessage.uuid(_("User Input Fields")))
+    image_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("picture")))
+    document_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("document")))
+    audio_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("Audio")))
+    child_node = serializers.DictField(required=False, allow_null=True,
+                                       error_messages=ErrMessage.dict(_("Child Nodes")))
+    node_data = serializers.DictField(required=False, allow_null=True, error_messages=ErrMessage.dict(_("Form Data")))
+
+
+class IApplicationNode(INode):
+    type = 'application-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ApplicationNodeSerializer
+
+    def _run(self):
+        question = self.workflow_manage.get_reference_field(
+            self.node_params_serializer.data.get('question_reference_address')[0],
+            self.node_params_serializer.data.get('question_reference_address')[1:])
+        kwargs = {}
+        for api_input_field in self.node_params_serializer.data.get('api_input_field_list', []):
+            value = api_input_field.get('value', [''])[0] if api_input_field.get('value') else ''
+            kwargs[api_input_field['variable']] = self.workflow_manage.get_reference_field(value,
+                                                                                           api_input_field['value'][
+                                                                                           1:]) if value != '' else ''
+
+        for user_input_field in self.node_params_serializer.data.get('user_input_field_list', []):
+            value = user_input_field.get('value', [''])[0] if user_input_field.get('value') else ''
+            kwargs[user_input_field['field']] = self.workflow_manage.get_reference_field(value,
+                                                                                         user_input_field['value'][
+                                                                                         1:]) if value != '' else ''
+        # 判断是否包含这个属性
+        app_document_list = self.node_params_serializer.data.get('document_list', [])
+        if app_document_list and len(app_document_list) > 0:
+            app_document_list = self.workflow_manage.get_reference_field(
+                app_document_list[0],
+                app_document_list[1:])
+            for document in app_document_list:
+                if 'file_id' not in document:
+                    raise ValueError(
+                        _("Parameter value error: The uploaded document lacks file_id, and the document upload fails"))
+        app_image_list = self.node_params_serializer.data.get('image_list', [])
+        if app_image_list and len(app_image_list) > 0:
+            app_image_list = self.workflow_manage.get_reference_field(
+                app_image_list[0],
+                app_image_list[1:])
+            for image in app_image_list:
+                if 'file_id' not in image:
+                    raise ValueError(
+                        _("Parameter value error: The uploaded image lacks file_id, and the image upload fails"))
+
+        app_audio_list = self.node_params_serializer.data.get('audio_list', [])
+        if app_audio_list and len(app_audio_list) > 0:
+            app_audio_list = self.workflow_manage.get_reference_field(
+                app_audio_list[0],
+                app_audio_list[1:])
+            for audio in app_audio_list:
+                if 'file_id' not in audio:
+                    raise ValueError(
+                        _("Parameter value error: The uploaded audio lacks file_id, and the audio upload fails."))
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data,
+                            app_document_list=app_document_list, app_image_list=app_image_list,
+                            app_audio_list=app_audio_list,
+                            message=str(question), **kwargs)
+
+    def execute(self, application_id, message, chat_id, chat_record_id, stream, re_chat, client_id, client_type,
+                app_document_list=None, app_image_list=None, app_audio_list=None, child_node=None, node_data=None,
+                **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/application_node/impl/__init__.py

@@ -0,0 +1,2 @@
+# coding=utf-8
+from .base_application_node import BaseApplicationNode

apps/application/flow/step_node/application_node/impl/base_application_node.py

@@ -0,0 +1,267 @@
+# coding=utf-8
+import json
+import re
+import time
+import uuid
+from typing import Dict, List
+
+from application.flow.common import Answer
+from application.flow.i_step_node import NodeResult, INode
+from application.flow.step_node.application_node.i_application_node import IApplicationNode
+from application.models import Chat
+
+
+def string_to_uuid(input_str):
+    return str(uuid.uuid5(uuid.NAMESPACE_DNS, input_str))
+
+
+def _is_interrupt_exec(node, node_variable: Dict, workflow_variable: Dict):
+    return node_variable.get('is_interrupt_exec', False)
+
+
+def _write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow, answer: str,
+                   reasoning_content: str):
+    result = node_variable.get('result')
+    node.context['application_node_dict'] = node_variable.get('application_node_dict')
+    node.context['node_dict'] = node_variable.get('node_dict', {})
+    node.context['is_interrupt_exec'] = node_variable.get('is_interrupt_exec')
+    node.context['message_tokens'] = result.get('usage', {}).get('prompt_tokens', 0)
+    node.context['answer_tokens'] = result.get('usage', {}).get('completion_tokens', 0)
+    node.context['answer'] = answer
+    node.context['result'] = answer
+    node.context['reasoning_content'] = reasoning_content
+    node.context['question'] = node_variable['question']
+    node.context['run_time'] = time.time() - node.context['start_time']
+    if workflow.is_result(node, NodeResult(node_variable, workflow_variable)):
+        node.answer_text = answer
+
+
+def write_context_stream(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据 (流式)
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result')
+    answer = ''
+    reasoning_content = ''
+    usage = {}
+    node_child_node = {}
+    application_node_dict = node.context.get('application_node_dict', {})
+    is_interrupt_exec = False
+    for chunk in response:
+        # 先把流转成字符串
+        response_content = chunk.decode('utf-8')[6:]
+        response_content = json.loads(response_content)
+        content = response_content.get('content', '')
+        runtime_node_id = response_content.get('runtime_node_id', '')
+        chat_record_id = response_content.get('chat_record_id', '')
+        child_node = response_content.get('child_node')
+        view_type = response_content.get('view_type')
+        node_type = response_content.get('node_type')
+        real_node_id = response_content.get('real_node_id')
+        node_is_end = response_content.get('node_is_end', False)
+        _reasoning_content = response_content.get('reasoning_content', '')
+        if node_type == 'form-node':
+            is_interrupt_exec = True
+        answer += content
+        reasoning_content += _reasoning_content
+        node_child_node = {'runtime_node_id': runtime_node_id, 'chat_record_id': chat_record_id,
+                           'child_node': child_node}
+
+        if real_node_id is not None:
+            application_node = application_node_dict.get(real_node_id, None)
+            if application_node is None:
+
+                application_node_dict[real_node_id] = {'content': content,
+                                                       'runtime_node_id': runtime_node_id,
+                                                       'chat_record_id': chat_record_id,
+                                                       'child_node': child_node,
+                                                       'index': len(application_node_dict),
+                                                       'view_type': view_type,
+                                                       'reasoning_content': _reasoning_content}
+            else:
+                application_node['content'] += content
+                application_node['reasoning_content'] += _reasoning_content
+
+        yield {'content': content,
+               'node_type': node_type,
+               'runtime_node_id': runtime_node_id, 'chat_record_id': chat_record_id,
+               'reasoning_content': _reasoning_content,
+               'child_node': child_node,
+               'real_node_id': real_node_id,
+               'node_is_end': node_is_end,
+               'view_type': view_type}
+        usage = response_content.get('usage', {})
+    node_variable['result'] = {'usage': usage}
+    node_variable['is_interrupt_exec'] = is_interrupt_exec
+    node_variable['child_node'] = node_child_node
+    node_variable['application_node_dict'] = application_node_dict
+    _write_context(node_variable, workflow_variable, node, workflow, answer, reasoning_content)
+
+
+def write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点实例对象
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result', {}).get('data', {})
+    node_variable['result'] = {'usage': {'completion_tokens': response.get('completion_tokens'),
+                                         'prompt_tokens': response.get('prompt_tokens')}}
+    answer = response.get('content', '') or "抱歉，没有查找到相关内容，请重新描述您的问题或提供更多信息。"
+    reasoning_content = response.get('reasoning_content', '')
+    answer_list = response.get('answer_list', [])
+    node_variable['application_node_dict'] = {answer.get('real_node_id'): {**answer, 'index': index} for answer, index
+                                              in
+                                              zip(answer_list, range(len(answer_list)))}
+    _write_context(node_variable, workflow_variable, node, workflow, answer, reasoning_content)
+
+
+def reset_application_node_dict(application_node_dict, runtime_node_id, node_data):
+    try:
+        if application_node_dict is None:
+            return
+        for key in application_node_dict:
+            application_node = application_node_dict[key]
+            if application_node.get('runtime_node_id') == runtime_node_id:
+                content: str = application_node.get('content')
+                match = re.search('<form_rander>.*?</form_rander>', content)
+                if match:
+                    form_setting_str = match.group().replace('<form_rander>', '').replace('</form_rander>', '')
+                    form_setting = json.loads(form_setting_str)
+                    form_setting['is_submit'] = True
+                    form_setting['form_data'] = node_data
+                    value = f'<form_rander>{json.dumps(form_setting)}</form_rander>'
+                    res = re.sub('<form_rander>.*?</form_rander>',
+                                 '${value}', content)
+                    application_node['content'] = res.replace('${value}', value)
+    except Exception as e:
+        pass
+
+
+class BaseApplicationNode(IApplicationNode):
+    def get_answer_list(self) -> List[Answer] | None:
+        if self.answer_text is None:
+            return None
+        application_node_dict = self.context.get('application_node_dict')
+        if application_node_dict is None or len(application_node_dict) == 0:
+            return [
+                Answer(self.answer_text, self.view_type, self.runtime_node_id, self.workflow_params['chat_record_id'],
+                       self.context.get('child_node'), self.runtime_node_id, '')]
+        else:
+            return [Answer(n.get('content'), n.get('view_type'), self.runtime_node_id,
+                           self.workflow_params['chat_record_id'], {'runtime_node_id': n.get('runtime_node_id'),
+                                                                    'chat_record_id': n.get('chat_record_id')
+                               , 'child_node': n.get('child_node')}, n.get('real_node_id'),
+                           n.get('reasoning_content', ''))
+                    for n in
+                    sorted(application_node_dict.values(), key=lambda item: item.get('index'))]
+
+    def save_context(self, details, workflow_manage):
+        self.context['answer'] = details.get('answer')
+        self.context['result'] = details.get('answer')
+        self.context['question'] = details.get('question')
+        self.context['type'] = details.get('type')
+        self.context['reasoning_content'] = details.get('reasoning_content')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('answer')
+
+    def execute(self, application_id, message, chat_id, chat_record_id, stream, re_chat, client_id, client_type,
+                app_document_list=None, app_image_list=None, app_audio_list=None, child_node=None, node_data=None,
+                **kwargs) -> NodeResult:
+        from application.serializers.chat_message_serializers import ChatMessageSerializer
+        # 生成嵌入应用的chat_id
+        current_chat_id = string_to_uuid(chat_id + application_id)
+        Chat.objects.get_or_create(id=current_chat_id, defaults={
+            'application_id': application_id,
+            'abstract': message[0:1024],
+            'client_id': client_id,
+        })
+        if app_document_list is None:
+            app_document_list = []
+        if app_image_list is None:
+            app_image_list = []
+        if app_audio_list is None:
+            app_audio_list = []
+        runtime_node_id = None
+        record_id = None
+        child_node_value = None
+        if child_node is not None:
+            runtime_node_id = child_node.get('runtime_node_id')
+            record_id = child_node.get('chat_record_id')
+            child_node_value = child_node.get('child_node')
+            application_node_dict = self.context.get('application_node_dict')
+            reset_application_node_dict(application_node_dict, runtime_node_id, node_data)
+
+        response = ChatMessageSerializer(
+            data={'chat_id': current_chat_id, 'message': message,
+                  're_chat': re_chat,
+                  'stream': stream,
+                  'application_id': application_id,
+                  'client_id': client_id,
+                  'client_type': client_type,
+                  'document_list': app_document_list,
+                  'image_list': app_image_list,
+                  'audio_list': app_audio_list,
+                  'runtime_node_id': runtime_node_id,
+                  'chat_record_id': record_id,
+                  'child_node': child_node_value,
+                  'node_data': node_data,
+                  'form_data': kwargs}).chat()
+        if response.status_code == 200:
+            if stream:
+                content_generator = response.streaming_content
+                return NodeResult({'result': content_generator, 'question': message}, {},
+                                  _write_context=write_context_stream, _is_interrupt=_is_interrupt_exec)
+            else:
+                data = json.loads(response.content)
+                return NodeResult({'result': data, 'question': message}, {},
+                                  _write_context=write_context, _is_interrupt=_is_interrupt_exec)
+
+    def get_details(self, index: int, **kwargs):
+        global_fields = []
+        for api_input_field in self.node_params_serializer.data.get('api_input_field_list', []):
+            value = api_input_field.get('value', [''])[0] if api_input_field.get('value') else ''
+            global_fields.append({
+                'label': api_input_field['variable'],
+                'key': api_input_field['variable'],
+                'value': self.workflow_manage.get_reference_field(
+                    value,
+                    api_input_field['value'][1:]
+                ) if value != '' else ''
+            })
+
+        for user_input_field in self.node_params_serializer.data.get('user_input_field_list', []):
+            value = user_input_field.get('value', [''])[0] if user_input_field.get('value') else ''
+            global_fields.append({
+                'label': user_input_field['label'],
+                'key': user_input_field['field'],
+                'value': self.workflow_manage.get_reference_field(
+                    value,
+                    user_input_field['value'][1:]
+                ) if value != '' else ''
+            })
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            "info": self.node.properties.get('node_data'),
+            'run_time': self.context.get('run_time'),
+            'question': self.context.get('question'),
+            'answer': self.context.get('answer'),
+            'reasoning_content': self.context.get('reasoning_content'),
+            'type': self.node.type,
+            'message_tokens': self.context.get('message_tokens'),
+            'answer_tokens': self.context.get('answer_tokens'),
+            'status': self.status,
+            'err_message': self.err_message,
+            'global_fields': global_fields,
+            'document_list': self.workflow_manage.document_list,
+            'image_list': self.workflow_manage.image_list,
+            'audio_list': self.workflow_manage.audio_list,
+            'application_node_dict': self.context.get('application_node_dict')
+        }

apps/application/flow/step_node/condition_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/6/7 14:43
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/condition_node/compare/__init__.py

@@ -0,0 +1,30 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/6/7 14:43
+    @desc:
+"""
+
+from .contain_compare import *
+from .equal_compare import *
+from .ge_compare import *
+from .gt_compare import *
+from .is_not_null_compare import *
+from .is_not_true import IsNotTrueCompare
+from .is_null_compare import *
+from .is_true import IsTrueCompare
+from .le_compare import *
+from .len_equal_compare import *
+from .len_ge_compare import *
+from .len_gt_compare import *
+from .len_le_compare import *
+from .len_lt_compare import *
+from .lt_compare import *
+from .not_contain_compare import *
+
+compare_handle_list = [GECompare(), GTCompare(), ContainCompare(), EqualCompare(), LTCompare(), LECompare(),
+                       LenLECompare(), LenGECompare(), LenEqualCompare(), LenGTCompare(), LenLTCompare(),
+                       IsNullCompare(),
+                       IsNotNullCompare(), NotContainCompare(), IsTrueCompare(), IsNotTrueCompare()]

apps/application/flow/step_node/condition_node/compare/compare.py

@@ -0,0 +1,20 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： compare.py
+    @date：2024/6/7 14:37
+    @desc:
+"""
+from abc import abstractmethod
+from typing import List
+
+
+class Compare:
+    @abstractmethod
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        pass
+
+    @abstractmethod
+    def compare(self, source_value, compare, target_value):
+        pass

apps/application/flow/step_node/condition_node/compare/contain_compare.py

@@ -0,0 +1,23 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： contain_compare.py
+    @date：2024/6/11 10:02
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class ContainCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'contain':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        if isinstance(source_value, str):
+            return str(target_value) in source_value
+        return any([str(item) == str(target_value) for item in source_value])

apps/application/flow/step_node/condition_node/compare/equal_compare.py

@@ -0,0 +1,21 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： equal_compare.py
+    @date：2024/6/7 14:44
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class EqualCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'eq':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        return str(source_value) == str(target_value)

apps/application/flow/step_node/condition_node/compare/ge_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 大于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class GECompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'ge':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return float(source_value) >= float(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/gt_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 大于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class GTCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'gt':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return float(source_value) > float(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/is_not_null_compare.py

@@ -0,0 +1,21 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： is_not_null_compare.py
+    @date：2024/6/28 10:45
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare import Compare
+
+
+class IsNotNullCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'is_not_null':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        return source_value is not None and len(source_value) > 0

apps/application/flow/step_node/condition_node/compare/is_not_true.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： is_not_true.py
+    @date：2025/4/7 13:44
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare import Compare
+
+
+class IsNotTrueCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'is_not_true':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return source_value is False
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/is_null_compare.py

@@ -0,0 +1,21 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： is_null_compare.py
+    @date：2024/6/28 10:45
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare import Compare
+
+
+class IsNullCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'is_null':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        return source_value is None or len(source_value) == 0

apps/application/flow/step_node/condition_node/compare/is_true.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： IsTrue.py
+    @date：2025/4/7 13:38
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare import Compare
+
+
+class IsTrueCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'is_true':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return source_value is True
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/le_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 小于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LECompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'le':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return float(source_value) <= float(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/len_equal_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： equal_compare.py
+    @date：2024/6/7 14:44
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LenEqualCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'len_eq':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return len(source_value) == int(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/len_ge_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 大于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LenGECompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'len_ge':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return len(source_value) >= int(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/len_gt_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 大于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LenGTCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'len_gt':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return len(source_value) > int(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/len_le_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 小于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LenLECompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'len_le':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return len(source_value) <= int(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/len_lt_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 小于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LenLTCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'len_lt':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return len(source_value) < int(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/lt_compare.py

@@ -0,0 +1,24 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： lt_compare.py
+    @date：2024/6/11 9:52
+    @desc: 小于比较器
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class LTCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'lt':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        try:
+            return float(source_value) < float(target_value)
+        except Exception as e:
+            return False

apps/application/flow/step_node/condition_node/compare/not_contain_compare.py

@@ -0,0 +1,23 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： contain_compare.py
+    @date：2024/6/11 10:02
+    @desc:
+"""
+from typing import List
+
+from application.flow.step_node.condition_node.compare.compare import Compare
+
+
+class NotContainCompare(Compare):
+
+    def support(self, node_id, fields: List[str], source_value, compare, target_value):
+        if compare == 'not_contain':
+            return True
+
+    def compare(self, source_value, compare, target_value):
+        if isinstance(source_value, str):
+            return str(target_value) not in source_value
+        return not any([str(item) == str(target_value) for item in source_value])

apps/application/flow/step_node/condition_node/i_condition_node.py

@@ -0,0 +1,39 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： i_condition_node.py
+    @date：2024/6/7 9:54
+    @desc:
+"""
+from typing import Type
+
+from django.utils.translation import gettext_lazy as _
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode
+from common.util.field_message import ErrMessage
+
+
+class ConditionSerializer(serializers.Serializer):
+    compare = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Comparator")))
+    value = serializers.CharField(required=True, error_messages=ErrMessage.char(_("value")))
+    field = serializers.ListField(required=True, error_messages=ErrMessage.char(_("Fields")))
+
+
+class ConditionBranchSerializer(serializers.Serializer):
+    id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Branch id")))
+    type = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Branch Type")))
+    condition = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Condition or|and")))
+    conditions = ConditionSerializer(many=True)
+
+
+class ConditionNodeParamsSerializer(serializers.Serializer):
+    branch = ConditionBranchSerializer(many=True)
+
+
+class IConditionNode(INode):
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ConditionNodeParamsSerializer
+
+    type = 'condition-node'

apps/application/flow/step_node/condition_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py
+    @date：2024/6/11 15:35
+    @desc:
+"""
+from .base_condition_node import BaseConditionNode

apps/application/flow/step_node/condition_node/impl/base_condition_node.py

@@ -0,0 +1,62 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： base_condition_node.py
+    @date：2024/6/7 11:29
+    @desc:
+"""
+from typing import List
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.condition_node.compare import compare_handle_list
+from application.flow.step_node.condition_node.i_condition_node import IConditionNode
+
+
+class BaseConditionNode(IConditionNode):
+    def save_context(self, details, workflow_manage):
+        self.context['branch_id'] = details.get('branch_id')
+        self.context['branch_name'] = details.get('branch_name')
+
+    def execute(self, **kwargs) -> NodeResult:
+        branch_list = self.node_params_serializer.data['branch']
+        branch = self._execute(branch_list)
+        r = NodeResult({'branch_id': branch.get('id'), 'branch_name': branch.get('type')}, {})
+        return r
+
+    def _execute(self, branch_list: List):
+        for branch in branch_list:
+            if self.branch_assertion(branch):
+                return branch
+
+    def branch_assertion(self, branch):
+        condition_list = [self.assertion(row.get('field'), row.get('compare'), row.get('value')) for row in
+                          branch.get('conditions')]
+        condition = branch.get('condition')
+        return all(condition_list) if condition == 'and' else any(condition_list)
+
+    def assertion(self, field_list: List[str], compare: str, value):
+        try:
+            value = self.workflow_manage.generate_prompt(value)
+        except Exception as e:
+            pass
+        field_value = None
+        try:
+            field_value = self.workflow_manage.get_reference_field(field_list[0], field_list[1:])
+        except  Exception as e:
+            pass
+        for compare_handler in compare_handle_list:
+            if compare_handler.support(field_list[0], field_list[1:], field_value, compare, value):
+                return compare_handler.compare(field_value, compare, value)
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'branch_id': self.context.get('branch_id'),
+            'branch_name': self.context.get('branch_name'),
+            'type': self.node.type,
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/direct_reply_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py
+    @date：2024/6/11 17:50
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/direct_reply_node/i_reply_node.py

@@ -0,0 +1,48 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： i_reply_node.py
+    @date：2024/6/11 16:25
+    @desc:
+"""
+from typing import Type
+
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.exception.app_exception import AppApiException
+from common.util.field_message import ErrMessage
+from django.utils.translation import gettext_lazy as _
+
+
+class ReplyNodeParamsSerializer(serializers.Serializer):
+    reply_type = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Response Type")))
+    fields = serializers.ListField(required=False, error_messages=ErrMessage.list(_("Reference Field")))
+    content = serializers.CharField(required=False, allow_blank=True, allow_null=True,
+                                    error_messages=ErrMessage.char(_("Direct answer content")))
+    is_result = serializers.BooleanField(required=False, error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    def is_valid(self, *, raise_exception=False):
+        super().is_valid(raise_exception=True)
+        if self.data.get('reply_type') == 'referencing':
+            if 'fields' not in self.data:
+                raise AppApiException(500, _("Reference field cannot be empty"))
+            if len(self.data.get('fields')) < 2:
+                raise AppApiException(500, _("Reference field error"))
+        else:
+            if 'content' not in self.data or self.data.get('content') is None:
+                raise AppApiException(500, _("Content cannot be empty"))
+
+
+class IReplyNode(INode):
+    type = 'reply-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ReplyNodeParamsSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, reply_type, stream, fields=None, content=None, **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/direct_reply_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： __init__.py
+    @date：2024/6/11 17:49
+    @desc:
+"""
+from .base_reply_node import *

apps/application/flow/step_node/direct_reply_node/impl/base_reply_node.py

@@ -0,0 +1,45 @@
+# coding=utf-8
+"""
+    @project: maxkb
+    @Author：虎
+    @file： base_reply_node.py
+    @date：2024/6/11 17:25
+    @desc:
+"""
+from typing import List
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.direct_reply_node.i_reply_node import IReplyNode
+
+
+class BaseReplyNode(IReplyNode):
+    def save_context(self, details, workflow_manage):
+        self.context['answer'] = details.get('answer')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('answer')
+
+    def execute(self, reply_type, stream, fields=None, content=None, **kwargs) -> NodeResult:
+        if reply_type == 'referencing':
+            result = self.get_reference_content(fields)
+        else:
+            result = self.generate_reply_content(content)
+        return NodeResult({'answer': result}, {})
+
+    def generate_reply_content(self, prompt):
+        return self.workflow_manage.generate_prompt(prompt)
+
+    def get_reference_content(self, fields: List[str]):
+        return str(self.workflow_manage.get_reference_field(
+            fields[0],
+            fields[1:]))
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'type': self.node.type,
+            'answer': self.context.get('answer'),
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/document_extract_node/__init__.py

@@ -0,0 +1 @@
+from .impl import *

apps/application/flow/step_node/document_extract_node/i_document_extract_node.py

@@ -0,0 +1,28 @@
+# coding=utf-8
+
+from typing import Type
+
+from django.utils.translation import gettext_lazy as _
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+
+
+class DocumentExtractNodeSerializer(serializers.Serializer):
+    document_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("document")))
+
+
+class IDocumentExtractNode(INode):
+    type = 'document-extract-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return DocumentExtractNodeSerializer
+
+    def _run(self):
+        res = self.workflow_manage.get_reference_field(self.node_params_serializer.data.get('document_list')[0],
+                                                       self.node_params_serializer.data.get('document_list')[1:])
+        return self.execute(document=res, **self.flow_params_serializer.data)
+
+    def execute(self, document, chat_id, **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/document_extract_node/impl/__init__.py

@@ -0,0 +1 @@
+from .base_document_extract_node import BaseDocumentExtractNode

apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py

@@ -0,0 +1,94 @@
+# coding=utf-8
+import io
+import mimetypes
+
+from django.core.files.uploadedfile import InMemoryUploadedFile
+from django.db.models import QuerySet
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.document_extract_node.i_document_extract_node import IDocumentExtractNode
+from dataset.models import File
+from dataset.serializers.document_serializers import split_handles, parse_table_handle_list, FileBufferHandle
+from dataset.serializers.file_serializers import FileSerializer
+
+
+def bytes_to_uploaded_file(file_bytes, file_name="file.txt"):
+    content_type, _ = mimetypes.guess_type(file_name)
+    if content_type is None:
+        # 如果未能识别，设置为默认的二进制文件类型
+        content_type = "application/octet-stream"
+    # 创建一个内存中的字节流对象
+    file_stream = io.BytesIO(file_bytes)
+
+    # 获取文件大小
+    file_size = len(file_bytes)
+
+    # 创建 InMemoryUploadedFile 对象
+    uploaded_file = InMemoryUploadedFile(
+        file=file_stream,
+        field_name=None,
+        name=file_name,
+        content_type=content_type,
+        size=file_size,
+        charset=None,
+    )
+    return uploaded_file
+
+
+splitter = '\n`-----------------------------------`\n'
+
+class BaseDocumentExtractNode(IDocumentExtractNode):
+    def save_context(self, details, workflow_manage):
+        self.context['content'] = details.get('content')
+
+
+    def execute(self, document, chat_id, **kwargs):
+        get_buffer = FileBufferHandle().get_buffer
+
+        self.context['document_list'] = document
+        content = []
+        if document is None or not isinstance(document, list):
+            return NodeResult({'content': ''}, {})
+
+        application = self.workflow_manage.work_flow_post_handler.chat_info.application
+
+        # doc文件中的图片保存
+        def save_image(image_list):
+            for image in image_list:
+                meta = {
+                    'debug': False if application.id else True,
+                    'chat_id': chat_id,
+                    'application_id': str(application.id) if application.id else None,
+                    'file_id': str(image.id)
+                }
+                file = bytes_to_uploaded_file(image.image, image.image_name)
+                FileSerializer(data={'file': file, 'meta': meta}).upload()
+
+        for doc in document:
+            file = QuerySet(File).filter(id=doc['file_id']).first()
+            buffer = io.BytesIO(file.get_byte().tobytes())
+            buffer.name = doc['name']  # this is the important line
+
+            for split_handle in (parse_table_handle_list + split_handles):
+                if split_handle.support(buffer, get_buffer):
+                    # 回到文件头
+                    buffer.seek(0)
+                    file_content = split_handle.get_content(buffer, save_image)
+                    content.append('### ' + doc['name'] + '\n' + file_content)
+                    break
+
+        return NodeResult({'content': splitter.join(content)}, {})
+
+    def get_details(self, index: int, **kwargs):
+        content = self.context.get('content', '').split(splitter)
+        # 不保存content全部内容，因为content内容可能会很大
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'type': self.node.type,
+            'content': [file_content[:500] for file_content in content],
+            'status': self.status,
+            'err_message': self.err_message,
+            'document_list': self.context.get('document_list')
+        }

apps/application/flow/step_node/form_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/11/4 14:48
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/form_node/i_form_node.py

@@ -0,0 +1,35 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： i_form_node.py
+    @date：2024/11/4 14:48
+    @desc:
+"""
+from typing import Type
+
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+from django.utils.translation import gettext_lazy as _
+
+
+class FormNodeParamsSerializer(serializers.Serializer):
+    form_field_list = serializers.ListField(required=True, error_messages=ErrMessage.list(_("Form Configuration")))
+    form_content_format = serializers.CharField(required=True, error_messages=ErrMessage.char(_('Form output content')))
+    form_data = serializers.DictField(required=False, allow_null=True, error_messages=ErrMessage.dict(_("Form Data")))
+
+
+class IFormNode(INode):
+    type = 'form-node'
+    view_type = 'single_view'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return FormNodeParamsSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, form_field_list, form_content_format, form_data, **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/form_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/11/4 14:49
+    @desc:
+"""
+from .base_form_node import BaseFormNode

apps/application/flow/step_node/form_node/impl/base_form_node.py

@@ -0,0 +1,107 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： base_form_node.py
+    @date：2024/11/4 14:52
+    @desc:
+"""
+import json
+import time
+from typing import Dict, List
+
+from langchain_core.prompts import PromptTemplate
+
+from application.flow.common import Answer
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.form_node.i_form_node import IFormNode
+
+
+def write_context(step_variable: Dict, global_variable: Dict, node, workflow):
+    if step_variable is not None:
+        for key in step_variable:
+            node.context[key] = step_variable[key]
+        if workflow.is_result(node, NodeResult(step_variable, global_variable)) and 'result' in step_variable:
+            result = step_variable['result']
+            yield result
+            node.answer_text = result
+    node.context['run_time'] = time.time() - node.context['start_time']
+
+
+class BaseFormNode(IFormNode):
+    def save_context(self, details, workflow_manage):
+        form_data = details.get('form_data', None)
+        self.context['result'] = details.get('result')
+        self.context['form_content_format'] = details.get('form_content_format')
+        self.context['form_field_list'] = details.get('form_field_list')
+        self.context['run_time'] = details.get('run_time')
+        self.context['start_time'] = details.get('start_time')
+        self.context['form_data'] = form_data
+        self.context['is_submit'] = details.get('is_submit')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('result')
+        if form_data is not None:
+            for key in form_data:
+                self.context[key] = form_data[key]
+
+    def execute(self, form_field_list, form_content_format, form_data, **kwargs) -> NodeResult:
+        if form_data is not None:
+            self.context['is_submit'] = True
+            self.context['form_data'] = form_data
+            for key in form_data:
+                self.context[key] = form_data.get(key)
+        else:
+            self.context['is_submit'] = False
+        form_setting = {"form_field_list": form_field_list, "runtime_node_id": self.runtime_node_id,
+                        "chat_record_id": self.flow_params_serializer.data.get("chat_record_id"),
+                        "is_submit": self.context.get("is_submit", False)}
+        form = f'<form_rander>{json.dumps(form_setting, ensure_ascii=False)}</form_rander>'
+        context = self.workflow_manage.get_workflow_content()
+        form_content_format = self.workflow_manage.reset_prompt(form_content_format)
+        prompt_template = PromptTemplate.from_template(form_content_format, template_format='jinja2')
+        value = prompt_template.format(form=form, context=context)
+        return NodeResult(
+            {'result': value, 'form_field_list': form_field_list, 'form_content_format': form_content_format}, {},
+            _write_context=write_context)
+
+    def get_answer_list(self) -> List[Answer] | None:
+        form_content_format = self.context.get('form_content_format')
+        form_field_list = self.context.get('form_field_list')
+        form_setting = {"form_field_list": form_field_list, "runtime_node_id": self.runtime_node_id,
+                        "chat_record_id": self.flow_params_serializer.data.get("chat_record_id"),
+                        'form_data': self.context.get('form_data', {}),
+                        "is_submit": self.context.get("is_submit", False)}
+        form = f'<form_rander>{json.dumps(form_setting, ensure_ascii=False)}</form_rander>'
+        context = self.workflow_manage.get_workflow_content()
+        form_content_format = self.workflow_manage.reset_prompt(form_content_format)
+        prompt_template = PromptTemplate.from_template(form_content_format, template_format='jinja2')
+        value = prompt_template.format(form=form, context=context)
+        return [Answer(value, self.view_type, self.runtime_node_id, self.workflow_params['chat_record_id'], None,
+                       self.runtime_node_id, '')]
+
+    def get_details(self, index: int, **kwargs):
+        form_content_format = self.context.get('form_content_format')
+        form_field_list = self.context.get('form_field_list')
+        form_setting = {"form_field_list": form_field_list, "runtime_node_id": self.runtime_node_id,
+                        "chat_record_id": self.flow_params_serializer.data.get("chat_record_id"),
+                        'form_data': self.context.get('form_data', {}),
+                        "is_submit": self.context.get("is_submit", False)}
+        form = f'<form_rander>{json.dumps(form_setting, ensure_ascii=False)}</form_rander>'
+        context = self.workflow_manage.get_workflow_content()
+        form_content_format = self.workflow_manage.reset_prompt(form_content_format)
+        prompt_template = PromptTemplate.from_template(form_content_format, template_format='jinja2')
+        value = prompt_template.format(form=form, context=context)
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            "result": value,
+            "form_content_format": self.context.get('form_content_format'),
+            "form_field_list": self.context.get('form_field_list'),
+            'form_data': self.context.get('form_data'),
+            'start_time': self.context.get('start_time'),
+            'is_submit': self.context.get('is_submit'),
+            'run_time': self.context.get('run_time'),
+            'type': self.node.type,
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/function_lib_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py
+    @date：2024/8/8 17:45
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/function_lib_node/i_function_lib_node.py

@@ -0,0 +1,48 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： i_function_lib_node.py
+    @date：2024/8/8 16:21
+    @desc:
+"""
+from typing import Type
+
+from django.db.models import QuerySet
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.field.common import ObjectField
+from common.util.field_message import ErrMessage
+from function_lib.models.function import FunctionLib
+from django.utils.translation import gettext_lazy as _
+
+
+class InputField(serializers.Serializer):
+    name = serializers.CharField(required=True, error_messages=ErrMessage.char(_('Variable Name')))
+    value = ObjectField(required=True, error_messages=ErrMessage.char(_("Variable Value")), model_type_list=[str, list])
+
+
+class FunctionLibNodeParamsSerializer(serializers.Serializer):
+    function_lib_id = serializers.UUIDField(required=True, error_messages=ErrMessage.uuid(_('Library ID')))
+    input_field_list = InputField(required=True, many=True)
+    is_result = serializers.BooleanField(required=False, error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    def is_valid(self, *, raise_exception=False):
+        super().is_valid(raise_exception=True)
+        f_lib = QuerySet(FunctionLib).filter(id=self.data.get('function_lib_id')).first()
+        if f_lib is None:
+            raise Exception(_('The function has been deleted'))
+
+
+class IFunctionLibNode(INode):
+    type = 'function-lib-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return FunctionLibNodeParamsSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, function_lib_id, input_field_list, **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/function_lib_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py
+    @date：2024/8/8 17:48
+    @desc:
+"""
+from .base_function_lib_node import BaseFunctionLibNodeNode

apps/application/flow/step_node/function_lib_node/impl/base_function_lib_node.py

@@ -0,0 +1,150 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： base_function_lib_node.py
+    @date：2024/8/8 17:49
+    @desc:
+"""
+import json
+import time
+from typing import Dict
+
+from django.db.models import QuerySet
+from django.utils.translation import gettext as _
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.function_lib_node.i_function_lib_node import IFunctionLibNode
+from common.exception.app_exception import AppApiException
+from common.util.function_code import FunctionExecutor
+from common.util.rsa_util import rsa_long_decrypt
+from function_lib.models.function import FunctionLib
+from smartdoc.const import CONFIG
+
+function_executor = FunctionExecutor(CONFIG.get('SANDBOX'))
+
+
+def write_context(step_variable: Dict, global_variable: Dict, node, workflow):
+    if step_variable is not None:
+        for key in step_variable:
+            node.context[key] = step_variable[key]
+        if workflow.is_result(node, NodeResult(step_variable, global_variable)) and 'result' in step_variable:
+            result = str(step_variable['result']) + '\n'
+            yield result
+            node.answer_text = result
+    node.context['run_time'] = time.time() - node.context['start_time']
+
+
+def get_field_value(debug_field_list, name, is_required):
+    result = [field for field in debug_field_list if field.get('name') == name]
+    if len(result) > 0:
+        return result[-1]['value']
+    if is_required:
+        raise AppApiException(500, _('Field: {name} No value set').format(name=name))
+    return None
+
+
+def valid_reference_value(_type, value, name):
+    if _type == 'int':
+        instance_type = int | float
+    elif _type == 'float':
+        instance_type = float | int
+    elif _type == 'dict':
+        instance_type = dict
+    elif _type == 'array':
+        instance_type = list
+    elif _type == 'string':
+        instance_type = str
+    else:
+        raise Exception(_('Field: {name} Type: {_type} Value: {value} Unsupported types').format(name=name,
+                                                                                                 _type=_type))
+    if not isinstance(value, instance_type):
+        raise Exception(
+            _('Field: {name} Type: {_type} Value: {value} Type error').format(name=name, _type=_type,
+                                                                              value=value))
+
+
+def convert_value(name: str, value, _type, is_required, source, node):
+    if not is_required and (value is None or (isinstance(value, str) and len(value) == 0)):
+        return None
+    if not is_required and source == 'reference' and (value is None or len(value) == 0):
+        return None
+    if source == 'reference':
+        value = node.workflow_manage.get_reference_field(
+            value[0],
+            value[1:])
+        valid_reference_value(_type, value, name)
+        if _type == 'int':
+            return int(value)
+        if _type == 'float':
+            return float(value)
+        return value
+    try:
+        if _type == 'int':
+            return int(value)
+        if _type == 'float':
+            return float(value)
+        if _type == 'dict':
+            v = json.loads(value)
+            if isinstance(v, dict):
+                return v
+            raise Exception(_('type error'))
+        if _type == 'array':
+            v = json.loads(value)
+            if isinstance(v, list):
+                return v
+            raise Exception(_('type error'))
+        return value
+    except Exception as e:
+        raise Exception(
+            _('Field: {name} Type: {_type} Value: {value} Type error').format(name=name, _type=_type,
+                                                                              value=value))
+
+
+def valid_function(function_lib, user_id):
+    if function_lib is None:
+        raise Exception(_('Function does not exist'))
+    if function_lib.permission_type == 'PRIVATE' and str(function_lib.user_id) != str(user_id):
+        raise Exception(_('No permission to use this function {name}').format(name=function_lib.name))
+    if not function_lib.is_active:
+        raise Exception(_('Function {name} is unavailable').format(name=function_lib.name))
+
+
+class BaseFunctionLibNodeNode(IFunctionLibNode):
+    def save_context(self, details, workflow_manage):
+        self.context['result'] = details.get('result')
+        if self.node_params.get('is_result'):
+            self.answer_text = str(details.get('result'))
+
+    def execute(self, function_lib_id, input_field_list, **kwargs) -> NodeResult:
+        function_lib = QuerySet(FunctionLib).filter(id=function_lib_id).first()
+        valid_function(function_lib, self.flow_params_serializer.data.get('user_id'))
+        params = {field.get('name'): convert_value(field.get('name'), field.get('value'), field.get('type'),
+                                                   field.get('is_required'),
+                                                   field.get('source'), self)
+                  for field in
+                  [{'value': get_field_value(input_field_list, field.get('name'), field.get('is_required'),
+                                             ), **field}
+                   for field in
+                   function_lib.input_field_list]}
+
+        self.context['params'] = params
+        # 合并初始化参数
+        if function_lib.init_params is not None:
+            all_params = json.loads(rsa_long_decrypt(function_lib.init_params)) | params
+        else:
+            all_params = params
+        result = function_executor.exec_code(function_lib.code, all_params)
+        return NodeResult({'result': result}, {}, _write_context=write_context)
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            "result": self.context.get('result'),
+            "params": self.context.get('params'),
+            'run_time': self.context.get('run_time'),
+            'type': self.node.type,
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/function_node/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/8/13 10:43
+    @desc:
+"""
+from .impl import *

apps/application/flow/step_node/function_node/i_function_node.py

@@ -0,0 +1,63 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： i_function_lib_node.py
+    @date：2024/8/8 16:21
+    @desc:
+"""
+import re
+from typing import Type
+
+from django.core import validators
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.exception.app_exception import AppApiException
+from common.field.common import ObjectField
+from common.util.field_message import ErrMessage
+from django.utils.translation import gettext_lazy as _
+from rest_framework.utils.formatting import lazy_format
+
+
+class InputField(serializers.Serializer):
+    name = serializers.CharField(required=True, error_messages=ErrMessage.char(_('Variable Name')))
+    is_required = serializers.BooleanField(required=True, error_messages=ErrMessage.boolean(_("Is this field required")))
+    type = serializers.CharField(required=True, error_messages=ErrMessage.char(_("type")), validators=[
+        validators.RegexValidator(regex=re.compile("^string|int|dict|array|float$"),
+                                  message=_("The field only supports string|int|dict|array|float"), code=500)
+    ])
+    source = serializers.CharField(required=True, error_messages=ErrMessage.char(_("source")), validators=[
+        validators.RegexValidator(regex=re.compile("^custom|reference$"),
+                                  message=_("The field only supports custom|reference"), code=500)
+    ])
+    value = ObjectField(required=True, error_messages=ErrMessage.char(_("Variable Value")), model_type_list=[str, list])
+
+    def is_valid(self, *, raise_exception=False):
+        super().is_valid(raise_exception=True)
+        is_required = self.data.get('is_required')
+        if is_required and self.data.get('value') is None:
+            message = lazy_format(_('{field}, this field is required.'), field=self.data.get("name"))
+            raise AppApiException(500, message)
+
+
+class FunctionNodeParamsSerializer(serializers.Serializer):
+    input_field_list = InputField(required=True, many=True)
+    code = serializers.CharField(required=True, error_messages=ErrMessage.char(_("function")))
+    is_result = serializers.BooleanField(required=False, error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    def is_valid(self, *, raise_exception=False):
+        super().is_valid(raise_exception=True)
+
+
+class IFunctionNode(INode):
+    type = 'function-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return FunctionNodeParamsSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, input_field_list, code, **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/function_node/impl/__init__.py

@@ -0,0 +1,9 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： __init__.py.py
+    @date：2024/8/13 11:19
+    @desc:
+"""
+from .base_function_node import BaseFunctionNodeNode

apps/application/flow/step_node/function_node/impl/base_function_node.py

@@ -0,0 +1,108 @@
+# coding=utf-8
+"""
+    @project: MaxKB
+    @Author：虎
+    @file： base_function_lib_node.py
+    @date：2024/8/8 17:49
+    @desc:
+"""
+import json
+import time
+
+from typing import Dict
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.function_node.i_function_node import IFunctionNode
+from common.exception.app_exception import AppApiException
+from common.util.function_code import FunctionExecutor
+from smartdoc.const import CONFIG
+
+function_executor = FunctionExecutor(CONFIG.get('SANDBOX'))
+
+
+def write_context(step_variable: Dict, global_variable: Dict, node, workflow):
+    if step_variable is not None:
+        for key in step_variable:
+            node.context[key] = step_variable[key]
+        if workflow.is_result(node, NodeResult(step_variable, global_variable)) and 'result' in step_variable:
+            result = str(step_variable['result']) + '\n'
+            yield result
+            node.answer_text = result
+    node.context['run_time'] = time.time() - node.context['start_time']
+
+
+def valid_reference_value(_type, value, name):
+    if _type == 'int':
+        instance_type = int | float
+    elif _type == 'float':
+        instance_type = float | int
+    elif _type == 'dict':
+        instance_type = dict
+    elif _type == 'array':
+        instance_type = list
+    elif _type == 'string':
+        instance_type = str
+    else:
+        raise Exception(500, f'字段:{name}类型:{_type} 不支持的类型')
+    if not isinstance(value, instance_type):
+        raise Exception(f'字段:{name}类型:{_type}值:{value}类型错误')
+
+
+def convert_value(name: str, value, _type, is_required, source, node):
+    if not is_required and (value is None or (isinstance(value, str) and len(value) == 0)):
+        return None
+    if source == 'reference':
+        value = node.workflow_manage.get_reference_field(
+            value[0],
+            value[1:])
+        valid_reference_value(_type, value, name)
+        if _type == 'int':
+            return int(value)
+        if _type == 'float':
+            return float(value)
+        return value
+    try:
+        if _type == 'int':
+            return int(value)
+        if _type == 'float':
+            return float(value)
+        if _type == 'dict':
+            v = json.loads(value)
+            if isinstance(v, dict):
+                return v
+            raise Exception("类型错误")
+        if _type == 'array':
+            v = json.loads(value)
+            if isinstance(v, list):
+                return v
+            raise Exception("类型错误")
+        return value
+    except Exception as e:
+        raise Exception(f'字段:{name}类型:{_type}值:{value}类型错误')
+
+
+class BaseFunctionNodeNode(IFunctionNode):
+    def save_context(self, details, workflow_manage):
+        self.context['result'] = details.get('result')
+        if self.node_params.get('is_result', False):
+            self.answer_text = str(details.get('result'))
+
+    def execute(self, input_field_list, code, **kwargs) -> NodeResult:
+        params = {field.get('name'): convert_value(field.get('name'), field.get('value'), field.get('type'),
+                                                   field.get('is_required'), field.get('source'), self)
+                  for field in input_field_list}
+        result = function_executor.exec_code(code, params)
+        self.context['params'] = params
+        return NodeResult({'result': result}, {}, _write_context=write_context)
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            "result": self.context.get('result'),
+            "params": self.context.get('params'),
+            'run_time': self.context.get('run_time'),
+            'type': self.node.type,
+            'status': self.status,
+            'err_message': self.err_message
+        }

apps/application/flow/step_node/image_generate_step_node/__init__.py

@@ -0,0 +1,3 @@
+# coding=utf-8
+
+from .impl import *

apps/application/flow/step_node/image_generate_step_node/i_image_generate_node.py

@@ -0,0 +1,45 @@
+# coding=utf-8
+
+from typing import Type
+
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+from django.utils.translation import gettext_lazy as _
+
+
+class ImageGenerateNodeSerializer(serializers.Serializer):
+    model_id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Model id")))
+
+    prompt = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Prompt word (positive)")))
+
+    negative_prompt = serializers.CharField(required=False, error_messages=ErrMessage.char(_("Prompt word (negative)")),
+                                            allow_null=True, allow_blank=True, )
+    # 多轮对话数量
+    dialogue_number = serializers.IntegerField(required=False, default=0,
+                                               error_messages=ErrMessage.integer(_("Number of multi-round conversations")))
+
+    dialogue_type = serializers.CharField(required=False, default='NODE',
+                                          error_messages=ErrMessage.char(_("Conversation storage type")))
+
+    is_result = serializers.BooleanField(required=False, error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    model_params_setting = serializers.JSONField(required=False, default=dict,
+                                                 error_messages=ErrMessage.json(_("Model parameter settings")))
+
+
+class IImageGenerateNode(INode):
+    type = 'image-generate-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ImageGenerateNodeSerializer
+
+    def _run(self):
+        return self.execute(**self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, model_id, prompt, negative_prompt, dialogue_number, dialogue_type, history_chat_record, chat_id,
+                model_params_setting,
+                chat_record_id,
+                **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/image_generate_step_node/impl/__init__.py

@@ -0,0 +1,3 @@
+# coding=utf-8
+
+from .base_image_generate_node import BaseImageGenerateNode

apps/application/flow/step_node/image_generate_step_node/impl/base_image_generate_node.py

@@ -0,0 +1,122 @@
+# coding=utf-8
+from functools import reduce
+from typing import List
+
+import requests
+from langchain_core.messages import BaseMessage, HumanMessage, AIMessage
+
+from application.flow.i_step_node import NodeResult
+from application.flow.step_node.image_generate_step_node.i_image_generate_node import IImageGenerateNode
+from common.util.common import bytes_to_uploaded_file
+from dataset.serializers.file_serializers import FileSerializer
+from setting.models_provider.tools import get_model_instance_by_model_user_id
+
+
+class BaseImageGenerateNode(IImageGenerateNode):
+    def save_context(self, details, workflow_manage):
+        self.context['answer'] = details.get('answer')
+        self.context['question'] = details.get('question')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('answer')
+
+    def execute(self, model_id, prompt, negative_prompt, dialogue_number, dialogue_type, history_chat_record, chat_id,
+                model_params_setting,
+                chat_record_id,
+                **kwargs) -> NodeResult:
+        print(model_params_setting)
+        application = self.workflow_manage.work_flow_post_handler.chat_info.application
+        tti_model = get_model_instance_by_model_user_id(model_id, self.flow_params_serializer.data.get('user_id'),
+                                                        **model_params_setting)
+        history_message = self.get_history_message(history_chat_record, dialogue_number)
+        self.context['history_message'] = history_message
+        question = self.generate_prompt_question(prompt)
+        self.context['question'] = question
+        message_list = self.generate_message_list(question, history_message)
+        self.context['message_list'] = message_list
+        self.context['dialogue_type'] = dialogue_type
+        print(message_list)
+        image_urls = tti_model.generate_image(question, negative_prompt)
+        # 保存图片
+        file_urls = []
+        for image_url in image_urls:
+            file_name = 'generated_image.png'
+            file = bytes_to_uploaded_file(requests.get(image_url).content, file_name)
+            meta = {
+                'debug': False if application.id else True,
+                'chat_id': chat_id,
+                'application_id': str(application.id) if application.id else None,
+            }
+            file_url = FileSerializer(data={'file': file, 'meta': meta}).upload()
+            file_urls.append(file_url)
+        self.context['image_list'] = [{'file_id': path.split('/')[-1], 'url': path} for path in file_urls]
+        answer = ' '.join([f"![Image]({path})" for path in file_urls])
+        return NodeResult({'answer': answer, 'chat_model': tti_model, 'message_list': message_list,
+                           'image': [{'file_id': path.split('/')[-1], 'url': path} for path in file_urls],
+                           'history_message': history_message, 'question': question}, {})
+
+    def generate_history_ai_message(self, chat_record):
+        for val in chat_record.details.values():
+            if self.node.id == val['node_id'] and 'image_list' in val:
+                if val['dialogue_type'] == 'WORKFLOW':
+                    return chat_record.get_ai_message()
+                image_list = val['image_list']
+                return AIMessage(content=[
+                    *[{'type': 'image_url', 'image_url': {'url': f'{file_url}'}} for file_url in image_list]
+                ])
+        return chat_record.get_ai_message()
+
+    def get_history_message(self, history_chat_record, dialogue_number):
+        start_index = len(history_chat_record) - dialogue_number
+        history_message = reduce(lambda x, y: [*x, *y], [
+            [self.generate_history_human_message(history_chat_record[index]),
+             self.generate_history_ai_message(history_chat_record[index])]
+            for index in
+            range(start_index if start_index > 0 else 0, len(history_chat_record))], [])
+        return history_message
+
+    def generate_history_human_message(self, chat_record):
+
+        for data in chat_record.details.values():
+            if self.node.id == data['node_id'] and 'image_list' in data:
+                image_list = data['image_list']
+                if len(image_list) == 0 or data['dialogue_type'] == 'WORKFLOW':
+                    return HumanMessage(content=chat_record.problem_text)
+                return HumanMessage(content=data['question'])
+        return HumanMessage(content=chat_record.problem_text)
+
+    def generate_prompt_question(self, prompt):
+        return self.workflow_manage.generate_prompt(prompt)
+
+    def generate_message_list(self, question: str, history_message):
+        return [
+            *history_message,
+            question
+        ]
+
+    @staticmethod
+    def reset_message_list(message_list: List[BaseMessage], answer_text):
+        result = [{'role': 'user' if isinstance(message, HumanMessage) else 'ai', 'content': message.content} for
+                  message
+                  in
+                  message_list]
+        result.append({'role': 'ai', 'content': answer_text})
+        return result
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'history_message': [{'content': message.content, 'role': message.type} for message in
+                                (self.context.get('history_message') if self.context.get(
+                                    'history_message') is not None else [])],
+            'question': self.context.get('question'),
+            'answer': self.context.get('answer'),
+            'type': self.node.type,
+            'message_tokens': self.context.get('message_tokens'),
+            'answer_tokens': self.context.get('answer_tokens'),
+            'status': self.status,
+            'err_message': self.err_message,
+            'image_list': self.context.get('image_list'),
+            'dialogue_type': self.context.get('dialogue_type')
+        }

apps/application/flow/step_node/image_understand_step_node/__init__.py

@@ -0,0 +1,3 @@
+# coding=utf-8
+
+from .impl import *

apps/application/flow/step_node/image_understand_step_node/i_image_understand_node.py

@@ -0,0 +1,46 @@
+# coding=utf-8
+
+from typing import Type
+
+from rest_framework import serializers
+
+from application.flow.i_step_node import INode, NodeResult
+from common.util.field_message import ErrMessage
+from django.utils.translation import gettext_lazy as _
+
+
+class ImageUnderstandNodeSerializer(serializers.Serializer):
+    model_id = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Model id")))
+    system = serializers.CharField(required=False, allow_blank=True, allow_null=True,
+                                   error_messages=ErrMessage.char(_("Role Setting")))
+    prompt = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Prompt word")))
+    # 多轮对话数量
+    dialogue_number = serializers.IntegerField(required=True, error_messages=ErrMessage.integer(_("Number of multi-round conversations")))
+
+    dialogue_type = serializers.CharField(required=True, error_messages=ErrMessage.char(_("Conversation storage type")))
+
+    is_result = serializers.BooleanField(required=False, error_messages=ErrMessage.boolean(_('Whether to return content')))
+
+    image_list = serializers.ListField(required=False, error_messages=ErrMessage.list(_("picture")))
+
+    model_params_setting = serializers.JSONField(required=False, default=dict,
+                                                 error_messages=ErrMessage.json(_("Model parameter settings")))
+
+
+class IImageUnderstandNode(INode):
+    type = 'image-understand-node'
+
+    def get_node_params_serializer_class(self) -> Type[serializers.Serializer]:
+        return ImageUnderstandNodeSerializer
+
+    def _run(self):
+        res = self.workflow_manage.get_reference_field(self.node_params_serializer.data.get('image_list')[0],
+                                                       self.node_params_serializer.data.get('image_list')[1:])
+        return self.execute(image=res, **self.node_params_serializer.data, **self.flow_params_serializer.data)
+
+    def execute(self, model_id, system, prompt, dialogue_number, dialogue_type, history_chat_record, stream, chat_id,
+                model_params_setting,
+                chat_record_id,
+                image,
+                **kwargs) -> NodeResult:
+        pass

apps/application/flow/step_node/image_understand_step_node/impl/__init__.py

@@ -0,0 +1,3 @@
+# coding=utf-8
+
+from .base_image_understand_node import BaseImageUnderstandNode

apps/application/flow/step_node/image_understand_step_node/impl/base_image_understand_node.py

@@ -0,0 +1,224 @@
+# coding=utf-8
+import base64
+import os
+import time
+from functools import reduce
+from typing import List, Dict
+
+from django.db.models import QuerySet
+from langchain_core.messages import BaseMessage, HumanMessage, SystemMessage, AIMessage
+
+from application.flow.i_step_node import NodeResult, INode
+from application.flow.step_node.image_understand_step_node.i_image_understand_node import IImageUnderstandNode
+from dataset.models import File
+from setting.models_provider.tools import get_model_instance_by_model_user_id
+from imghdr import what
+
+
+def _write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow, answer: str):
+    chat_model = node_variable.get('chat_model')
+    message_tokens = node_variable['usage_metadata']['output_tokens'] if 'usage_metadata' in node_variable else 0
+    answer_tokens = chat_model.get_num_tokens(answer)
+    node.context['message_tokens'] = message_tokens
+    node.context['answer_tokens'] = answer_tokens
+    node.context['answer'] = answer
+    node.context['history_message'] = node_variable['history_message']
+    node.context['question'] = node_variable['question']
+    node.context['run_time'] = time.time() - node.context['start_time']
+    if workflow.is_result(node, NodeResult(node_variable, workflow_variable)):
+        node.answer_text = answer
+
+
+def write_context_stream(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据 (流式)
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result')
+    answer = ''
+    for chunk in response:
+        answer += chunk.content
+        yield chunk.content
+    _write_context(node_variable, workflow_variable, node, workflow, answer)
+
+
+def write_context(node_variable: Dict, workflow_variable: Dict, node: INode, workflow):
+    """
+    写入上下文数据
+    @param node_variable:      节点数据
+    @param workflow_variable:  全局数据
+    @param node:               节点实例对象
+    @param workflow:           工作流管理器
+    """
+    response = node_variable.get('result')
+    answer = response.content
+    _write_context(node_variable, workflow_variable, node, workflow, answer)
+
+
+def file_id_to_base64(file_id: str):
+    file = QuerySet(File).filter(id=file_id).first()
+    file_bytes = file.get_byte()
+    base64_image = base64.b64encode(file_bytes).decode("utf-8")
+    return [base64_image, what(None, file_bytes.tobytes())]
+
+
+class BaseImageUnderstandNode(IImageUnderstandNode):
+    def save_context(self, details, workflow_manage):
+        self.context['answer'] = details.get('answer')
+        self.context['question'] = details.get('question')
+        if self.node_params.get('is_result', False):
+            self.answer_text = details.get('answer')
+
+    def execute(self, model_id, system, prompt, dialogue_number, dialogue_type, history_chat_record, stream, chat_id,
+                model_params_setting,
+                chat_record_id,
+                image,
+                **kwargs) -> NodeResult:
+        # 处理不正确的参数
+        if image is None or not isinstance(image, list):
+            image = []
+        print(model_params_setting)
+        image_model = get_model_instance_by_model_user_id(model_id, self.flow_params_serializer.data.get('user_id'), **model_params_setting)
+        # 执行详情中的历史消息不需要图片内容
+        history_message = self.get_history_message_for_details(history_chat_record, dialogue_number)
+        self.context['history_message'] = history_message
+        question = self.generate_prompt_question(prompt)
+        self.context['question'] = question.content
+        # 生成消息列表, 真实的history_message
+        message_list = self.generate_message_list(image_model, system, prompt,
+                                                  self.get_history_message(history_chat_record, dialogue_number), image)
+        self.context['message_list'] = message_list
+        self.context['image_list'] = image
+        self.context['dialogue_type'] = dialogue_type
+        if stream:
+            r = image_model.stream(message_list)
+            return NodeResult({'result': r, 'chat_model': image_model, 'message_list': message_list,
+                               'history_message': history_message, 'question': question.content}, {},
+                              _write_context=write_context_stream)
+        else:
+            r = image_model.invoke(message_list)
+            return NodeResult({'result': r, 'chat_model': image_model, 'message_list': message_list,
+                               'history_message': history_message, 'question': question.content}, {},
+                              _write_context=write_context)
+
+    def get_history_message_for_details(self, history_chat_record, dialogue_number):
+        start_index = len(history_chat_record) - dialogue_number
+        history_message = reduce(lambda x, y: [*x, *y], [
+            [self.generate_history_human_message_for_details(history_chat_record[index]),
+             self.generate_history_ai_message(history_chat_record[index])]
+            for index in
+            range(start_index if start_index > 0 else 0, len(history_chat_record))], [])
+        return history_message
+
+    def generate_history_ai_message(self, chat_record):
+        for val in chat_record.details.values():
+            if self.node.id == val['node_id'] and 'image_list' in val:
+                if val['dialogue_type'] == 'WORKFLOW':
+                    return chat_record.get_ai_message()
+                return AIMessage(content=val['answer'])
+        return chat_record.get_ai_message()
+
+    def generate_history_human_message_for_details(self, chat_record):
+        for data in chat_record.details.values():
+            if self.node.id == data['node_id'] and 'image_list' in data:
+                image_list = data['image_list']
+                if len(image_list) == 0 or data['dialogue_type'] == 'WORKFLOW':
+                    return HumanMessage(content=chat_record.problem_text)
+                file_id_list = [image.get('file_id') for image in image_list]
+                return HumanMessage(content=[
+                    {'type': 'text', 'text': data['question']},
+                    *[{'type': 'image_url', 'image_url': {'url': f'/api/file/{file_id}'}} for file_id in file_id_list]
+
+                ])
+        return HumanMessage(content=chat_record.problem_text)
+
+    def get_history_message(self, history_chat_record, dialogue_number):
+        start_index = len(history_chat_record) - dialogue_number
+        history_message = reduce(lambda x, y: [*x, *y], [
+            [self.generate_history_human_message(history_chat_record[index]),
+             self.generate_history_ai_message(history_chat_record[index])]
+            for index in
+            range(start_index if start_index > 0 else 0, len(history_chat_record))], [])
+        return history_message
+
+    def generate_history_human_message(self, chat_record):
+
+        for data in chat_record.details.values():
+            if self.node.id == data['node_id'] and 'image_list' in data:
+                image_list = data['image_list']
+                if len(image_list) == 0 or data['dialogue_type'] == 'WORKFLOW':
+                    return HumanMessage(content=chat_record.problem_text)
+                image_base64_list = [file_id_to_base64(image.get('file_id')) for image in image_list]
+                return HumanMessage(
+                    content=[
+                        {'type': 'text', 'text': data['question']},
+                        *[{'type': 'image_url', 'image_url': {'url': f'data:image/{base64_image[1]};base64,{base64_image[0]}'}} for
+                          base64_image in image_base64_list]
+                    ])
+        return HumanMessage(content=chat_record.problem_text)
+
+    def generate_prompt_question(self, prompt):
+        return HumanMessage(self.workflow_manage.generate_prompt(prompt))
+
+    def generate_message_list(self, image_model, system: str, prompt: str, history_message, image):
+        if image is not None and len(image) > 0:
+            # 处理多张图片
+            images = []
+            for img in image:
+                file_id = img['file_id']
+                file = QuerySet(File).filter(id=file_id).first()
+                image_bytes = file.get_byte()
+                base64_image = base64.b64encode(image_bytes).decode("utf-8")
+                image_format = what(None, image_bytes.tobytes())
+                images.append({'type': 'image_url', 'image_url': {'url': f'data:image/{image_format};base64,{base64_image}'}})
+            messages = [HumanMessage(
+                content=[
+                    {'type': 'text', 'text': self.workflow_manage.generate_prompt(prompt)},
+                    *images
+                ])]
+        else:
+            messages = [HumanMessage(self.workflow_manage.generate_prompt(prompt))]
+
+        if system is not None and len(system) > 0:
+            return [
+                SystemMessage(self.workflow_manage.generate_prompt(system)),
+                *history_message,
+                *messages
+            ]
+        else:
+            return [
+                *history_message,
+                *messages
+            ]
+
+    @staticmethod
+    def reset_message_list(message_list: List[BaseMessage], answer_text):
+        result = [{'role': 'user' if isinstance(message, HumanMessage) else 'ai', 'content': message.content} for
+                  message
+                  in
+                  message_list]
+        result.append({'role': 'ai', 'content': answer_text})
+        return result
+
+    def get_details(self, index: int, **kwargs):
+        return {
+            'name': self.node.properties.get('stepName'),
+            "index": index,
+            'run_time': self.context.get('run_time'),
+            'system': self.node_params.get('system'),
+            'history_message': [{'content': message.content, 'role': message.type} for message in
+                                (self.context.get('history_message') if self.context.get(
+                                    'history_message') is not None else [])],
+            'question': self.context.get('question'),
+            'answer': self.context.get('answer'),
+            'type': self.node.type,
+            'message_tokens': self.context.get('message_tokens'),
+            'answer_tokens': self.context.get('answer_tokens'),
+            'status': self.status,
+            'err_message': self.err_message,
+            'image_list': self.context.get('image_list'),
+            'dialogue_type': self.context.get('dialogue_type')
+        }

apps/application/flow/step_node/mcp_node/__init__.py

@@ -0,0 +1,3 @@
+# coding=utf-8
+
+from .impl import *