feat: bump version to 0.3.31-fix2 (#1592)

Co-authored-by: jyong <jyong@dify.ai>

.devcontainer/Dockerfile

@@ -1,11 +1,8 @@
-FROM mcr.microsoft.com/devcontainers/anaconda:0-3
+FROM mcr.microsoft.com/devcontainers/python:3.10
+
+COPY . .
 
-# Copy environment.yml (if found) to a temp location so we update the environment. Also
-# copy "noop.txt" so the COPY instruction does not fail if no environment.yml exists.
-COPY environment.yml* .devcontainer/noop.txt /tmp/conda-tmp/
-RUN if [ -f "/tmp/conda-tmp/environment.yml" ]; then umask 0002 && /opt/conda/bin/conda env update -n base -f /tmp/conda-tmp/environment.yml; fi \
-    && rm -rf /tmp/conda-tmp
 
 # [Optional] Uncomment this section to install additional OS packages.
 # RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
-#     && apt-get -y install --no-install-recommends <your-package-list-here>
+#     && apt-get -y install --no-install-recommends <your-package-list-here>

.devcontainer/devcontainer.json

@@ -1,13 +1,12 @@
 // For format details, see https://aka.ms/devcontainer.json. For config options, see the
 // README at: https://github.com/devcontainers/templates/tree/main/src/anaconda
 {
-	"name": "Anaconda (Python 3)",
+	"name": "Python 3.10",
 	"build": { 
 		"context": "..",
 		"dockerfile": "Dockerfile"
 	},
 	"features": {
-		"ghcr.io/dhoeric/features/act:1": {},
 		"ghcr.io/devcontainers/features/node:1": {
 			"nodeGypDependencies": true,
 			"version": "lts"

.github/CODE_OF_CONDUCT.md

@@ -0,0 +1,43 @@
+# Dify 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, caste, color, 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
+
+## Language Policy
+
+To facilitate clear and effective communication, all discussions, comments, documentation, and pull requests in this project should be conducted in English. This ensures that all contributors can participate and collaborate effectively.
+
+

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -1,5 +1,5 @@
 name: "🕷️ Bug report"
-description: Report errors or unexpected behavior
+description: Report errors or unexpected behavior [please use English :）]
 labels:
 - bug
 body:

.github/ISSUE_TEMPLATE/document_issue.yml

@@ -1,5 +1,5 @@
 name: "📚 Documentation Issue"
-description: Report issues in our documentation
+description: Report issues in our documentation [please use English :）]
 labels: 
 - ducumentation
 body:

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -1,5 +1,5 @@
 name: "⭐ Feature or enhancement request"
-description: Propose something new.
+description: Propose something new. [please use English :）]
 labels:
 - enhancement
 body:

.github/ISSUE_TEMPLATE/help_wanted.yml

@@ -1,5 +1,5 @@
 name: "🤝 Help Wanted"
-description: "Request help from the community"
+description: "Request help from the community" [please use English :）]
 labels:
 - help-wanted
 body:

.github/ISSUE_TEMPLATE/translation_issue.yml

@@ -1,5 +1,5 @@
 name: "🌐 Localization/Translation issue"
-description: Report incorrect translations.
+description: Report incorrect translations. [please use English :）]
 labels:
 - translation
 body:

.github/workflows/build-api-image.yml

@@ -31,12 +31,10 @@ jobs:
       with:
         images: langgenius/dify-api
         tags: |
-          type=raw,value=latest,enable={{is_default_branch}}
+          type=raw,value=latest,enable=${{ startsWith(github.ref, 'refs/tags/') }}
           type=ref,event=branch
           type=sha,enable=true,priority=100,prefix=,suffix=,format=long
-          type=semver,pattern={{major}}.{{minor}}.{{patch}}
-          type=semver,pattern={{major}}.{{minor}}
-          type=semver,pattern={{major}}
+          type=raw,value=${{ github.ref_name }},enable=${{ startsWith(github.ref, 'refs/tags/') }}
 
     - name: Build and push
       uses: docker/build-push-action@v4

.github/workflows/build-web-image.yml

@@ -31,12 +31,10 @@ jobs:
       with:
         images: langgenius/dify-web
         tags: |
-          type=raw,value=latest,enable={{is_default_branch}}
+          type=raw,value=latest,enable=${{ startsWith(github.ref, 'refs/tags/') }}
           type=ref,event=branch
           type=sha,enable=true,priority=100,prefix=,suffix=,format=long
-          type=semver,pattern={{major}}.{{minor}}.{{patch}}
-          type=semver,pattern={{major}}.{{minor}}
-          type=semver,pattern={{major}}
+          type=raw,value=${{ github.ref_name }},enable=${{ startsWith(github.ref, 'refs/tags/') }}
 
     - name: Build and push
       uses: docker/build-push-action@v4

.github/workflows/check_no_chinese_comments.py

@@ -1,37 +0,0 @@
-import os
-import re
-from zhon.hanzi import punctuation
-
-def has_chinese_characters(text):
-    for char in text:
-        if '\u4e00' <= char <= '\u9fff' or char in punctuation:
-            return True
-    return False
-
-def check_file_for_chinese_comments(file_path):
-    with open(file_path, 'r', encoding='utf-8') as file:
-        for line_number, line in enumerate(file, start=1):
-            if has_chinese_characters(line):
-                print(f"Found Chinese characters in {file_path} on line {line_number}:")
-                print(line.strip())
-                return True
-    return False
-
-def main():
-    has_chinese = False
-    excluded_files = ["model_template.py", 'stopwords.py', 'commands.py',
-                      'indexing_runner.py', 'web_reader_tool.py', 'spark_provider.py',
-                      'prompts.py']
-
-    for root, _, files in os.walk("."):
-        for file in files:
-            if file.endswith(".py") and file not in excluded_files:
-                file_path = os.path.join(root, file)
-                if check_file_for_chinese_comments(file_path):
-                    has_chinese = True
-
-    if has_chinese:
-        raise Exception("Found Chinese characters in Python files. Please remove them.")
-
-if __name__ == "__main__":
-    main()

.github/workflows/check_no_chinese_comments.yml

@@ -1,31 +0,0 @@
-name: Check for Chinese comments
-
-on:
-  push:
-    branches:
-      - 'main'
-  pull_request:
-    branches:
-      - main
-
-jobs:
-  check-chinese-comments:
-    runs-on: ubuntu-latest
-
-    steps:
-    - name: Check out repository
-      uses: actions/checkout@v2
-
-    - name: Set up Python
-      uses: actions/setup-python@v2
-      with:
-        python-version: 3.9
-
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        pip install zhon
-
-    - name: Run script to check for Chinese comments
-      run: |
-        python .github/workflows/check_no_chinese_comments.py

.github/workflows/stale.yml

@@ -20,11 +20,11 @@ jobs:
     steps:
     - uses: actions/stale@v5
       with:
-        days-before-issue-stale: 30
+        days-before-issue-stale: 15
         days-before-issue-close: 3
         repo-token: ${{ secrets.GITHUB_TOKEN }}
         stale-issue-message: "Close due to it's no longer active, if you have any questions, you can reopen it."
         stale-pr-message: "Close due to it's no longer active, if you have any questions, you can reopen it."
         stale-issue-label: 'no-issue-activity'
         stale-pr-label: 'no-pr-activity'
-        any-of-labels: 'duplicate,question,invalid,wontfix,no-issue-activity,no-pr-activity,enhancement'
+        any-of-labels: 'duplicate,question,invalid,wontfix,no-issue-activity,no-pr-activity,enhancement,cant-reproduce,help-wanted'

.gitignore

@@ -144,6 +144,7 @@ docker/volumes/app/storage/*
 docker/volumes/db/data/*
 docker/volumes/redis/data/*
 docker/volumes/weaviate/*
+docker/volumes/qdrant/*
 
 sdks/python-client/build
 sdks/python-client/dist

README.md

@@ -1,66 +1,67 @@
-![](./images/describe-en.png)
+[![](./images/describe.png)](https://dify.ai)
 <p align="center">
   <a href="./README.md">English</a> |
   <a href="./README_CN.md">简体中文</a> |
   <a href="./README_JA.md">日本語</a> |
-  <a href="./README_ES.md">Español</a>
+  <a href="./README_ES.md">Español</a> |
+  <a href="./README_KL.md">Klingon</a>
 </p>
 
-#### [Website](https://dify.ai) • [Docs](https://docs.dify.ai) • [Deployment Docs](https://docs.dify.ai/getting-started/install-self-hosted) •  [FAQ](https://docs.dify.ai/getting-started/faq) • [Twitter](https://twitter.com/dify_ai) • [Discord](https://discord.gg/FngNHpbcY7)
+<p align="center">
+    <a href="https://dify.ai" target="_blank">
+        <img alt="Static Badge" src="https://img.shields.io/badge/AI-Dify?logo=AI&logoColor=%20%23f5f5f5&label=Dify&labelColor=%20%23155EEF&color=%23EAECF0"></a>
+    <a href="https://discord.gg/FngNHpbcY7" target="_blank">
+        <img src="https://img.shields.io/discord/1082486657678311454?logo=discord"
+            alt="chat on Discord"></a>
+    <a href="https://twitter.com/intent/follow?screen_name=dify_ai" target="_blank">
+        <img src="https://img.shields.io/twitter/follow/dify_ai?style=social&logo=X"
+            alt="follow on Twitter"></a>
+    <a href="https://hub.docker.com/u/langgenius" target="_blank">
+        <img alt="Docker Pulls" src="https://img.shields.io/docker/pulls/langgenius/dify-web"></a>
+</p>
 
-**Dify** is an easy-to-use LLMOps platform designed to empower more people to create sustainable, AI-native applications. With visual orchestration for various application types, Dify offers out-of-the-box, ready-to-use applications that can also serve as Backend-as-a-Service APIs. Unify your development process with one API for plugins and datasets integration, and streamline your operations using a single interface for prompt engineering, visual analytics, and continuous improvement.
+**Dify** is an LLM application development platform that has already seen over **100,000** applications built on Dify.AI. It integrates the concepts of Backend as a Service and LLMOps, covering the core tech stack required for building generative AI-native applications, including a built-in RAG engine. With Dify, **you can self-deploy capabilities similar to Assistants API and GPTs based on any LLMs.**
 
-Applications created with Dify include:
-
-Out-of-the-box web sites supporting form mode and chat conversation mode
-A single API encompassing plugin capabilities, context enhancement, and more, saving you backend coding effort
-Visual data analysis, log review, and annotation for applications
-
-
-https://github.com/langgenius/dify/assets/100913391/f6e658d5-31b3-4c16-a0af-9e191da4d0f6
-
-
-## Highlighted Features
-**1. LLMs support:** Choose capabilities based on different models when building your Dify AI apps. Dify is compatible with Langchain, meaning it will support various LLMs. Currently supported:
-
-- [x] **OpenAI**: GPT4, GPT3.5-turbo, GPT3.5-turbo-16k, text-davinci-003 
-- [x] **Azure OpenAI Service**
-- [x] **Anthropic**: Claude2, Claude-instant
-- [x] **Replicate**
-- [x] **Hugging Face Hub**
-- [x] **ChatGLM**
-- [x] **Llama2**
-- [x] **MiniMax**
-- [x] **Spark**
-- [x] **Wenxin**
-- [x] **Tongyi**
-
-
-We provide the following free resources for registered Dify cloud users (sign up at [dify.ai](https://dify.ai)):
-* 600,000 free Claude model tokens to build Claude-powered apps
-* 200 free OpenAI queries to build OpenAI-based apps
-
-  
-**2. Visual orchestration:** Build an AI app in minutes by writing and debugging prompts visually.
-
-**3. Text embedding:** Fully automated text preprocessing embeds your data as context without complex concepts. Supports PDF, TXT, and syncing data from Notion, webpages, APIs.
-
-**4. API-based:**  Backend-as-a-service. Access web apps directly or integrate via APIs without complex backend setup.
-
-**5. Plugins:** Dify "Smart Chat" now supports first-party plugins like web browsing, Google search, Wikipedia to enable online lookup, analyzing web content, and explaining the AI's reasoning process conversationally.
-
-**6. Team workspaces:** Team members can join workspaces to collaboratively edit, manage, and use team AI apps.
-
-**7. Data labeling and improvement:**  Visually inspect AI app logs and improve data via labeling. Observe the AI's reasoning process to continuously enhance performance. (Coming soon)
-
-## Use cases
-* [Create an AI ChatBot with Business Data in Minutes.](https://docs.dify.ai/use-cases/create-an-ai-chatbot-with-business-data-in-minutes)
-* [How to Build an Notion AI Assistant Based on Your Own Notes?](https://docs.dify.ai/use-cases/build-an-notion-ai-assistant)
-* [Create a Midjoureny Prompt Bot Without Code in Just a Few Minutes.](https://docs.dify.ai/use-cases/create-a-midjoureny-prompt-bot-with-dify)
+![](./images/demo.png)
 
 ## Use Cloud Services
 
-Visit [Dify.ai](https://dify.ai)
+Using [Dify.AI Cloud](https://dify.ai) provides all the capabilities of the open-source version, and includes a complimentary 200 GPT trial credits.
+
+## Why Dify
+
+Dify features model neutrality and is a complete, engineered tech stack compared to hardcoded development libraries like LangChain. Unlike OpenAI's Assistants API, Dify allows for full local deployment of services.
+
+| Feature | Dify.AI | Assistants API | LangChain |
+|---------|---------|----------------|-----------|
+| **Programming Approach** | API-oriented | API-oriented | Python Code-oriented |
+| **Ecosystem Strategy** | Open Source | Closed and Commercial | Open Source |
+| **RAG Engine** | Supported | Supported | Not Supported |
+| **Prompt IDE** | Included | Included | None |
+| **Supported LLMs** | Rich Variety | Only GPT | Rich Variety |
+| **Local Deployment** | Supported | Not Supported | Not Applicable |
+
+## Features
+
+![](./images/models.png)
+
+**1. LLM Support**: Integration with OpenAI's GPT family of models, or the open-source Llama2 family models. In fact, Dify supports mainstream commercial models and open-source models (locally deployed or based on MaaS).
+
+**2. Prompt IDE**: Visual orchestration of applications and services based on LLMs with your team.
+
+**3. RAG Engine**: Includes various RAG capabilities based on full-text indexing or vector database embeddings, allowing direct upload of PDFs, TXTs, and other text formats.
+
+**4. Agents**: A Function Calling based Agent framework that allows users to configure what they see is what they get. Dify includes basic plugin capabilities like Google Search.
+
+**5. Continuous Operations**: Monitor and analyze application logs and performance, continuously improving Prompts, datasets, or models using production data.
+
+## Before You Start
+
+- [Website](https://dify.ai)
+- [Docs](https://docs.dify.ai)
+- [Deployment Docs](https://docs.dify.ai/getting-started/install-self-hosted)
+- [FAQ](https://docs.dify.ai/getting-started/faq) 
+
 
 ## Install the Community Edition
 
@@ -89,88 +90,28 @@ You can go to https://github.com/BorisPolonsky/dify-helm for deployment informat
 
 ### Configuration
 
-If you need to customize the configuration, please refer to the comments in our [docker-compose.yml](docker/docker-compose.yaml) file and manually set the environment configuration. After making the changes, please run 'docker-compose up -d' again.
+If you need to customize the configuration, please refer to the comments in our [docker-compose.yml](docker/docker-compose.yaml) file and manually set the environment configuration. After making the changes, please run `docker-compose up -d` again. You can see the full list of environment variables in our [docs](https://docs.dify.ai/getting-started/install-self-hosted/environments).
 
-## Roadmap
-
-Features under development:
-
-- **Datasets**, supporting more datasets, e.g. syncing content from Notion or webpages
-We will support more datasets, including text, webpages, and even Notion content. Users can build AI applications based on their own data sources.
-- **Plugins**, introducing ChatGPT Plugin-standard plugins for applications, or using Dify-produced plugins
-We will release plugins complying with ChatGPT standard, or Dify's own plugins to enable more capabilities in applications. 
-
-
-## Q&A
-
-**Q: What can I do with Dify?**
-
-A: Dify is a simple yet powerful LLM development and operations tool. You can use it to build commercial-grade applications, personal assistants. If you want to develop your own applications, LangDifyGenius can save you backend work in integrating with OpenAI and offer visual operations capabilities, allowing you to continuously improve and train your GPT model.
-
-**Q: How do I use Dify to "train" my own model?**
-
-A: A valuable application consists of Prompt Engineering, context enhancement, and Fine-tuning. We've created a hybrid programming approach combining Prompts with programming languages (similar to a template engine), making it easy to accomplish long-text embedding or capturing subtitles from a user-input Youtube video - all of which will be submitted as context for LLMs to process. We place great emphasis on application operability, with data generated by users during App usage available for analysis, annotation, and continuous training. Without the right tools, these steps can be time-consuming.
-
-**Q: What do I need to prepare if I want to create my own application?**
-
-A: We assume you already have an OpenAI API Key; if not, please register for one. If you already have some content that can serve as training context, that's great!
-
-**Q: What interface languages are available?**
-
-A: English and Chinese are currently supported, and you can contribute language packs to us.
 
 ## Star History
 
 [![Star History Chart](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
 
 
-## Contributing 
+## Community & Support
 
-We welcome you to contribute to Dify to help make Dify better. We welcome contributions in various ways, submitting code, issues, new ideas, or sharing the interesting and useful AI applications you have created based on Dify. At the same time, we also welcome you to share Dify at different events, conferences, and social media.
+We welcome you to contribute to Dify to help make Dify better in various ways, submitting code, issues, new ideas, or sharing the interesting and useful AI applications you have created based on Dify. At the same time, we also welcome you to share Dify at different events, conferences, and social media.
 
-### Submit a Pull Request 
+- [GitHub Issues](https://github.com/langgenius/dify/issues). Best for: bugs and errors you encounter using Dify.AI, see the [Contribution Guide](CONTRIBUTING.md).
+- [Email Support](mailto:hello@dify.ai?subject=[GitHub]Questions%20About%20Dify). Best for: questions you have about using Dify.AI.
+- [Discord](https://discord.gg/FngNHpbcY7). Best for: sharing your applications and hanging out with the community.
+- [Twitter](https://twitter.com/dify_ai). Best for: sharing your applications and hanging out with the community.
+- [Business License](mailto:business@dify.ai?subject=[GitHub]Business%20License%20Inquiry). Best for: business inquiries of licensing Dify.AI for commercial use.
 
-To ensure proper review, all code contributions, including from contributors with direct commit access, must be submitted as PR requests and approved by core developers before merging branches. 
-We welcome PRs from everyone! If you're willing to help out, you can learn more about how to contribute code to the project in the [Contribution Guide](CONTRIBUTING.md).  
-
-### Submit issues or ideas  
-
-You can submit your issues or ideas by adding issues to the Dify repository. If you encounter issues, please describe the steps you took to encounter the issue as much as possible so we can better discover it. If you have any new ideas for our product, we also welcome your feedback. Please share your insights as much as possible so we can get more feedback and further discussion in the community.  
-
-### Share your applications
-
-We encourage all community members to share their AI applications built on Dify, which can be applied to different scenarios or different users. This will provide powerful inspiration for people who want to create AI capabilities! You can share your experience by [submitting an issue in the Dify-user-case repository](https://github.com/langgenius/dify-user-case/issues).  
-
-### Share Dify with others
-
-We encourage community contributors to actively demonstrate different aspects of using Dify. You can talk or share any feature of using Dify at  meetups and conferences, blogs or social media. We believe your unique sharing will be of great help to others!  Mention @Dify.AI on Twitter and/or communicate on [Discord](https://discord.gg/FngNHpbcY7) so we can give pointers and tips and help you spread the word by promoting your content on the different Dify communication channels.
-
-### Help others 
-You can also help people in need of help on Discord, GitHub issues or other social platforms, guide others to solve problems encountered during use and share usage experiences. This is also a great contribution! If you want to become a maintainer of the Dify community, please contact the official team via [Discord](https://discord.gg/FngNHpbcY7) or email us at support@dify.ai. 
-
-
-## Contact Us
-
-If you have any questions, suggestions, or partnership inquiries, feel free to contact us through the following channels:
-
-- Submit an Issue or PR on our GitHub Repo
-- Join the discussion in our [Discord](https://discord.gg/FngNHpbcY7) Community
-- Send an email to hello@dify.ai
-
-We're eager to assist you and together create more fun and useful AI applications!
-
-## Security
+## Security Disclosure
 
 To protect your privacy, please avoid posting security issues on GitHub. Instead, send your questions to security@dify.ai and we will provide you with a more detailed answer.
 
-## Citation
-
-This software uses the following open-source software:
-
-- Chase, H. (2022). LangChain [Computer software]. https://github.com/hwchase17/langchain
-
-For more information, please refer to the official website or license text of the respective software.
-
 ## License
 
-This repository is available under the [Dify Open Source License](LICENSE).
+This repository is available under the [Dify Open Source License](LICENSE), which is essentially Apache 2.0 with a few additional restrictions.

README_CN.md

@@ -1,61 +1,63 @@
-![](./images/describe-cn.jpg)
+[![](./images/describe.png)](https://dify.ai)
 <p align="center">
   <a href="./README.md">English</a> |
   <a href="./README_CN.md">简体中文</a> |
   <a href="./README_JA.md">日本語</a> |
-  <a href="./README_ES.md">Español</a>
+  <a href="./README_ES.md">Español</a> |
+  <a href="./README_KL.md">Klingon</a>
 </p>
 
+<p align="center">
+    <a href="https://dify.ai" target="_blank">
+        <img alt="Static Badge" src="https://img.shields.io/badge/AI-Dify?logo=AI&logoColor=%20%23f5f5f5&label=Dify&labelColor=%20%23155EEF&color=%23EAECF0"></a>
+    <a href="https://discord.gg/FngNHpbcY7" target="_blank">
+        <img src="https://img.shields.io/discord/1082486657678311454?logo=discord"
+            alt="chat on Discord"></a>
+    <a href="https://twitter.com/intent/follow?screen_name=dify_ai" target="_blank">
+        <img src="https://img.shields.io/twitter/follow/dify_ai?style=social&logo=X"
+            alt="follow on Twitter"></a>
+    <a href="https://hub.docker.com/u/langgenius" target="_blank">
+        <img alt="Docker Pulls" src="https://img.shields.io/docker/pulls/langgenius/dify-web"></a>
+</p>
 
-####  [官方网站](https://dify.ai) •  [使用文档](https://docs.dify.ai/v/zh-hans)  · [部署文档](https://docs.dify.ai/v/zh-hans/getting-started/install-self-hosted)  ·  [FAQ](https://docs.dify.ai/v/zh-hans/getting-started/faq) • [Twitter](https://twitter.com/dify_ai) •  [Discord](https://discord.gg/FngNHpbcY7)
+Dify 是一个 LLM 应用开发平台，已经有超过 10 万个应用基于 Dify.AI 构建。它融合了 Backend as Service 和 LLMOps 的理念，涵盖了构建生成式 AI 原生应用所需的核心技术栈，包括一个内置 RAG 引擎。使用 Dify，你可以基于任何模型自部署类似 Assistants API 和 GPTs 的能力。
 
-**Dify** 是一个易用的 LLMOps 平台，基于不同的大型语言模型能力，让更多人可以简易地创建可持续运营的原生 AI 应用。Dify 提供多种类型应用的可视化编排，应用可开箱即用，也能以“后端即服务”的 API 提供服务。
+![](./images/demo.png)
 
-通过 Dify 创建的应用包含了：
+## 为什么选择 Dify
 
-- 开箱即用的的 Web 站点，支持表单模式和聊天对话模式
-- 一套 API 即可包含插件、上下文增强等能力，替你省下了后端代码的编写工作
-- 可视化的对应用进行数据分析，查阅日志或进行标注
+Dify 具有模型中立性，相较 LangChain 等硬编码开发库 Dify 是一个完整的、工程化的技术栈，而相较于 OpenAI 的 Assistants API 你可以完全将服务部署在本地。
 
-https://github.com/langgenius/dify/assets/100913391/f6e658d5-31b3-4c16-a0af-9e191da4d0f6
-
-## 核心能力
-1. **模型支持：** 你可以在 Dify 上选择基于不同模型的能力来开发你的 AI 应用。Dify 兼容 Langchain，这意味着我们将逐步支持多种 LLMs ，目前支持的模型供应商：
-
-- [x] **OpenAI**：GPT4、GPT3.5-turbo、GPT3.5-turbo-16k、text-davinci-003 
-- [x] **Azure OpenAI Service**
-- [x] **Anthropic**：Claude2、Claude-instant
-- [x] **Replicate**
-- [x] **Hugging Face Hub**
-- [x] **ChatGLM**
-- [x] **Llama2**
-- [x] **MiniMax**
-- [x] **讯飞星火大模型**
-- [x] **文心一言**
-- [x] **通义千问**
+| 功能 | Dify.AI | Assistants API | LangChain |
+| --- | --- | --- | --- |
+| 编程方式 | 面向 API | 面向 API | 面向 Python 代码 |
+| 生态策略 | 开源 | 封闭且商用 | 开源 |
+| RAG 引擎 | 支持 | 支持 | 不支持 |
+| Prompt IDE | 包含 | 包含 | 没有 |
+| 支持的 LLMs | 丰富 | 仅 GPT | 丰富 |
+| 本地部署 | 支持 | 不支持 | 不适用 |
 
 
-我们为所有注册云端版的用户免费提供以下资源（登录 [dify.ai](https://cloud.dify.ai) 即可使用）：
-* 60 万 Tokens Claude 模型的消息调用额度，用于创建基于 Claude 模型的 AI 应用
-* 200 次 OpenAI 模型的消息调用额度，用于创建基于 OpenAI 模型的 AI 应用
-* 300 万 讯飞星火大模型 Token 的调用额度，用于创建基于讯飞星火大模型的 AI 应用
-* 100 万 MiniMax Token 的调用额度，用于创建基于 MiniMax 模型的 AI 应用
-2. **可视化编排 Prompt：** 通过界面化编写 prompt 并调试，只需几分钟即可发布一个 AI 应用。
-3. **文本 Embedding 处理（数据集）**：全自动完成文本预处理，使用你的数据作为上下文，无需理解晦涩的概念和技术处理。支持 PDF、txt 等文件格式，支持从 Notion、网页、API 同步数据。
-4. **基于 API 开发：** 后端即服务。您可以直接访问网页应用，也可以接入 API 集成到您的应用中，无需关注复杂的后端架构和部署过程。
-5. **插件能力：** Dify 「智聊」平台已支持网页浏览、Google 搜索、Wikipedia 查询等第一方插件，可在对话中实现联网搜索、分析网页内容、展示 AI 的推理过程。
-6. **团队 Workspace：** 团队成员可加入 Workspace 编辑、管理和使用团队内的 AI 应用。
-6. **数据标注与改进：** 可视化查阅 AI 应用日志并对数据进行改进标注，观测 AI 的推理过程，不断提高其性能。（Coming soon）
- -----------------------------
- ## Use cases
- * [几分钟创建一个带有业务数据的官网 AI 智能客服](https://docs.dify.ai/v/zh-hans/use-cases/create-an-ai-chatbot-with-business-data-in-minutes)
- * [构建一个 Notion AI 助手](https://docs.dify.ai/v/zh-hans/use-cases/build-an-notion-ai-assistant)
- * [创建 Midjoureny 提示词机器人](https://docs.dify.ai/v/zh-hans/use-cases/create-a-midjoureny-prompt-word-robot-with-zero-code)
+## 特点
 
+![](./images/models.png)
 
-## 使用云服务
+**1. LLM支持**：与 OpenAI 的 GPT 系列模型集成,或者与开源的 Llama2 系列模型集成。事实上，Dify支持主流的商业模型和开源模型(本地部署或基于 MaaS)。
 
-访问 [Dify.ai](https://cloud.dify.ai) 使用云端版。
+**2. Prompt IDE**：和团队一起在 Dify 协作，通过可视化的 Prompt 和应用编排工具开发 AI 应用。 支持无缝切换多种大型语言模型。
+
+**3. RAG引擎**：包括各种基于全文索引或向量数据库嵌入的 RAG 能力，允许直接上传 PDF、TXT 等各种文本格式。
+
+**4. Agent**：基于函数调用的 Agent框架，允许用户自定义配置，所见即所得。Dify 提供了基本的插件能力，如谷歌搜索。
+
+**5. 持续运营**：监控和分析应用日志和性能，使用生产数据持续改进 Prompt、数据集或模型。
+
+## 在开始之前
+
+- [网站](https://dify.ai)
+- [文档](https://docs.dify.ai)
+- [部署文档](https://docs.dify.ai/getting-started/install-self-hosted)
+- [常见问题](https://docs.dify.ai/getting-started/faq)
 
 ## 安装社区版
 
@@ -84,80 +86,29 @@ docker compose up -d
 
 ### 配置
 
-需要自定义配置，请参考我们的 [docker-compose.yml](docker/docker-compose.yaml) 文件中的注释，并手动设置环境配置，修改完毕后，请再次执行 `docker-compose up -d`。
-
-## Roadmap
-
-我们正在开发中的功能：
-
-- **数据集**，支持更多的数据集，通过网页、API 同步内容。用户可以根据自己的数据源构建 AI 应用程序。
-- **插件**，我们将发布符合 ChatGPT 标准的插件，支持更多 Dify 自己的插件，支持用户自定义插件能力，以在应用程序中启用更多功能，例如以支持以目标为导向的分解推理任务。
-
-## Q&A
-
-**Q: 我能用 Dify 做什么？**
-
-A: Dify 是一个简单且能力丰富的 LLM 开发和运营工具。你可以用它搭建商用级应用，个人助理。如果你想自己开发应用，Dify 也能为你省下接入 OpenAI 的后端工作，使用我们逐步提供的可视化运营能力，你可以持续的改进和训练你的 GPT 模型。
-
-**Q: 如何使用 Dify “训练”自己的模型？**
-
-A: 一个有价值的应用由 Prompt Engineering、上下文增强和 Fine-tune 三个环节组成。我们创造了一种 Prompt 结合编程语言的 Hybrid 编程方式（类似一个模版引擎），你可以轻松的完成长文本嵌入，或抓取用户输入的一个 Youtube 视频的字幕——这些都将作为上下文提交给 LLMs 进行计算。我们十分注重应用的可运营性，你的用户在使用 App 期间产生的数据，可进行分析、标记和持续训练。以上环节如果没有好的工具支持，可能会消耗你大量的时间。
-
-**Q: 如果要创建一个自己的应用，我需要准备什么？**
-
-A: 我们假定你已经有了 OpenAI 或 Claude 等模型的 API Key，如果没有请去注册一个。如果你已经有了一些内容可以作为训练上下文，就太好了。
-
-**Q: 提供哪些界面语言？**
-
-A:  支持英文、中文，你可以为我们贡献语言包并提供维护支持。
+如果您需要自定义配置，请参考我们的 [docker-compose.yml](docker/docker-compose.yaml) 文件中的注释，并手动设置环境配置。更改后，请再次运行 `docker-compose up -d`。您可以在我们的[文档](https://docs.dify.ai/getting-started/install-self-hosted/environments)中查看所有环境变量的完整列表。
 
 ## Star History
 
 [![Star History Chart](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
 
 
-## 贡献
+## 社区与支持
 
-我们欢迎你为 Dify 作出贡献帮助 Dify 变得更好。我们欢迎各种方式的贡献，提交代码、问题、新想法、或者分享你基于 Dify 创建出的各种有趣有用的 AI 应用。同时，我们也欢迎你在不同的活动、研讨会、社交媒体上分享 Dify。
+我们欢迎您为 Dify 做出贡献，以帮助改善 Dify。包括：提交代码、问题、新想法，或分享您基于 Dify 创建的有趣且有用的 AI 应用程序。同时，我们也欢迎您在不同的活动、会议和社交媒体上分享 Dify。
 
-### 贡献代码
-为了确保正确审查，所有代码贡献 - 包括来自具有直接提交更改权限的贡献者 - 都必须提交 PR 请求并在合并分支之前得到核心开发人员的批准。
+- [GitHub Issues](https://github.com/langgenius/dify/issues)。👉：使用 Dify.AI 时遇到的错误和问题，请参阅[贡献指南](CONTRIBUTING.md)。
+- [电子邮件支持](mailto:hello@dify.ai?subject=[GitHub]Questions%20About%20Dify)。👉：关于使用 Dify.AI 的问题。
+- [Discord](https://discord.gg/FngNHpbcY7)。👉：分享您的应用程序并与社区交流。
+- [Twitter](https://twitter.com/dify_ai)。👉：分享您的应用程序并与社区交流。
+- [商业许可](mailto:business@dify.ai?subject=[GitHub]Business%20License%20Inquiry)。👉：有关商业用途许可 Dify.AI 的商业咨询。
+ - [微信]() 👉：扫描下方二维码，添加微信好友，备注 Dify，我们将邀请您加入 Dify 社区。  
+<img src="./images/wechat.png" alt="wechat" width="100"/>
 
-我们欢迎所有人提交 PR！如果您愿意提供帮助，可以在 [贡献指南](CONTRIBUTING_CN.md) 中了解有关如何为项目做出代码贡献的更多信息。
-
-### 提交问题或想法
-你可以通过 Dify 代码仓库新增 issues 来提交你的问题或想法。如遇到问题，请尽可能描述你遇到问题的操作步骤，以便我们更好地发现它。如果你对我们的产品有任何新想法，也欢迎向我们反馈，请尽可能多地分享你的见解，以便我们在社区中获得更多反馈和进一步讨论。
-
-### 分享你的应用
-我们鼓励所有社区成员分享他们基于 Dify 创造出的 AI 应用，它们可以是应用于不同情景或不同用户，这将有助于为希望基于 AI 能力创造的人们提供强大灵感！你可以通过 [Dify-user-case 仓库项目提交 issue](https://github.com/langgenius/dify-user-case) 来分享你的应用案例。
-
-### 向别人分享 Dify
-我们鼓励社区贡献者们积极展示你使用 Dify 的不同角度。你可以通过线下研讨会、博客或社交媒体上谈论或分享你使用 Dify 的任意功能，相信你独特的使用分享会给别人带来非常大的帮助！如果你需要任何指导帮助，欢迎联系我们 support@dify.ai ,你也可以在 twitter @Dify.AI 或在 [Discord 社区](https://discord.gg/FngNHpbcY7)交流来帮助你传播信息。
-
-### 帮助别人
-你还可以在 Discord、GitHub issues或其他社交平台上帮助需要帮助的人，指导别人解决使用过程中遇到的问题和分享使用经验。这也是个非常了不起的贡献！如果你希望成为 Dify 社区的维护者，请通过[Discord 社区](https://discord.gg/FngNHpbcY7) 联系官方团队或邮件联系我们 support@dify.ai.
-
-
-## 联系我们
-
-如果您有任何问题、建议或合作意向，欢迎通过以下方式联系我们：
-
-- 在我们的 [GitHub Repo](https://github.com/langgenius/dify) 上提交 Issue 或 PR
-- 在我们的 [Discord 社区](https://discord.gg/FngNHpbcY7) 上加入讨论
-- 发送邮件至 hello@dify.ai
-
-## 安全
+## 安全问题
 
 为了保护您的隐私，请避免在 GitHub 上发布安全问题。发送问题至 security@dify.ai，我们将为您做更细致的解答。
 
-## Citation
-
-本软件使用了以下开源软件：
-
-- Chase, H. (2022). LangChain [Computer software]. https://github.com/hwchase17/langchain
-
-更多信息，请参考相应软件的官方网站或许可证文本。
-
 ## License
 
 本仓库遵循 [Dify Open Source License](LICENSE) 开源协议。

README_ES.md

@@ -1,4 +1,4 @@
-![](./images/describe-en.png)
+[![](./images/describe.png)](https://dify.ai)
 <p align="center">
   <a href="./README.md">English</a> |
   <a href="./README_CN.md">简体中文</a> |
@@ -6,36 +6,71 @@
   <a href="./README_ES.md">Español</a>
 </p>
 
-[Sitio web](https://dify.ai) • [Documentación](https://docs.dify.ai) • [Twitter](https://twitter.com/dify_ai) • [Discord](https://discord.gg/FngNHpbcY7)
+<p align="center">
+    <a href="https://dify.ai" target="_blank">
+        <img alt="Static Badge" src="https://img.shields.io/badge/AI-Dify?logo=AI&logoColor=%20%23f5f5f5&label=Dify&labelColor=%20%23155EEF&color=%23EAECF0"></a>
+    <a href="https://discord.gg/FngNHpbcY7" target="_blank">
+        <img src="https://img.shields.io/discord/1082486657678311454?logo=discord"
+            alt="chat on Discord"></a>
+    <a href="https://twitter.com/intent/follow?screen_name=dify_ai" target="_blank">
+        <img src="https://img.shields.io/twitter/follow/dify_ai?style=social&logo=X"
+            alt="follow on Twitter"></a>
+    <a href="https://hub.docker.com/u/langgenius" target="_blank">
+        <img alt="Docker Pulls" src="https://img.shields.io/docker/pulls/langgenius/dify-web"></a>
+</p>
 
-**Dify** es una plataforma LLMOps fácil de usar diseñada para capacitar a más personas para que creen aplicaciones sostenibles basadas en IA. Con orquestación visual para varios tipos de aplicaciones, Dify ofrece aplicaciones listas para usar que también pueden funcionar como APIs de Backend-as-a-Service. Unifica tu proceso de desarrollo con una API para la integración de complementos y conjuntos de datos, y agiliza tus operaciones utilizando una interfaz única para la ingeniería de indicaciones, análisis visual y mejora continua.
+**Dify** es una plataforma de desarrollo de aplicaciones para modelos de lenguaje de gran tamaño (LLM) que ya ha visto la creación de más de **100,000** aplicaciones basadas en Dify.AI. Integra los conceptos de Backend como Servicio y LLMOps, cubriendo el conjunto de tecnologías esenciales requerido para construir aplicaciones nativas de inteligencia artificial generativa, incluyendo un motor RAG incorporado. Con Dify, **puedes auto-desplegar capacidades similares a las de Assistants API y GPTs basadas en cualquier LLM.**
 
-Las aplicaciones creadas con Dify incluyen:
+![](./images/demo.png)
 
-- Sitios web listos para usar que admiten el modo de formulario y el modo de conversación por chat.
-- Una API única que abarca capacidades de complementos, mejora de contexto y más, lo que te ahorra esfuerzo de programación en el backend.
-- Análisis visual de datos, revisión de registros y anotación para aplicaciones.
+## Utilizar Servicios en la Nube
 
-Dify es compatible con Langchain, lo que significa que gradualmente admitiremos múltiples LLMs, actualmente compatibles con:
+Usar [Dify.AI Cloud](https://dify.ai) proporciona todas las capacidades de la versión de código abierto, e incluye un complemento de 200 créditos de prueba para GPT.
 
-- GPT 3 (text-davinci-003)
-- GPT 3.5 Turbo (ChatGPT)
-- GPT-4
+## Por qué Dify
 
-## Usar servicios en la nube
+Dify se caracteriza por su neutralidad de modelo y es un conjunto tecnológico completo e ingenierizado, en comparación con las bibliotecas de desarrollo codificadas como LangChain. A diferencia de la API de Assistants de OpenAI, Dify permite el despliegue local completo de los servicios.
 
-Visita [Dify.ai](https://dify.ai)
+| Característica | Dify.AI | API de Assistants | LangChain |
+|----------------|---------|------------------|-----------|
+| **Enfoque de Programación** | Orientado a API | Orientado a API | Orientado a Código en Python |
+| **Estrategia del Ecosistema** | Código Abierto | Cerrado y Comercial | Código Abierto |
+| **Motor RAG** | Soportado | Soportado | No Soportado |
+| **IDE de Prompts** | Incluido | Incluido | Ninguno |
+| **LLMs Soportados** | Gran Variedad | Solo GPT | Gran Variedad |
+| **Despliegue Local** | Soportado | No Soportado | No Aplicable |
+
+## Características
+
+![](./images/models.png)
+
+**1. Soporte LLM**: Integración con la familia de modelos GPT de OpenAI, o los modelos de la familia Llama2 de código abierto. De hecho, Dify soporta modelos comerciales convencionales y modelos de código abierto (desplegados localmente o basados en MaaS).
+
+**2. IDE de Prompts**: Orquestación visual de aplicaciones y servicios basados en LLMs con tu equipo.
+
+**3. Motor RAG**: Incluye varias capacidades RAG basadas en indexación de texto completo o incrustaciones de base de datos vectoriales, permitiendo la carga directa de PDFs, TXTs y otros formatos de texto.
+
+**4. Agentes**: Un marco de Agentes basado en Llamadas de Función que permite a los usuarios configurar lo que ven es lo que obtienen. Dify incluye capacidades básicas de plugins como la Búsqueda de Google.
+
+**5. Operaciones Continuas**: Monitorear y analizar registros de aplicaciones y rendimiento, mejorando continuamente Prompts, conjuntos de datos o modelos usando datos de producción.
+
+## Antes de Empezar
+
+- [Sitio web](https://dify.ai)
+- [Documentación](https://docs.dify.ai)
+- [Documentación de Implementación](https://docs.dify.ai/getting-started/install-self-hosted)
+- [Preguntas Frecuentes](https://docs.dify.ai/getting-started/faq)
 
 ## Instalar la Edición Comunitaria
 
-### Requisitos del sistema
+### Requisitos del Sistema
 
-Antes de instalar Dify, asegúrate de que tu máquina cumple con los siguientes requisitos mínimos del sistema:
+Antes de instalar Dify, asegúrate de que tu máquina cumpla con los siguientes requisitos mínimos del sistema:
 
-- CPU >= 2 Core
+- CPU >= 2 núcleos
 - RAM >= 4GB
 
-### Inicio rápido
+### Inicio Rápido
 
 La forma más sencilla de iniciar el servidor de Dify es ejecutar nuestro archivo [docker-compose.yml](docker/docker-compose.yaml). Antes de ejecutar el comando de instalación, asegúrate de que [Docker](https://docs.docker.com/get-docker/) y [Docker Compose](https://docs.docker.com/compose/install/) estén instalados en tu máquina:
 
@@ -44,80 +79,34 @@ cd docker
 docker compose up -d
 ```
 
-Después de ejecutarlo, puedes acceder al panel de control de Dify en tu navegador desde [http://localhost/install](http://localhost/install) y comenzar el proceso de instalación de inicialización.
+Después de ejecutarlo, puedes acceder al panel de control de Dify en tu navegador en [http://localhost/install](http://localhost/install) y comenzar el proceso de instalación de inicialización.
 
-### Helm Chart
+### Gráfico Helm
 
-Un gran agradecimiento a @BorisPolonsky por proporcionarnos una versión de [Helm Chart](https://helm.sh/), que permite desplegar Dify en Kubernetes.
-Puede ir a https://github.com/BorisPolonsky/dify-helm para obtener información de despliegue.
+Un gran agradecimiento a @BorisPolonsky por proporcionarnos una versión del [Gráfico Helm](https://helm.sh/), que permite implementar Dify en Kubernetes. Puedes visitar https://github.com/BorisPolonsky/dify-helm para obtener información sobre la implementación.
 
 ### Configuración
 
-Si necesitas personalizar la configuración, consulta los comentarios en nuestro archivo [docker-compose.yml](docker/docker-compose.yaml) y configura manualmente la configuración del entorno. Después de realizar los cambios, ejecuta nuevamente 'docker-compose up -d'.
+Si necesitas personalizar la configuración, consulta los comentarios en nuestro archivo [docker-compose.yml](docker/docker-compose.yaml) y configura manualmente la configuración del entorno. Después de realizar los cambios, ejecuta nuevamente `docker-compose up -d`. Puedes ver la lista completa de variables de entorno en nuestra [documentación](https://docs.dify.ai/getting-started/install-self-hosted/environments).
 
-## Hoja de ruta
+## Historial de Estrellas
 
-Funciones en desarrollo:
+[![Gráfico de Historial de Estrellas](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
 
-- **Conjuntos de datos**, admitiendo más conjuntos de datos, por ejemplo, sincronización de contenido desde Notion o páginas web.
-Admitiremos más conjuntos de datos, incluidos texto, páginas web e incluso contenido de Notion. Los usuarios pueden construir aplicaciones de IA basadas en sus propias fuentes de datos
-- **Complementos**, introduciendo complementos estándar de ChatGPT para aplicaciones, o utilizando complementos producidos por Dify.
-Lanzaremos complementos que cumplan con el estándar de ChatGPT, o nuestros propios complementos de Dify para habilitar más capacidades en las aplicaciones.
-- **Modelos de código abierto**, por ejemplo, adoptar Llama como proveedor de modelos o para un ajuste adicional.
-Trabajaremos con excelentes modelos de código abierto como Llama, proporcionándolos como opciones de modelos en nuestra plataforma o utilizándolos para un ajuste adicional.
+## Comunidad y Soporte
 
-## Preguntas y respuestas
+Te damos la bienvenida a contribuir a Dify para ayudar a hacer que Dify sea mejor de diversas maneras, enviando código, informando problemas, proponiendo nuevas ideas o compartiendo las aplicaciones de inteligencia artificial interesantes y útiles que hayas creado basadas en Dify. Al mismo tiempo, también te invitamos a compartir Dify en diferentes eventos, conferencias y redes sociales.
 
-**P: ¿Qué puedo hacer con Dify?**
+- [Problemas en GitHub](https://github.com/langgenius/dify/issues). Lo mejor para: errores y problemas que encuentres al usar Dify.AI, consulta la [Guía de Contribución](CONTRIBUTING.md).
+- [Soporte por Correo Electrónico](mailto:hello@dify.ai?subject=[GitHub]Preguntas%20sobre%20Dify). Lo mejor para: preguntas que tengas sobre el uso de Dify.AI.
+- [Discord](https://discord.gg/FngNHpbcY7). Lo mejor para: compartir tus aplicaciones y socializar con la comunidad.
+- [Twitter](https://twitter.com/dify_ai). Lo mejor para: compartir tus aplicaciones y socializar con la comunidad.
+- [Licencia Comercial](mailto:business@dify.ai?subject=[GitHub]Consulta%20de%20Licencia%20Comercial). Lo mejor para: consultas comerciales sobre la licencia de Dify.AI para uso comercial.
 
-R: Dify es una herramienta de desarrollo y operaciones de LLM, simple pero poderosa. Puedes usarla para construir aplicaciones de calidad comercial y asistentes personales. Si deseas desarrollar tus propias aplicaciones, LangDifyGenius puede ahorrarte trabajo en el backend al integrar con OpenAI y ofrecer capacidades de operaciones visuales, lo que te permite mejorar y entrenar continuamente tu modelo GPT.
-
-**P: ¿Cómo uso Dify para "entrenar" mi propio modelo?**
-
-R: Una aplicación valiosa consta de Ingeniería de indicaciones, mejora de contexto y ajuste fino. Hemos creado un enfoque de programación híbrida que combina las indicaciones con lenguajes de programación (similar a un motor de plantillas), lo que facilita la incorporación de texto largo o la captura de subtítulos de un video de YouTube ingresado por el usuario, todo lo cual se enviará como contexto para que los LLM lo procesen. Damos gran importancia a la operabilidad de la aplicación, con los datos generados por los usuarios durante el uso de la aplicación disponibles para análisis, anotación y entrenamiento continuo. Sin las herramientas adecuadas, estos pasos pueden llevar mucho tiempo.
-
-**P: ¿Qué necesito preparar si quiero crear mi propia aplicación?**
-
-R: Suponemos que ya tienes una clave de API de OpenAI; si no la tienes, por favor regístrate. ¡Si ya tienes contenido que pueda servir como contexto de entrenamiento, eso es genial!
-
-**P: ¿Qué idiomas de interfaz están disponibles?**
-
-R: Actualmente se admiten inglés y chino, y puedes contribuir con paquetes de idiomas.
-
-## Historial de estrellas
-
-[![Gráfico de historial de estrellas](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
-
-## Contáctanos
-
-Si tienes alguna pregunta, sugerencia o consulta sobre asociación, no dudes en contactarnos a través de los siguientes canales:
-
-- Presentar un problema o una solicitud de extracción en nuestro repositorio de GitHub.
-- Únete a la discusión en nuestra comunidad de [Discord](https://discord.gg/FngNHpbcY7).
-- Envía un correo electrónico a hello@dify.ai.
-
-¡Estamos ansiosos por ayudarte y crear juntos aplicaciones de IA más divertidas y útiles!
-
-## Contribuciones
-
-Para garantizar una revisión adecuada, todas las contribuciones de código, incluidas las de los colaboradores con acceso directo a los compromisos, deben enviarse mediante solicitudes de extracción y ser aprobadas por el equipo principal de
-
- desarrollo antes de fusionarse.
-
-¡Agradecemos todas las solicitudes de extracción! Si deseas ayudar, consulta la [Guía de Contribución](CONTRIBUTING.md) para obtener más información sobre cómo comenzar.
-
-## Seguridad
+## Divulgación de Seguridad
 
 Para proteger tu privacidad, evita publicar problemas de seguridad en GitHub. En su lugar, envía tus preguntas a security@dify.ai y te proporcionaremos una respuesta más detallada.
 
-## Citación
-
-Este software utiliza el siguiente software de código abierto:
-
-- Chase, H. (2022). LangChain [Software de computadora]. https://github.com/hwchase17/langchain
-
-Para obtener más información, consulta el sitio web oficial o el texto de la licencia del software correspondiente.
-
 ## Licencia
 
 Este repositorio está disponible bajo la [Licencia de código abierto de Dify](LICENSE).

README_JA.md

@@ -1,122 +1,103 @@
-![](./images/describe-en.png)
+[![](./images/describe.png)](https://dify.ai)
 <p align="center">
   <a href="./README.md">English</a> |
   <a href="./README_CN.md">简体中文</a> |
   <a href="./README_JA.md">日本語</a> |
-  <a href="./README_ES.md">Español</a>
+  <a href="./README_ES.md">Español</a> |
+  <a href="./README_KL.md">Klingon</a>
 </p>
 
-[Web サイト](https://dify.ai) • [ドキュメント](https://docs.dify.ai) • [Twitter](https://twitter.com/dify_ai) • [Discord](https://discord.gg/FngNHpbcY7)
+<p align="center">
+    <a href="https://dify.ai" target="_blank">
+        <img alt="Static Badge" src="https://img.shields.io/badge/AI-Dify?logo=AI&logoColor=%20%23f5f5f5&label=Dify&labelColor=%20%23155EEF&color=%23EAECF0"></a>
+    <a href="https://discord.gg/FngNHpbcY7" target="_blank">
+        <img src="https://img.shields.io/discord/1082486657678311454?logo=discord"
+            alt="chat on Discord"></a>
+    <a href="https://twitter.com/intent/follow?screen_name=dify_ai" target="_blank">
+        <img src="https://img.shields.io/twitter/follow/dify_ai?style=social&logo=X"
+            alt="follow on Twitter"></a>
+    <a href="https://hub.docker.com/u/langgenius" target="_blank">
+        <img alt="Docker Pulls" src="https://img.shields.io/docker/pulls/langgenius/dify-web"></a>
+</p>
 
+"Difyは、既にDify.AI上で10万以上のアプリケーションが構築されているLLMアプリケーション開発プラットフォームです。バックエンド・アズ・ア・サービスとLLMOpsの概念を統合し、組み込みのRAGエンジンを含む、生成AIネイティブアプリケーションを構築するためのコアテックスタックをカバーしています。Difyを使用すると、どのLLMに基づいても、Assistants APIやGPTのような機能を自己デプロイすることができます。"
 
-**Dify** は、より多くの人々が持続可能な AI ネイティブアプリケーションを作成できるように設計された、使いやすい LLMOps プラットフォームです。様々なアプリケーションタイプに対応したビジュアルオーケストレーションにより Dify は Backend-as-a-Service API としても機能する、すぐに使えるアプリケーションを提供します。プラグインやデータセットを統合するための1つの API で開発プロセスを統一し、プロンプトエンジニアリング、ビジュアル分析、継続的な改善のための1つのインターフェイスを使って業務を合理化します。
-
-Difyで作成したアプリケーションは以下の通りです:
-
-フォームモードとチャット会話モードをサポートする、すぐに使える Web サイト
-プラグイン機能、コンテキストの強化などを網羅する単一の API により、バックエンドのコーディングの手間を省きます。
-アプリケーションの視覚的なデータ分析、ログレビュー、アノテーションが可能です。
-Dify は LangChain と互換性があり、複数の LLM を徐々にサポートします:
-
-- GPT 3 (text-davinci-003)
-- GPT 3.5 Turbo(ChatGPT)
-- GPT-4
+Please note that translating complex technical terms can sometimes result in slight variations in meaning due to differences in language nuances.
 
 ## クラウドサービスの利用
 
-[Dify.ai](https://dify.ai) をご覧ください
+[Dify.AI Cloud](https://dify.ai) を使用すると、オープンソース版の全機能を利用でき、さらに200GPTのトライアルクレジットが無料で提供されます。
 
-## Community Edition のインストール
+## Difyの利点
+
+Difyはモデルニュートラルであり、LangChainのようなハードコードされた開発ライブラリと比較して、完全にエンジニアリングされた技術スタックを特徴としています。OpenAIのAssistants APIとは異なり、Difyではサービスの完全なローカルデプロイメントが可能です。
+
+| 機能 | Dify.AI | Assistants API | LangChain |
+|---------|---------|----------------|-----------|
+| **プログラミングアプローチ** | API指向 | API指向 | Pythonコード指向 |
+| **エコシステム戦略** | オープンソース | 閉鎖的かつ商業的 | オープンソース |
+| **RAGエンジン** | サポート済み | サポート済み | 非サポート |
+| **プロンプトIDE** | 含まれる | 含まれる | なし |
+| **サポートされるLLMs** | 豊富な種類 | GPTのみ | 豊富な種類 |
+| **ローカルデプロイメント** | サポート済み | 非サポート | 該当なし |
+
+
+## 開始する前に
+
+- [Website](https://dify.ai)
+- [Docs](https://docs.dify.ai)
+- [Deployment Docs](https://docs.dify.ai/getting-started/install-self-hosted)
+- [FAQ](https://docs.dify.ai/getting-started/faq) 
+
+
+## コミュニティエディションのインストール
 
 ### システム要件
 
-Dify をインストールする前に、お使いのマシンが以下の最低システム要件を満たしていることを確認してください:
+Difyをインストールする前に、以下の最低限のシステム要件を満たしていることを確認してください：
 
-- CPU >= 1 Core
+- CPU >= 2コア
 - RAM >= 4GB
 
 ### クイックスタート
 
-Dify サーバーを起動する最も簡単な方法は、[docker-compose.yml](docker/docker-compose.yaml) ファイルを実行することです。インストールコマンドを実行する前に、[Docker](https://docs.docker.com/get-docker/) と [Docker Compose](https://docs.docker.com/compose/install/) がお使いのマシンにインストールされていることを確認してください:
+Difyサーバーを始める最も簡単な方法は、[docker-compose.yml](docker/docker-compose.yaml) ファイルを実行することです。インストールコマンドを実行する前に、マシンに [Docker](https://docs.docker.com/get-docker/) と [Docker Compose](https://docs.docker.com/compose/install/) がインストールされていることを確認してください：
 
 ```bash
 cd docker
 docker compose up -d
 ```
 
-実行後、ブラウザで [http://localhost/install](http://localhost/install) にアクセスし、初期化インストール作業を開始することができます。
+実行後、ブラウザで [http://localhost/install](http://localhost/install) にアクセスし、初期化インストールプロセスを開始できます。
 
 ### Helm Chart
 
-@BorisPolonsky に大感謝します。彼は Dify を Kubernetes 上にデプロイするための [Helm Chart](https://helm.sh/) バージョンを提供してくれました。
+@BorisPolonskyによる[Helm Chart](https://helm.sh/) バージョンを提供してくれて、大変感謝しています。これにより、DifyはKubernetes上にデプロイすることができます。
 デプロイ情報については、https://github.com/BorisPolonsky/dify-helm をご覧ください。
 
-### 構成
+### 設定
 
-カスタマイズが必要な場合は、[docker-compose.yml](docker/docker-compose.yaml) ファイルのコメントを参照し、手動で環境設定をお願いします。変更後、再度 'docker-compose up -d' を実行してください。
-
-## ロードマップ
-
-開発中の機能:
-
-- **データセット**, Notionやウェブページからのコンテンツ同期など、より多くのデータセットをサポートします
-テキスト、ウェブページ、さらには Notion コンテンツなど、より多くのデータセットをサポートする予定です。ユーザーは、自分のデータソースをもとに AI アプリケーションを構築することができます。
-- **プラグイン**, アプリケーションに ChatGPT プラグイン標準のプラグインを導入する、または Dify 制作のプラグインを利用する
-今後、ChatGPT 規格に準拠したプラグインや、ディファイ独自のプラグインを公開し、より多くの機能をアプリケーションで実現できるようにします。
-- **オープンソースモデル**, 例えばモデルプロバイダーとして Llama を採用したり、さらにファインチューニングを行う
-Llama のような優れたオープンソースモデルを、私たちのプラットフォームのモデルオプションとして提供したり、さらなる微調整のために使用したりすることで、協力していきます。
+設定をカスタマイズする必要がある場合は、[docker-compose.yml](docker/docker-compose.yaml) ファイルのコメントを参照し、環境設定を手動で行ってください。変更を行った後は、もう一度 `docker-compose up -d` を実行してください。環境変数の完全なリストは、[ドキュメント](https://docs.dify.ai/getting-started/install-self-hosted/environments)で確認できます。
 
 
-## Q&A
-
-**Q: Dify で何ができるのか？**
-
-A: Dify はシンプルでパワフルな LLM 開発・運用ツールです。商用グレードのアプリケーション、パーソナルアシスタントを構築するために使用することができます。独自のアプリケーションを開発したい場合、LangDifyGenius は OpenAI と統合する際のバックエンド作業を省き、視覚的な操作機能を提供し、GPT モデルを継続的に改善・訓練することが可能です。
-
-**Q: Dify を使って、自分のモデルを「トレーニング」するにはどうすればいいのでしょうか？**
-
-A: プロンプトエンジニアリング、コンテキスト拡張、ファインチューニングからなる価値あるアプリケーションです。プロンプトとプログラミング言語を組み合わせたハイブリッドプログラミングアプローチ（テンプレートエンジンのようなもの）で、長文の埋め込みやユーザー入力の YouTube 動画からの字幕取り込みなどを簡単に実現し、これらはすべて LLM が処理するコンテキストとして提出される予定です。また、アプリケーションの操作性を重視し、ユーザーがアプリケーションを使用する際に生成したデータを分析、アノテーション、継続的なトレーニングに利用できるようにしました。適切なツールがなければ、これらのステップに時間がかかることがあります。
-
-**Q: 自分でアプリケーションを作りたい場合、何を準備すればよいですか？**
-
-A: すでに OpenAI API Key をお持ちだと思いますが、お持ちでない場合はご登録ください。もし、すでにトレーニングのコンテキストとなるコンテンツをお持ちでしたら、それは素晴らしいことです！
-
-**Q: インターフェイスにどの言語が使えますか？**
-
-A: 現在、英語と中国語に対応しており、言語パックを寄贈することも可能です。
-
-## Star ヒストリー
+## スターヒストリー
 
 [![Star History Chart](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
 
-## お問合せ
+## コミュニティとサポート
 
-ご質問、ご提案、パートナーシップに関するお問い合わせは、以下のチャンネルからお気軽にご連絡ください:
+Difyに貢献していただき、コードの提出、問題の報告、新しいアイデアの提供、またはDifyを基に作成した興味深く有用なAIアプリケーションの共有により、Difyをより良いものにするお手伝いを歓迎します。同時に、さまざまなイベント、会議、ソーシャルメディアでDifyを共有することも歓迎します。
 
-- GitHub Repo で Issue や PR を提出する
-- [Discord](https://discord.gg/FngNHpbcY7) コミュニティで議論に参加する。
-- hello@dify.ai にメールを送信します
-
-私たちは、皆様のお手伝いをさせていただき、より楽しく、より便利な AI アプリケーションを一緒に作っていきたいと思っています！
-
-## コントリビュート
-
-適切なレビューを行うため、コミットへの直接アクセスが可能なコントリビュータを含むすべてのコードコントリビュータは、プルリクエストで提出し、マージされる前にコア開発チームによって承認される必要があります。
-
-私たちはすべてのプルリクエストを歓迎します！協力したい方は、[コントリビューションガイド](CONTRIBUTING.md) をチェックしてみてください。
+- [GitHub Issues](https://github.com/langgenius/dify/issues)。最適な使用法：Dify.AIの使用中に遭遇するバグやエラー、[貢献ガイド](CONTRIBUTING.md)を参照。
+- [Email サポート](mailto:hello@dify.ai?subject=[GitHub]Questions%20About%20Dify)。最適な使用法：Dify.AIの使用に関する質問。
+- [Discord](https://discord.gg/FngNHpbcY7)。最適な使用法：アプリケーションの共有とコミュニティとの交流。
+- [Twitter](https://twitter.com/dify_ai)。最適な使用法：アプリケーションの共有とコミュニティとの交流。
+- [ビジネスライセンス](mailto:business@dify.ai?subject=[GitHub]Business%20License%20Inquiry)。最適な使用法：Dify.AIを商業利用するためのビジネス関連の問い合わせ。
 
 ## セキュリティ
 
 プライバシー保護のため、GitHub へのセキュリティ問題の投稿は避けてください。代わりに、あなたの質問を security@dify.ai に送ってください。より詳細な回答を提供します。
 
-## 引用
-
-本ソフトウェアは、以下のオープンソースソフトウェアを使用しています:
-
-- Chase, H. (2022). LangChain [Computer software]. https://github.com/hwchase17/langchain
-
-詳しくは、各ソフトウェアの公式サイトまたはライセンス文をご参照ください。
-
 ## ライセンス
 
 このリポジトリは、[Dify Open Source License](LICENSE) のもとで利用できます。

README_KL.md

@@ -0,0 +1,114 @@
+[![](./images/describe.png)](https://dify.ai)
+<p align="center">
+  <a href="./README.md">English</a> |
+  <a href="./README_CN.md">简体中文</a> |
+  <a href="./README_JA.md">日本語</a> |
+  <a href="./README_ES.md">Español</a> |
+  <a href="./README_KL.md">Klingon</a>
+</p>
+
+<p align="center">
+    <a href="https://dify.ai" target="_blank">
+        <img alt="Static Badge" src="https://img.shields.io/badge/AI-Dify?logo=AI&logoColor=%20%23f5f5f5&label=Dify&labelColor=%20%23155EEF&color=%23EAECF0"></a>
+    <a href="https://discord.gg/FngNHpbcY7" target="_blank">
+        <img src="https://img.shields.io/discord/1082486657678311454?logo=discord"
+            alt="chat on Discord"></a>
+    <a href="https://twitter.com/intent/follow?screen_name=dify_ai" target="_blank">
+        <img src="https://img.shields.io/twitter/follow/dify_ai?style=social&logo=X"
+            alt="follow on Twitter"></a>
+    <a href="https://hub.docker.com/u/langgenius" target="_blank">
+        <img alt="Docker Pulls" src="https://img.shields.io/docker/pulls/langgenius/dify-web"></a>
+</p>
+
+**Dify** Hoch LLM qorwI' pIqoDvam pagh laHta' je **100,000** pIqoDvamvam Dify.AI De'wI'. Dify leghpu' Backend chu' a Service teH LLMOps vItlhutlh, generative AI-native pIqoD teq wa'vam, vIyoD Built-in RAG engine. Dify, **'ej chenmoHmoH Hoch 'oHna' Assistant API 'ej GPTmey HoStaHbogh LLMmey.**
+
+![](./images/demo.png)
+
+## ngIl QaQ
+
+[Dify.AI ngIl](https://dify.ai) pIm neHlaH 'ej ghaH. cha'logh wa' DIvI' 200 GPT trial credits.
+
+## Dify WovmoH
+
+Dify Daq rIn neutrality 'ej Hoch, LangChain tInHar HubwI'. maH Daqbe'law' Qawqar, OpenAI's Assistant API Daq local neH deployment.
+
+| Qo'logh | Dify.AI | Assistants API | LangChain |
+|---------|---------|----------------|-----------|
+| **qet QaS** | API-oriented | API-oriented | Python Code-oriented |
+| **Ecosystem Strategy** | Open Source | Closed and Commercial | Open Source |
+| **RAG Engine** | Ha'qu' | Ha'qu' | ghoS Ha'qu' |
+| **Prompt IDE** | jaH Include | jaH Include | qeylIS qaq |
+| **qet LLMmey** | bo'Degh Hoch | GPTmey tIn | bo'Degh Hoch |
+| **local deployment** | Ha'qu' | tInHa'qu' | tInHa'qu' ghogh |
+
+## ruch
+
+![](./images/models.png)
+
+**1. LLM tIq**: OpenAI's GPT Hur nISmoHvam neH vIngeH, wa' Llama2 Hur nISmoHvam. Heghlu'lu'pu' Dify mIw 'oH choH qay'be'.Daq commercial Hurmey 'ej Open Source Hurmey (maqtaHvIS pagh locally neH neH deployment HoSvam).
+
+**2. Prompt IDE**: cha'logh wa' LLMmey Hoch janlu'pu' 'ej lughpu' choH qay'be'.
+
+**3. RAG Engine**: RAG vaD tIqpu' lo'taH indexing qor neH vector database wa' embeddings wIj, PDFs, TXTs, 'ej ghojmoHmoH HIq qorlIj je upload.
+
+**4. jenSuvpu'**: jenbe' SuDqang naQ moDwu' jenSuvpu' porgh cha'logh choHvam. Dify Google Search Hur vItlhutlh plugin choH.
+
+**5. QaS muDHa'wI': cha'logh wa' pIq mI' logs 'ej quv yIn, vItlhutlh tIq 'e'wIj lo'taHmoHmoH Prompts, vItlhutlh, Hurmey ghaH production data jatlh.
+
+## Do'wI' qabmey lo'taH
+
+- [Website](https://dify.ai)
+- [Docs](https://docs.dify.ai)
+- [lo'taHmoH Docs](https://docs.dify.ai/getting-started/install-self-hosted)
+- [FAQ](https://docs.dify.ai/getting-started/faq) 
+
+## Community Edition tu' yo'
+
+### System Qab
+
+Dify yo' yo' qaqmeH SuS chenmoH 'oH qech!
+
+- CPU >= 2 Cores
+- RAM >= 4GB
+
+### Quick Start
+
+Dify server luHoHtaHlu' vIngeH lo'laHbe'chugh vIyoD [docker-compose.yml](docker/docker-compose.yaml) QorwI'ghach. toH yItlhutlh chenmoH luH!chugh 'ay' vaj vIneHmeH, 'ej [Docker](https://docs.docker.com/get-docker/) 'ej [Docker Compose](https://docs.docker.com/compose/install/) vaj 'oH 'e' vIneHmeH:
+
+```bash
+cd docker
+docker compose up -d
+```
+
+luHoHtaHmeH HoHtaHvIS, Dify dashboard vIneHmeH vIngeH lI'wI' [http://localhost/install](http://localhost/install) 'ej 'oH initialization 'e' vIneHmeH.
+
+### Helm Chart
+
+@BorisPolonsky Dify wIq tIq ['ay'var (Helm Chart)](https://helm.sh/) version Hur yIn chu' Dify luHoHchu'. Heghlu'lu' vIneHmeH [https://github.com/BorisPolonsky/dify-helm](https://github.com/BorisPolonsky/dify-helm) 'ej vaj QaS deployment information.
+
+### veS config
+
+chenmoHDI' config lo'taH ghaH, vItlhutlh HIq wIgharghbe'lu'pu'. toH lo'taHvIS pagh vay' vIneHmeH, 'ej `docker-compose up -d` wa'DIch. tIqmoHmeH list full wa' lo'taHvo'lu'pu' ghaH [docs](https://docs.dify.ai/getting-started/install-self-hosted/environments).
+
+## tIng qem
+
+[![tIng qem Hur Chart](https://api.star-history.com/svg?repos=langgenius/dify&type=Date)](https://star-history.com/#langgenius/dify&Date)
+
+## choHmoH 'ej vItlhutlh
+
+Dify choHmoH je mIw Dify puqloD, Dify ghaHta'bogh vItlhutlh, HurDI' code, ghItlh, ghItlh qo'lu'pu'pu' qej. tIqmeH, Hurmey je, Dify Hur tIqDI' woDDaj, DuD QangmeH 'ej HInobDaq vItlhutlh HImej Dify'e'.
+
+- [GitHub vItlhutlh](https://github.com/langgenius/dify/issues). Hurmey: bugs 'ej errors Dify.AI tIqmeH. yImej [Contribution Guide](CONTRIBUTING.md).
+- [Email QaH](mailto:hello@dify.ai?subject=[GitHub]Questions%20About%20Dify). Hurmey: questions vItlhutlh Dify.AI chaw'.
+- [Discord](https://discord.gg/FngNHpbcY7). Hurmey: jIpuv 'ej jImej mIw Dify vItlhutlh.
+- [Twitter](https://twitter.com/dify_ai). Hurmey: jIpuv 'ej jImej mIw Dify vItlhutlh.
+- [Business License](mailto:business@dify.ai?subject=[GitHub]Business%20License%20Inquiry). Hurmey: qurgh vItlhutlh Hurmey Dify.AI tIqbe'law'.
+
+## bIQDaqmey bom
+
+taghlI' vIngeH'a'? pong security 'oH posting GitHub. yItlhutlh, toH security@dify.ai 'ej vIngeH'a'.
+
+## License
+
+ghItlh puqloD chenmoH [Dify vItlhutlh Hur](LICENSE), ghaH nIvbogh Apache 2.0.
+

api/.env.example

@@ -18,6 +18,9 @@ SERVICE_API_URL=http://127.0.0.1:5001
 APP_API_URL=http://127.0.0.1:5001
 APP_WEB_URL=http://127.0.0.1:3000
 
+# Files URL
+FILES_URL=http://127.0.0.1:5001
+
 # celery configuration
 CELERY_BROKER_URL=redis://:difyai123456@localhost:6379/1
 
@@ -50,7 +53,7 @@ S3_REGION=your-region
 WEB_API_CORS_ALLOW_ORIGINS=http://127.0.0.1:3000,*
 CONSOLE_CORS_ALLOW_ORIGINS=http://127.0.0.1:3000,*
 
-# Vector database configuration, support: weaviate, qdrant
+# Vector database configuration, support: weaviate, qdrant, milvus
 VECTOR_STORE=weaviate
 
 # Weaviate configuration
@@ -59,9 +62,24 @@ WEAVIATE_API_KEY=WVF5YThaHlkYwhGUSmCRgsX3tD5ngdN8pkih
 WEAVIATE_GRPC_ENABLED=false
 WEAVIATE_BATCH_SIZE=100
 
-# Qdrant configuration, use `path:` prefix for local mode or `https://your-qdrant-cluster-url.qdrant.io` for remote mode
-QDRANT_URL=path:storage/qdrant
-QDRANT_API_KEY=your-qdrant-api-key
+# Qdrant configuration, use `http://localhost:6333` for local mode or `https://your-qdrant-cluster-url.qdrant.io` for remote mode
+QDRANT_URL=http://localhost:6333
+QDRANT_API_KEY=difyai123456
+
+# Milvus configuration
+MILVUS_HOST=127.0.0.1
+MILVUS_PORT=19530
+MILVUS_USER=root
+MILVUS_PASSWORD=Milvus
+MILVUS_SECURE=false
+
+# Upload configuration
+UPLOAD_FILE_SIZE_LIMIT=15
+UPLOAD_FILE_BATCH_LIMIT=5
+UPLOAD_IMAGE_FILE_SIZE_LIMIT=10
+
+# Model Configuration
+MULTIMODAL_SEND_IMAGE_FORMAT=base64
 
 # Mail configuration, support: resend
 MAIL_TYPE=

api/.vscode/launch.json

@@ -10,7 +10,7 @@
             "request": "launch",
             "module": "flask",
             "env": {
-                "FLASK_APP": "api/app.py",
+                "FLASK_APP": "app.py",
                 "FLASK_DEBUG": "1",
                 "GEVENT_SUPPORT": "True"
             },

api/app.py

@@ -6,7 +6,11 @@ from werkzeug.exceptions import Unauthorized
 if not os.environ.get("DEBUG") or os.environ.get("DEBUG").lower() != 'true':
     from gevent import monkey
     monkey.patch_all()
+    if os.environ.get("VECTOR_STORE") == 'milvus':
+        import grpc.experimental.gevent
+        grpc.experimental.gevent.init_gevent()
 
+import time
 import logging
 import json
 import threading
@@ -16,7 +20,7 @@ from flask_cors import CORS
 
 from core.model_providers.providers import hosted
 from extensions import ext_celery, ext_sentry, ext_redis, ext_login, ext_migrate, \
-    ext_database, ext_storage, ext_mail, ext_stripe
+    ext_database, ext_storage, ext_mail, ext_stripe, ext_code_based_extension
 from extensions.ext_database import db
 from extensions.ext_login import login_manager
 
@@ -33,6 +37,13 @@ from libs.passport import PassportService
 import warnings
 warnings.simplefilter("ignore", ResourceWarning)
 
+# fix windows platform
+if os.name == "nt":
+    os.system('tzutil /s "UTC"')    
+else:
+    os.environ['TZ'] = 'UTC'
+    time.tzset()
+
 
 class DifyApp(Flask):
     pass
@@ -76,6 +87,7 @@ def create_app(test_config=None) -> Flask:
 def initialize_extensions(app):
     # Since the application instance is now created, pass it to each Flask
     # extension instance to bind it to the Flask application instance (app)
+    ext_code_based_extension.init()
     ext_database.init_app(app)
     ext_migrate.init(app, db)
     ext_redis.init_app(app)
@@ -122,6 +134,7 @@ def register_blueprints(app):
     from controllers.service_api import bp as service_api_bp
     from controllers.web import bp as web_bp
     from controllers.console import bp as console_app_bp
+    from controllers.files import bp as files_bp
 
     CORS(service_api_bp,
          allow_headers=['Content-Type', 'Authorization', 'X-App-Code'],
@@ -151,6 +164,12 @@ def register_blueprints(app):
 
     app.register_blueprint(console_app_bp)
 
+    CORS(files_bp,
+         allow_headers=['Content-Type'],
+         methods=['GET', 'PUT', 'POST', 'DELETE', 'OPTIONS', 'PATCH']
+         )
+    app.register_blueprint(files_bp)
+
 
 # create app
 app = create_app()

api/commands.py

@@ -8,6 +8,8 @@ import time
 import uuid
 
 import click
+import qdrant_client
+from qdrant_client.http.models import TextIndexParams, TextIndexType, TokenizerType
 from tqdm import tqdm
 from flask import current_app, Flask
 from langchain.embeddings import OpenAIEmbeddings
@@ -484,6 +486,38 @@ def normalization_collections():
     click.echo(click.style('Congratulations! restore {} dataset indexes.'.format(len(normalization_count)), fg='green'))
 
 
+@click.command('add-qdrant-full-text-index', help='add qdrant full text index')
+def add_qdrant_full_text_index():
+    click.echo(click.style('Start add full text index.', fg='green'))
+    binds = db.session.query(DatasetCollectionBinding).all()
+    if binds and current_app.config['VECTOR_STORE'] == 'qdrant':
+        qdrant_url = current_app.config['QDRANT_URL']
+        qdrant_api_key = current_app.config['QDRANT_API_KEY']
+        client = qdrant_client.QdrantClient(
+            qdrant_url,
+            api_key=qdrant_api_key,  # For Qdrant Cloud, None for local instance
+        )
+        for bind in binds:
+            try:
+                text_index_params = TextIndexParams(
+                    type=TextIndexType.TEXT,
+                    tokenizer=TokenizerType.MULTILINGUAL,
+                    min_token_len=2,
+                    max_token_len=20,
+                    lowercase=True
+                )
+                client.create_payload_index(bind.collection_name, 'page_content',
+                                            field_schema=text_index_params)
+            except Exception as e:
+                click.echo(
+                    click.style('Create full text index error: {} {}'.format(e.__class__.__name__, str(e)),
+                                fg='red'))
+            click.echo(
+                click.style(
+                    'Congratulations! add collection {} full text index successful.'.format(bind.collection_name),
+                    fg='green'))
+
+
 def deal_dataset_vector(flask_app: Flask, dataset: Dataset, normalization_count: list):
     with flask_app.app_context():
         try:
@@ -647,10 +681,10 @@ def update_app_model_configs(batch_size):
 
             pbar.update(len(data_batch))
 
+
 @click.command('migrate_default_input_to_dataset_query_variable')
 @click.option("--batch-size", default=500, help="Number of records to migrate in each batch.")
 def migrate_default_input_to_dataset_query_variable(batch_size):
-
     click.secho("Starting...", fg='green')
 
     total_records = db.session.query(AppModelConfig) \
@@ -658,13 +692,13 @@ def migrate_default_input_to_dataset_query_variable(batch_size):
         .filter(App.mode == 'completion') \
         .filter(AppModelConfig.dataset_query_variable == None) \
         .count()
-    
+
     if total_records == 0:
         click.secho("No data to migrate.", fg='green')
         return
 
     num_batches = (total_records + batch_size - 1) // batch_size
-    
+
     with tqdm(total=total_records, desc="Migrating Data") as pbar:
         for i in range(num_batches):
             offset = i * batch_size
@@ -697,14 +731,14 @@ def migrate_default_input_to_dataset_query_variable(batch_size):
                     for form in user_input_form:
                         paragraph = form.get('paragraph')
                         if paragraph \
-                            and paragraph.get('variable') == 'query':
-                                data.dataset_query_variable = 'query'
-                                break
-                        
+                                and paragraph.get('variable') == 'query':
+                            data.dataset_query_variable = 'query'
+                            break
+
                         if paragraph \
-                            and paragraph.get('variable') == 'default_input':
-                                data.dataset_query_variable = 'default_input'
-                                break
+                                and paragraph.get('variable') == 'default_input':
+                            data.dataset_query_variable = 'default_input'
+                            break
 
                 db.session.commit()
 
@@ -712,7 +746,7 @@ def migrate_default_input_to_dataset_query_variable(batch_size):
                 click.secho(f"Error while migrating data: {e}, app_id: {data.app_id}, app_model_config_id: {data.id}",
                             fg='red')
                 continue
-            
+
             click.secho(f"Successfully migrated batch {i + 1}/{num_batches}.", fg='green')
 
             pbar.update(len(data_batch))
@@ -731,3 +765,4 @@ def register_commands(app):
     app.cli.add_command(update_app_model_configs)
     app.cli.add_command(normalization_collections)
     app.cli.add_command(migrate_default_input_to_dataset_query_variable)
+    app.cli.add_command(add_qdrant_full_text_index)

api/config.py

@@ -26,6 +26,7 @@ DEFAULTS = {
     'SERVICE_API_URL': 'https://api.dify.ai',
     'APP_WEB_URL': 'https://udify.app',
     'APP_API_URL': 'https://udify.app',
+    'FILES_URL': '',
     'STORAGE_TYPE': 'local',
     'STORAGE_LOCAL_PATH': 'storage',
     'CHECK_UPDATE_URL': 'https://updates.dify.ai',
@@ -57,6 +58,10 @@ DEFAULTS = {
     'CLEAN_DAY_SETTING': 30,
     'UPLOAD_FILE_SIZE_LIMIT': 15,
     'UPLOAD_FILE_BATCH_LIMIT': 5,
+    'UPLOAD_IMAGE_FILE_SIZE_LIMIT': 10,
+    'OUTPUT_MODERATION_BUFFER_SIZE': 300,
+    'MULTIMODAL_SEND_IMAGE_FORMAT': 'base64',
+    'INVITE_EXPIRY_HOURS': 72
 }
 
 
@@ -83,78 +88,65 @@ class Config:
     """Application configuration class."""
 
     def __init__(self):
-        # app settings
-        self.CONSOLE_API_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_API_URL')
-        self.CONSOLE_WEB_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_WEB_URL')
-        self.SERVICE_API_URL = get_env('API_URL') if get_env('API_URL') else get_env('SERVICE_API_URL')
-        self.APP_WEB_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_WEB_URL')
-        self.APP_API_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_API_URL')
-        self.CONSOLE_URL = get_env('CONSOLE_URL')
-        self.API_URL = get_env('API_URL')
-        self.APP_URL = get_env('APP_URL')
-        self.CURRENT_VERSION = "0.3.24"
+        # ------------------------
+        # General Configurations.
+        # ------------------------
+        self.CURRENT_VERSION = "0.3.31"
         self.COMMIT_SHA = get_env('COMMIT_SHA')
         self.EDITION = "SELF_HOSTED"
         self.DEPLOY_ENV = get_env('DEPLOY_ENV')
         self.TESTING = False
         self.LOG_LEVEL = get_env('LOG_LEVEL')
 
+        # The backend URL prefix of the console API.
+        # used to concatenate the login authorization callback or notion integration callback.
+        self.CONSOLE_API_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_API_URL')
+
+        # The front-end URL prefix of the console web.
+        # used to concatenate some front-end addresses and for CORS configuration use.
+        self.CONSOLE_WEB_URL = get_env('CONSOLE_URL') if get_env('CONSOLE_URL') else get_env('CONSOLE_WEB_URL')
+
+        # WebApp API backend Url prefix.
+        # used to declare the back-end URL for the front-end API.
+        self.APP_API_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_API_URL')
+
+        # WebApp Url prefix.
+        # used to display WebAPP API Base Url to the front-end.
+        self.APP_WEB_URL = get_env('APP_URL') if get_env('APP_URL') else get_env('APP_WEB_URL')
+
+        # Service API Url prefix.
+        # used to display Service API Base Url to the front-end.
+        self.SERVICE_API_URL = get_env('API_URL') if get_env('API_URL') else get_env('SERVICE_API_URL')
+
+        # File preview or download Url prefix.
+        # used to display File preview or download Url to the front-end or as Multi-model inputs;
+        # Url is signed and has expiration time.
+        self.FILES_URL = get_env('FILES_URL') if get_env('FILES_URL') else self.CONSOLE_API_URL
+
+        # Fallback Url prefix.
+        # Will be deprecated in the future.
+        self.CONSOLE_URL = get_env('CONSOLE_URL')
+        self.API_URL = get_env('API_URL')
+        self.APP_URL = get_env('APP_URL')
+
         # Your App secret key will be used for securely signing the session cookie
         # Make sure you are changing this key for your deployment with a strong key.
         # You can generate a strong key using `openssl rand -base64 42`.
         # Alternatively you can set it with `SECRET_KEY` environment variable.
         self.SECRET_KEY = get_env('SECRET_KEY')
 
-        # redis settings
-        self.REDIS_HOST = get_env('REDIS_HOST')
-        self.REDIS_PORT = get_env('REDIS_PORT')
-        self.REDIS_USERNAME = get_env('REDIS_USERNAME')
-        self.REDIS_PASSWORD = get_env('REDIS_PASSWORD')
-        self.REDIS_DB = get_env('REDIS_DB')
-        self.REDIS_USE_SSL = get_bool_env('REDIS_USE_SSL')
-
-        # storage settings
-        self.STORAGE_TYPE = get_env('STORAGE_TYPE')
-        self.STORAGE_LOCAL_PATH = get_env('STORAGE_LOCAL_PATH')
-        self.S3_ENDPOINT = get_env('S3_ENDPOINT')
-        self.S3_BUCKET_NAME = get_env('S3_BUCKET_NAME')
-        self.S3_ACCESS_KEY = get_env('S3_ACCESS_KEY')
-        self.S3_SECRET_KEY = get_env('S3_SECRET_KEY')
-        self.S3_REGION = get_env('S3_REGION')
-
-        # vector store settings, only support weaviate, qdrant
-        self.VECTOR_STORE = get_env('VECTOR_STORE')
-
-        # weaviate settings
-        self.WEAVIATE_ENDPOINT = get_env('WEAVIATE_ENDPOINT')
-        self.WEAVIATE_API_KEY = get_env('WEAVIATE_API_KEY')
-        self.WEAVIATE_GRPC_ENABLED = get_bool_env('WEAVIATE_GRPC_ENABLED')
-        self.WEAVIATE_BATCH_SIZE = int(get_env('WEAVIATE_BATCH_SIZE'))
-
-        # qdrant settings
-        self.QDRANT_URL = get_env('QDRANT_URL')
-        self.QDRANT_API_KEY = get_env('QDRANT_API_KEY')
-
         # cors settings
         self.CONSOLE_CORS_ALLOW_ORIGINS = get_cors_allow_origins(
             'CONSOLE_CORS_ALLOW_ORIGINS', self.CONSOLE_WEB_URL)
         self.WEB_API_CORS_ALLOW_ORIGINS = get_cors_allow_origins(
             'WEB_API_CORS_ALLOW_ORIGINS', '*')
 
-        # mail settings
-        self.MAIL_TYPE = get_env('MAIL_TYPE')
-        self.MAIL_DEFAULT_SEND_FROM = get_env('MAIL_DEFAULT_SEND_FROM')
-        self.RESEND_API_KEY = get_env('RESEND_API_KEY')
-
-        # sentry settings
-        self.SENTRY_DSN = get_env('SENTRY_DSN')
-        self.SENTRY_TRACES_SAMPLE_RATE = float(get_env('SENTRY_TRACES_SAMPLE_RATE'))
-        self.SENTRY_PROFILES_SAMPLE_RATE = float(get_env('SENTRY_PROFILES_SAMPLE_RATE'))
-
         # check update url
         self.CHECK_UPDATE_URL = get_env('CHECK_UPDATE_URL')
 
-        # database settings
+        # ------------------------
+        # Database Configurations.
+        # ------------------------
         db_credentials = {
             key: get_env(key) for key in
             ['DB_USERNAME', 'DB_PASSWORD', 'DB_HOST', 'DB_PORT', 'DB_DATABASE']
@@ -168,14 +160,107 @@ class Config:
 
         self.SQLALCHEMY_ECHO = get_bool_env('SQLALCHEMY_ECHO')
 
-        # celery settings
+        # ------------------------
+        # Redis Configurations.
+        # ------------------------
+        self.REDIS_HOST = get_env('REDIS_HOST')
+        self.REDIS_PORT = get_env('REDIS_PORT')
+        self.REDIS_USERNAME = get_env('REDIS_USERNAME')
+        self.REDIS_PASSWORD = get_env('REDIS_PASSWORD')
+        self.REDIS_DB = get_env('REDIS_DB')
+        self.REDIS_USE_SSL = get_bool_env('REDIS_USE_SSL')
+
+        # ------------------------
+        # Celery worker Configurations.
+        # ------------------------
         self.CELERY_BROKER_URL = get_env('CELERY_BROKER_URL')
         self.CELERY_BACKEND = get_env('CELERY_BACKEND')
         self.CELERY_RESULT_BACKEND = 'db+{}'.format(self.SQLALCHEMY_DATABASE_URI) \
             if self.CELERY_BACKEND == 'database' else self.CELERY_BROKER_URL
         self.BROKER_USE_SSL = self.CELERY_BROKER_URL.startswith('rediss://')
 
-        # hosted provider credentials
+        # ------------------------
+        # File Storage Configurations.
+        # ------------------------
+        self.STORAGE_TYPE = get_env('STORAGE_TYPE')
+        self.STORAGE_LOCAL_PATH = get_env('STORAGE_LOCAL_PATH')
+        self.S3_ENDPOINT = get_env('S3_ENDPOINT')
+        self.S3_BUCKET_NAME = get_env('S3_BUCKET_NAME')
+        self.S3_ACCESS_KEY = get_env('S3_ACCESS_KEY')
+        self.S3_SECRET_KEY = get_env('S3_SECRET_KEY')
+        self.S3_REGION = get_env('S3_REGION')
+
+        # ------------------------
+        # Vector Store Configurations.
+        # Currently, only support: qdrant, milvus, zilliz, weaviate
+        # ------------------------
+        self.VECTOR_STORE = get_env('VECTOR_STORE')
+
+        # qdrant settings
+        self.QDRANT_URL = get_env('QDRANT_URL')
+        self.QDRANT_API_KEY = get_env('QDRANT_API_KEY')
+
+        # milvus / zilliz setting
+        self.MILVUS_HOST = get_env('MILVUS_HOST')
+        self.MILVUS_PORT = get_env('MILVUS_PORT')
+        self.MILVUS_USER = get_env('MILVUS_USER')
+        self.MILVUS_PASSWORD = get_env('MILVUS_PASSWORD')
+        self.MILVUS_SECURE = get_env('MILVUS_SECURE')
+
+        # weaviate settings
+        self.WEAVIATE_ENDPOINT = get_env('WEAVIATE_ENDPOINT')
+        self.WEAVIATE_API_KEY = get_env('WEAVIATE_API_KEY')
+        self.WEAVIATE_GRPC_ENABLED = get_bool_env('WEAVIATE_GRPC_ENABLED')
+        self.WEAVIATE_BATCH_SIZE = int(get_env('WEAVIATE_BATCH_SIZE'))
+
+        # ------------------------
+        # Mail Configurations.
+        # ------------------------
+        self.MAIL_TYPE = get_env('MAIL_TYPE')
+        self.MAIL_DEFAULT_SEND_FROM = get_env('MAIL_DEFAULT_SEND_FROM')
+        self.RESEND_API_KEY = get_env('RESEND_API_KEY')
+        
+        # ------------------------
+        # Workpace Configurations.
+        # ------------------------
+        self.INVITE_EXPIRY_HOURS = int(get_env('INVITE_EXPIRY_HOURS'))
+
+        # ------------------------
+        # Sentry Configurations.
+        # ------------------------
+        self.SENTRY_DSN = get_env('SENTRY_DSN')
+        self.SENTRY_TRACES_SAMPLE_RATE = float(get_env('SENTRY_TRACES_SAMPLE_RATE'))
+        self.SENTRY_PROFILES_SAMPLE_RATE = float(get_env('SENTRY_PROFILES_SAMPLE_RATE'))
+
+        # ------------------------
+        # Business Configurations.
+        # ------------------------
+
+        # multi model send image format, support base64, url, default is base64
+        self.MULTIMODAL_SEND_IMAGE_FORMAT = get_env('MULTIMODAL_SEND_IMAGE_FORMAT')
+
+        # Dataset Configurations.
+        self.TENANT_DOCUMENT_COUNT = get_env('TENANT_DOCUMENT_COUNT')
+        self.CLEAN_DAY_SETTING = get_env('CLEAN_DAY_SETTING')
+
+        # File upload Configurations.
+        self.UPLOAD_FILE_SIZE_LIMIT = int(get_env('UPLOAD_FILE_SIZE_LIMIT'))
+        self.UPLOAD_FILE_BATCH_LIMIT = int(get_env('UPLOAD_FILE_BATCH_LIMIT'))
+        self.UPLOAD_IMAGE_FILE_SIZE_LIMIT = int(get_env('UPLOAD_IMAGE_FILE_SIZE_LIMIT'))
+
+        # Moderation in app Configurations.
+        self.OUTPUT_MODERATION_BUFFER_SIZE = int(get_env('OUTPUT_MODERATION_BUFFER_SIZE'))
+
+        # Notion integration setting
+        self.NOTION_CLIENT_ID = get_env('NOTION_CLIENT_ID')
+        self.NOTION_CLIENT_SECRET = get_env('NOTION_CLIENT_SECRET')
+        self.NOTION_INTEGRATION_TYPE = get_env('NOTION_INTEGRATION_TYPE')
+        self.NOTION_INTERNAL_SECRET = get_env('NOTION_INTERNAL_SECRET')
+        self.NOTION_INTEGRATION_TOKEN = get_env('NOTION_INTEGRATION_TOKEN')
+
+        # ------------------------
+        # Platform Configurations.
+        # ------------------------
         self.HOSTED_OPENAI_ENABLED = get_bool_env('HOSTED_OPENAI_ENABLED')
         self.HOSTED_OPENAI_API_KEY = get_env('HOSTED_OPENAI_API_KEY')
         self.HOSTED_OPENAI_API_BASE = get_env('HOSTED_OPENAI_API_BASE')
@@ -203,23 +288,6 @@ class Config:
         self.HOSTED_MODERATION_ENABLED = get_bool_env('HOSTED_MODERATION_ENABLED')
         self.HOSTED_MODERATION_PROVIDERS = get_env('HOSTED_MODERATION_PROVIDERS')
 
-        self.STRIPE_API_KEY = get_env('STRIPE_API_KEY')
-        self.STRIPE_WEBHOOK_SECRET = get_env('STRIPE_WEBHOOK_SECRET')
-
-        # notion import setting
-        self.NOTION_CLIENT_ID = get_env('NOTION_CLIENT_ID')
-        self.NOTION_CLIENT_SECRET = get_env('NOTION_CLIENT_SECRET')
-        self.NOTION_INTEGRATION_TYPE = get_env('NOTION_INTEGRATION_TYPE')
-        self.NOTION_INTERNAL_SECRET = get_env('NOTION_INTERNAL_SECRET')
-        self.NOTION_INTEGRATION_TOKEN = get_env('NOTION_INTEGRATION_TOKEN')
-
-        self.TENANT_DOCUMENT_COUNT = get_env('TENANT_DOCUMENT_COUNT')
-        self.CLEAN_DAY_SETTING = get_env('CLEAN_DAY_SETTING')
-
-        # uploading settings
-        self.UPLOAD_FILE_SIZE_LIMIT = int(get_env('UPLOAD_FILE_SIZE_LIMIT'))
-        self.UPLOAD_FILE_BATCH_LIMIT = int(get_env('UPLOAD_FILE_BATCH_LIMIT'))
-
 
 class CloudEditionConfig(Config):
 
@@ -234,18 +302,5 @@ class CloudEditionConfig(Config):
         self.GOOGLE_CLIENT_SECRET = get_env('GOOGLE_CLIENT_SECRET')
         self.OAUTH_REDIRECT_PATH = get_env('OAUTH_REDIRECT_PATH')
 
-
-class TestConfig(Config):
-
-    def __init__(self):
-        super().__init__()
-
-        self.EDITION = "SELF_HOSTED"
-        self.TESTING = True
-
-        db_credentials = {
-            key: get_env(key) for key in ['DB_USERNAME', 'DB_PASSWORD', 'DB_HOST', 'DB_PORT']
-        }
-
-        # use a different database for testing: dify_test
-        self.SQLALCHEMY_DATABASE_URI = f"postgresql://{db_credentials['DB_USERNAME']}:{db_credentials['DB_PASSWORD']}@{db_credentials['DB_HOST']}:{db_credentials['DB_PORT']}/dify_test"
+        self.STRIPE_API_KEY = get_env('STRIPE_API_KEY')
+        self.STRIPE_WEBHOOK_SECRET = get_env('STRIPE_WEBHOOK_SECRET')

api/constants/model_template.py

@@ -31,6 +31,7 @@ model_templates = {
             'model': json.dumps({
                 "provider": "openai",
                 "name": "gpt-3.5-turbo-instruct",
+                "mode": "completion",
                 "completion_params": {
                     "max_tokens": 512,
                     "temperature": 1,
@@ -81,6 +82,7 @@ model_templates = {
             'model': json.dumps({
                 "provider": "openai",
                 "name": "gpt-3.5-turbo",
+                "mode": "chat",
                 "completion_params": {
                     "max_tokens": 512,
                     "temperature": 1,
@@ -137,10 +139,11 @@ demo_model_templates = {
                 },
                 opening_statement='',
                 suggested_questions=None,
-                pre_prompt="Please translate the following text into {{target_language}}:\n",
+                pre_prompt="Please translate the following text into {{target_language}}:\n{{query}}\ntranslate:",
                 model=json.dumps({
                     "provider": "openai",
                     "name": "gpt-3.5-turbo-instruct",
+                    "mode": "completion",
                     "completion_params": {
                         "max_tokens": 1000,
                         "temperature": 0,
@@ -169,6 +172,13 @@ demo_model_templates = {
                                 'Italian',
                             ]
                         }
+                    },{
+                        "paragraph": {
+                            "label": "Query",
+                            "variable": "query",
+                            "required": True,
+                            "default": ""
+                        }
                     }
                 ])
             )
@@ -200,6 +210,7 @@ demo_model_templates = {
                 model=json.dumps({
                     "provider": "openai",
                     "name": "gpt-3.5-turbo",
+                    "mode": "chat",
                     "completion_params": {
                         "max_tokens": 300,
                         "temperature": 0.8,
@@ -255,10 +266,11 @@ demo_model_templates = {
                 },
                 opening_statement='',
                 suggested_questions=None,
-                pre_prompt="请将以下文本翻译为{{target_language}}:\n",
+                pre_prompt="请将以下文本翻译为{{target_language}}:\n{{query}}\n翻译:",
                 model=json.dumps({
                     "provider": "openai",
                     "name": "gpt-3.5-turbo-instruct",
+                    "mode": "completion",
                     "completion_params": {
                         "max_tokens": 1000,
                         "temperature": 0,
@@ -287,6 +299,13 @@ demo_model_templates = {
                                 "意大利语",
                             ]
                         }
+                    },{
+                        "paragraph": {
+                            "label": "文本内容",
+                            "variable": "query",
+                            "required": True,
+                            "default": ""
+                        }
                     }
                 ])
             )
@@ -318,6 +337,7 @@ demo_model_templates = {
                 model=json.dumps({
                     "provider": "openai",
                     "name": "gpt-3.5-turbo",
+                    "mode": "chat",
                     "completion_params": {
                         "max_tokens": 300,
                         "temperature": 0.8,

api/controllers/console/__init__.py

@@ -6,10 +6,10 @@ bp = Blueprint('console', __name__, url_prefix='/console/api')
 api = ExternalApi(bp)
 
 # Import other controllers
-from . import setup, version, apikey, admin
+from . import extension, setup, version, apikey, admin
 
 # Import app controllers
-from .app import app, site, completion, model_config, statistic, conversation, message, generator, audio
+from .app import advanced_prompt_template, app, site, completion, model_config, statistic, conversation, message, generator, audio
 
 # Import auth controllers
 from .auth import login, oauth, data_source_oauth, activate

api/controllers/console/apikey.py

@@ -1,5 +1,5 @@
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 import flask_restful
 from flask_restful import Resource, fields, marshal_with
 from werkzeug.exceptions import Forbidden

api/controllers/console/app/advanced_prompt_template.py

@@ -0,0 +1,25 @@
+from flask_restful import Resource, reqparse
+
+from controllers.console import api
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from libs.login import login_required
+from services.advanced_prompt_template_service import AdvancedPromptTemplateService
+
+class AdvancedPromptTemplateList(Resource):
+    
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+         
+        parser = reqparse.RequestParser()
+        parser.add_argument('app_mode', type=str, required=True, location='args')
+        parser.add_argument('model_mode', type=str, required=True, location='args')
+        parser.add_argument('has_context', type=str, required=False, default='true', location='args')
+        parser.add_argument('model_name', type=str, required=True, location='args')
+        args = parser.parse_args()
+
+        return AdvancedPromptTemplateService.get_prompt(args)
+
+api.add_resource(AdvancedPromptTemplateList, '/app/prompt-templates')

api/controllers/console/app/app.py

@@ -3,10 +3,9 @@ import json
 import logging
 from datetime import datetime
 
-import flask
 from flask_login import current_user
-from core.login.login import login_required
-from flask_restful import Resource, reqparse, fields, marshal_with, abort, inputs
+from libs.login import login_required
+from flask_restful import Resource, reqparse, marshal_with, abort, inputs
 from werkzeug.exceptions import Forbidden
 
 from constants.model_template import model_templates, demo_model_templates
@@ -17,11 +16,9 @@ from controllers.console.wraps import account_initialization_required
 from core.model_providers.error import ProviderTokenNotInitError, LLMBadRequestError
 from core.model_providers.model_factory import ModelFactory
 from core.model_providers.model_provider_factory import ModelProviderFactory
-from core.model_providers.models.entity.model_params import ModelType
 from events.app_event import app_was_created, app_was_deleted
 from fields.app_fields import app_pagination_fields, app_detail_fields, template_list_fields, \
     app_detail_fields_with_site
-from libs.helper import TimestampField
 from extensions.ext_database import db
 from models.model import App, AppModelConfig, Site
 from services.app_model_config_service import AppModelConfigService

api/controllers/console/app/audio.py

@@ -2,8 +2,8 @@
 import logging
 
 from flask import request
-from core.login.login import login_required
-from werkzeug.exceptions import InternalServerError, NotFound
+from libs.login import login_required
+from werkzeug.exceptions import InternalServerError
 
 import services
 from controllers.console import api

api/controllers/console/app/completion.py

@@ -5,7 +5,7 @@ from typing import Generator, Union
 
 import flask_login
 from flask import Response, stream_with_context
-from core.login.login import login_required
+from libs.login import login_required
 from werkzeug.exceptions import InternalServerError, NotFound
 
 import services
@@ -40,12 +40,14 @@ class CompletionMessageApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('model_config', type=dict, required=True, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] != 'blocking'
+        args['auto_generate_name'] = False
 
         account = flask_login.current_user
 
@@ -113,6 +115,7 @@ class ChatMessageApi(Resource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('model_config', type=dict, required=True, location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
@@ -120,6 +123,7 @@ class ChatMessageApi(Resource):
         args = parser.parse_args()
 
         streaming = args['response_mode'] != 'blocking'
+        args['auto_generate_name'] = False
 
         account = flask_login.current_user
 

api/controllers/console/app/conversation.py

@@ -2,8 +2,8 @@ from datetime import datetime
 
 import pytz
 from flask_login import current_user
-from core.login.login import login_required
-from flask_restful import Resource, reqparse, fields, marshal_with
+from libs.login import login_required
+from flask_restful import Resource, reqparse, marshal_with
 from flask_restful.inputs import int_range
 from sqlalchemy import or_, func
 from sqlalchemy.orm import joinedload
@@ -15,7 +15,7 @@ from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
 from fields.conversation_fields import conversation_pagination_fields, conversation_detail_fields, \
     conversation_message_detail_fields, conversation_with_summary_pagination_fields
-from libs.helper import TimestampField, datetime_string, uuid_value
+from libs.helper import datetime_string
 from extensions.ext_database import db
 from models.model import Message, MessageAnnotation, Conversation
 
@@ -108,7 +108,7 @@ class CompletionConversationDetailApi(Resource):
         conversation_id = str(conversation_id)
 
         return _get_conversation(app_id, conversation_id, 'completion')
-    
+
     @setup_required
     @login_required
     @account_initialization_required
@@ -230,7 +230,7 @@ class ChatConversationDetailApi(Resource):
         conversation_id = str(conversation_id)
 
         return _get_conversation(app_id, conversation_id, 'chat')
-    
+
     @setup_required
     @login_required
     @account_initialization_required
@@ -253,8 +253,6 @@ class ChatConversationDetailApi(Resource):
         return {'result': 'success'}, 204
 
 
-
-
 api.add_resource(CompletionConversationApi, '/apps/<uuid:app_id>/completion-conversations')
 api.add_resource(CompletionConversationDetailApi, '/apps/<uuid:app_id>/completion-conversations/<uuid:conversation_id>')
 api.add_resource(ChatConversationApi, '/apps/<uuid:app_id>/chat-conversations')

api/controllers/console/app/generator.py

@@ -1,5 +1,5 @@
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse
 
 from controllers.console import api
@@ -12,35 +12,6 @@ from core.model_providers.error import ProviderTokenNotInitError, QuotaExceededE
     LLMAPIUnavailableError, LLMRateLimitError, LLMAuthorizationError, ModelCurrentlyNotSupportError
 
 
-class IntroductionGenerateApi(Resource):
-    @setup_required
-    @login_required
-    @account_initialization_required
-    def post(self):
-        parser = reqparse.RequestParser()
-        parser.add_argument('prompt_template', type=str, required=True, location='json')
-        args = parser.parse_args()
-
-        account = current_user
-
-        try:
-            answer = LLMGenerator.generate_introduction(
-                account.current_tenant_id,
-                args['prompt_template']
-            )
-        except ProviderTokenNotInitError as ex:
-            raise ProviderNotInitializeError(ex.description)
-        except QuotaExceededError:
-            raise ProviderQuotaExceededError()
-        except ModelCurrentlyNotSupportError:
-            raise ProviderModelCurrentlyNotSupportError()
-        except (LLMBadRequestError, LLMAPIConnectionError, LLMAPIUnavailableError,
-                LLMRateLimitError, LLMAuthorizationError) as e:
-            raise CompletionRequestError(str(e))
-
-        return {'introduction': answer}
-
-
 class RuleGenerateApi(Resource):
     @setup_required
     @login_required
@@ -72,5 +43,4 @@ class RuleGenerateApi(Resource):
         return rules
 
 
-api.add_resource(IntroductionGenerateApi, '/introduction-generate')
 api.add_resource(RuleGenerateApi, '/rule-generate')

api/controllers/console/app/message.py

@@ -16,9 +16,9 @@ from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
 from core.model_providers.error import LLMRateLimitError, LLMBadRequestError, LLMAuthorizationError, LLMAPIConnectionError, \
     ProviderTokenNotInitError, LLMAPIUnavailableError, QuotaExceededError, ModelCurrentlyNotSupportError
-from core.login.login import login_required
+from libs.login import login_required
 from fields.conversation_fields import message_detail_fields
-from libs.helper import uuid_value, TimestampField
+from libs.helper import uuid_value
 from libs.infinite_scroll_pagination import InfiniteScrollPagination
 from extensions.ext_database import db
 from models.model import MessageAnnotation, Conversation, Message, MessageFeedback
@@ -295,8 +295,8 @@ class MessageSuggestedQuestionApi(Resource):
         try:
             questions = MessageService.get_suggested_questions_after_answer(
                 app_model=app_model,
-                user=current_user,
                 message_id=message_id,
+                user=current_user,
                 check_enabled=False
             )
         except MessageNotExistsError:
@@ -329,7 +329,7 @@ class MessageApi(Resource):
         message_id = str(message_id)
 
         # get app info
-        app_model = _get_app(app_id, 'chat')
+        app_model = _get_app(app_id)
 
         message = db.session.query(Message).filter(
             Message.id == message_id,

api/controllers/console/app/model_config.py

@@ -1,5 +1,4 @@
 # -*- coding:utf-8 -*-
-import json
 
 from flask import request
 from flask_restful import Resource
@@ -9,7 +8,7 @@ from controllers.console import api
 from controllers.console.app import _get_app
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
-from core.login.login import login_required
+from libs.login import login_required
 from events.app_event import app_model_config_was_updated
 from extensions.ext_database import db
 from models.model import AppModelConfig

api/controllers/console/app/site.py

@@ -1,7 +1,7 @@
 # -*- coding:utf-8 -*-
 from flask_login import current_user
-from core.login.login import login_required
-from flask_restful import Resource, reqparse, fields, marshal_with
+from libs.login import login_required
+from flask_restful import Resource, reqparse, marshal_with
 from werkzeug.exceptions import NotFound, Forbidden
 
 from controllers.console import api

api/controllers/console/app/statistic.py

@@ -5,7 +5,7 @@ from datetime import datetime
 import pytz
 from flask import jsonify
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse
 
 from controllers.console import api

api/controllers/console/auth/data_source_oauth.py

@@ -1,16 +1,13 @@
 import logging
-from datetime import datetime
-from typing import Optional
 
-import flask_login
 import requests
-from flask import request, redirect, current_app, session
+from flask import request, redirect, current_app
 from flask_login import current_user
 
 from flask_restful import Resource
 from werkzeug.exceptions import Forbidden
 
-from core.login.login import login_required
+from libs.login import login_required
 from libs.oauth_data_source import NotionOAuth
 from controllers.console import api
 from ..setup import setup_required

api/controllers/console/datasets/data_source.py

@@ -1,11 +1,10 @@
 import datetime
 import json
 
-from cachetools import TTLCache
-from flask import request, current_app
+from flask import request
 from flask_login import current_user
-from core.login.login import login_required
-from flask_restful import Resource, marshal_with, fields, reqparse, marshal
+from libs.login import login_required
+from flask_restful import Resource, marshal_with, reqparse
 from werkzeug.exceptions import NotFound
 
 from controllers.console import api
@@ -15,14 +14,11 @@ from core.data_loader.loader.notion import NotionLoader
 from core.indexing_runner import IndexingRunner
 from extensions.ext_database import db
 from fields.data_source_fields import integrate_notion_info_list_fields, integrate_list_fields
-from libs.helper import TimestampField
 from models.dataset import Document
 from models.source import DataSourceBinding
 from services.dataset_service import DatasetService, DocumentService
 from tasks.document_indexing_sync_task import document_indexing_sync_task
 
-cache = TTLCache(maxsize=None, ttl=30)
-
 
 class DataSourceApi(Resource):
 

api/controllers/console/datasets/datasets.py

@@ -4,8 +4,8 @@ from flask import request, current_app
 from flask_login import current_user
 
 from controllers.console.apikey import api_key_list, api_key_fields
-from core.login.login import login_required
-from flask_restful import Resource, reqparse, fields, marshal, marshal_with
+from libs.login import login_required
+from flask_restful import Resource, reqparse, marshal, marshal_with
 from werkzeug.exceptions import NotFound, Forbidden
 import services
 from controllers.console import api
@@ -170,6 +170,7 @@ class DatasetApi(Resource):
                             help='Invalid indexing technique.')
         parser.add_argument('permission', type=str, location='json', choices=(
             'only_me', 'all_team_members'), help='Invalid permission.')
+        parser.add_argument('retrieval_model', type=dict, location='json', help='Invalid retrieval model.')
         args = parser.parse_args()
 
         # The role of the current user in the ta table must be admin or owner
@@ -401,6 +402,7 @@ class DatasetApiKeyApi(Resource):
 
 class DatasetApiDeleteApi(Resource):
     resource_type = 'dataset'
+
     @setup_required
     @login_required
     @account_initialization_required
@@ -436,6 +438,50 @@ class DatasetApiBaseUrlApi(Resource):
         }
 
 
+class DatasetRetrievalSettingApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        vector_type = current_app.config['VECTOR_STORE']
+        if vector_type == 'milvus':
+            return {
+                'retrieval_method': [
+                    'semantic_search'
+                ]
+            }
+        elif vector_type == 'qdrant' or vector_type == 'weaviate':
+            return {
+                'retrieval_method': [
+                    'semantic_search', 'full_text_search', 'hybrid_search'
+                ]
+            }
+        else:
+            raise ValueError("Unsupported vector db type.")
+
+
+class DatasetRetrievalSettingMockApi(Resource):
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self, vector_type):
+
+        if vector_type == 'milvus':
+            return {
+                'retrieval_method': [
+                    'semantic_search'
+                ]
+            }
+        elif vector_type == 'qdrant' or vector_type == 'weaviate':
+            return {
+                'retrieval_method': [
+                    'semantic_search', 'full_text_search', 'hybrid_search'
+                ]
+            }
+        else:
+            raise ValueError("Unsupported vector db type.")
+
+
 api.add_resource(DatasetListApi, '/datasets')
 api.add_resource(DatasetApi, '/datasets/<uuid:dataset_id>')
 api.add_resource(DatasetQueryApi, '/datasets/<uuid:dataset_id>/queries')
@@ -445,3 +491,5 @@ api.add_resource(DatasetIndexingStatusApi, '/datasets/<uuid:dataset_id>/indexing
 api.add_resource(DatasetApiKeyApi, '/datasets/api-keys')
 api.add_resource(DatasetApiDeleteApi, '/datasets/api-keys/<uuid:api_key_id>')
 api.add_resource(DatasetApiBaseUrlApi, '/datasets/api-base-info')
+api.add_resource(DatasetRetrievalSettingApi, '/datasets/retrieval-setting')
+api.add_resource(DatasetRetrievalSettingMockApi, '/datasets/retrieval-setting/<string:vector_type>')

api/controllers/console/datasets/datasets_document.py

@@ -1,11 +1,10 @@
 # -*- coding:utf-8 -*-
-import random
 from datetime import datetime
 from typing import List
 
 from flask import request, current_app
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, fields, marshal, marshal_with, reqparse
 from sqlalchemy import desc, asc
 from werkzeug.exceptions import NotFound, Forbidden
@@ -25,7 +24,6 @@ from core.model_providers.model_factory import ModelFactory
 from extensions.ext_redis import redis_client
 from fields.document_fields import document_with_segments_fields, document_fields, \
     dataset_and_document_fields, document_status_fields
-from libs.helper import TimestampField
 from extensions.ext_database import db
 from models.dataset import DatasetProcessRule, Dataset
 from models.dataset import Document, DocumentSegment
@@ -223,6 +221,8 @@ class DatasetDocumentListApi(Resource):
         parser.add_argument('doc_form', type=str, default='text_model', required=False, nullable=False, location='json')
         parser.add_argument('doc_language', type=str, default='English', required=False, nullable=False,
                             location='json')
+        parser.add_argument('retrieval_model', type=dict, required=False, nullable=False,
+                            location='json')
         args = parser.parse_args()
 
         if not dataset.indexing_technique and not args['indexing_technique']:
@@ -265,6 +265,8 @@ class DatasetInitApi(Resource):
         parser.add_argument('doc_form', type=str, default='text_model', required=False, nullable=False, location='json')
         parser.add_argument('doc_language', type=str, default='English', required=False, nullable=False,
                             location='json')
+        parser.add_argument('retrieval_model', type=dict, required=False, nullable=False,
+                            location='json')
         args = parser.parse_args()
         if args['indexing_technique'] == 'high_quality':
             try:

api/controllers/console/datasets/datasets_segments.py

@@ -14,13 +14,12 @@ from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required
 from core.model_providers.error import LLMBadRequestError, ProviderTokenNotInitError
 from core.model_providers.model_factory import ModelFactory
-from core.login.login import login_required
+from libs.login import login_required
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from fields.segment_fields import segment_fields
 from models.dataset import DocumentSegment
 
-from libs.helper import TimestampField
 from services.dataset_service import DatasetService, DocumentService, SegmentService
 from tasks.enable_segment_to_index_task import enable_segment_to_index_task
 from tasks.disable_segment_from_index_task import disable_segment_from_index_task

api/controllers/console/datasets/file.py

@@ -1,9 +1,9 @@
-from cachetools import TTLCache
 from flask import request, current_app
+from flask_login import current_user
 
 import services
-from core.login.login import login_required
-from flask_restful import Resource, marshal_with, fields
+from libs.login import login_required
+from flask_restful import Resource, marshal_with
 
 from controllers.console import api
 from controllers.console.datasets.error import NoFileUploadedError, TooManyFilesError, FileTooLargeError, \
@@ -15,9 +15,6 @@ from fields.file_fields import upload_config_fields, file_fields
 
 from services.file_service import FileService
 
-cache = TTLCache(maxsize=None, ttl=30)
-
-ALLOWED_EXTENSIONS = ['txt', 'markdown', 'md', 'pdf', 'html', 'htm', 'xlsx', 'docx', 'csv']
 PREVIEW_WORDS_LIMIT = 3000
 
 
@@ -30,9 +27,11 @@ class FileApi(Resource):
     def get(self):
         file_size_limit = current_app.config.get("UPLOAD_FILE_SIZE_LIMIT")
         batch_count_limit = current_app.config.get("UPLOAD_FILE_BATCH_LIMIT")
+        image_file_size_limit = current_app.config.get("UPLOAD_IMAGE_FILE_SIZE_LIMIT")
         return {
             'file_size_limit': file_size_limit,
-            'batch_count_limit': batch_count_limit
+            'batch_count_limit': batch_count_limit,
+            'image_file_size_limit': image_file_size_limit
         }, 200
 
     @setup_required
@@ -51,7 +50,7 @@ class FileApi(Resource):
         if len(request.files) > 1:
             raise TooManyFilesError()
         try:
-            upload_file = FileService.upload_file(file)
+            upload_file = FileService.upload_file(file, current_user)
         except services.errors.file.FileTooLargeError as file_too_large_error:
             raise FileTooLargeError(file_too_large_error.description)
         except services.errors.file.UnsupportedFileTypeError:

api/controllers/console/datasets/hit_testing.py

@@ -1,7 +1,7 @@
 import logging
 
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse, marshal
 from werkzeug.exceptions import InternalServerError, NotFound, Forbidden
 
@@ -42,19 +42,18 @@ class HitTestingApi(Resource):
 
         parser = reqparse.RequestParser()
         parser.add_argument('query', type=str, location='json')
+        parser.add_argument('retrieval_model', type=dict, required=False, location='json')
         args = parser.parse_args()
 
-        query = args['query']
-
-        if not query or len(query) > 250:
-            raise ValueError('Query is required and cannot exceed 250 characters')
+        HitTestingService.hit_testing_args_check(args)
 
         try:
             response = HitTestingService.retrieve(
                 dataset=dataset,
-                query=query,
+                query=args['query'],
                 account=current_user,
-                limit=10,
+                retrieval_model=args['retrieval_model'],
+                limit=10
             )
 
             return {"query": response['query'], 'records': marshal(response['records'], hit_testing_record_fields)}
@@ -68,7 +67,7 @@ class HitTestingApi(Resource):
             raise ProviderModelCurrentlyNotSupportError()
         except LLMBadRequestError:
             raise ProviderNotInitializeError(
-                f"No Embedding Model available. Please configure a valid provider "
+                f"No Embedding Model or Reranking Model available. Please configure a valid provider "
                 f"in the Settings -> Model Provider.")
         except ValueError as e:
             raise ValueError(str(e))

api/controllers/console/explore/completion.py

@@ -1,6 +1,7 @@
 # -*- coding:utf-8 -*-
 import json
 import logging
+from datetime import datetime
 from typing import Generator, Union
 
 from flask import Response, stream_with_context
@@ -17,6 +18,7 @@ from controllers.console.explore.wraps import InstalledAppResource
 from core.conversation_message_task import PubHandler
 from core.model_providers.error import LLMBadRequestError, LLMAPIUnavailableError, LLMAuthorizationError, LLMAPIConnectionError, \
     LLMRateLimitError, ProviderTokenNotInitError, QuotaExceededError, ModelCurrentlyNotSupportError
+from extensions.ext_database import db
 from libs.helper import uuid_value
 from services.completion_service import CompletionService
 
@@ -32,11 +34,16 @@ class CompletionApi(InstalledAppResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='explore_app', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
+
+        installed_app.last_used_at = datetime.utcnow()
+        db.session.commit()
 
         try:
             response = CompletionService.completion(
@@ -91,12 +98,17 @@ class ChatApi(InstalledAppResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='explore_app', location='json')
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
+
+        installed_app.last_used_at = datetime.utcnow()
+        db.session.commit()
 
         try:
             response = CompletionService.completion(

api/controllers/console/explore/conversation.py

@@ -38,7 +38,8 @@ class ConversationListApi(InstalledAppResource):
                 user=current_user,
                 last_id=args['last_id'],
                 limit=args['limit'],
-                pinned=pinned
+                pinned=pinned,
+                exclude_debug_conversation=True
             )
         except LastConversationNotExistsError:
             raise NotFound("Last Conversation Not Exists.")
@@ -71,11 +72,18 @@ class ConversationRenameApi(InstalledAppResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, current_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                current_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/console/explore/installed_app.py

@@ -2,8 +2,8 @@
 from datetime import datetime
 
 from flask_login import current_user
-from core.login.login import login_required
-from flask_restful import Resource, reqparse, fields, marshal_with, inputs
+from libs.login import login_required
+from flask_restful import Resource, reqparse, marshal_with, inputs
 from sqlalchemy import and_
 from werkzeug.exceptions import NotFound, Forbidden, BadRequest
 
@@ -12,7 +12,6 @@ from controllers.console.explore.wraps import InstalledAppResource
 from controllers.console.wraps import account_initialization_required
 from extensions.ext_database import db
 from fields.installed_app_fields import installed_app_list_fields
-from libs.helper import TimestampField
 from models.model import App, InstalledApp, RecommendedApp
 from services.account_service import TenantService
 
@@ -40,8 +39,9 @@ class InstalledAppsListApi(Resource):
             }
             for installed_app in installed_apps
         ]
-        installed_apps.sort(key=lambda app: (-app['is_pinned'], app['last_used_at']
-                            if app['last_used_at'] is not None else datetime.min))
+        installed_apps.sort(key=lambda app: (-app['is_pinned'],
+                                             app['last_used_at'] is None,
+                                             -app['last_used_at'].timestamp() if app['last_used_at'] is not None else 0))
 
         return {'installed_apps': installed_apps}
 

api/controllers/console/explore/parameter.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import marshal_with, fields
+from flask import current_app
 
 from controllers.console import api
 from controllers.console.explore.wraps import InstalledAppResource
@@ -19,6 +20,10 @@ class AppParameterApi(InstalledAppResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -27,6 +32,9 @@ class AppParameterApi(InstalledAppResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -42,7 +50,12 @@ class AppParameterApi(InstalledAppResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/console/explore/recommended_app.py

@@ -1,6 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, fields, marshal_with
 from sqlalchemy import and_
 

api/controllers/console/explore/saved_message.py

@@ -9,6 +9,7 @@ from controllers.console.explore.wraps import InstalledAppResource
 from libs.helper import uuid_value, TimestampField
 from services.errors.message import MessageNotExistsError
 from services.saved_message_service import SavedMessageService
+from fields.conversation_fields import message_file_fields
 
 feedback_fields = {
     'rating': fields.String
@@ -19,6 +20,7 @@ message_fields = {
     'inputs': fields.Raw,
     'query': fields.String,
     'answer': fields.String,
+    'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
     'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
     'created_at': TimestampField
 }

api/controllers/console/explore/wraps.py

@@ -1,5 +1,5 @@
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource
 from functools import wraps
 

api/controllers/console/extension.py

@@ -0,0 +1,114 @@
+from flask_restful import Resource, reqparse, marshal_with
+from flask_login import current_user
+
+from controllers.console import api
+from controllers.console.setup import setup_required
+from controllers.console.wraps import account_initialization_required
+from libs.login import login_required
+from models.api_based_extension import APIBasedExtension
+from fields.api_based_extension_fields import api_based_extension_fields
+from services.code_based_extension_service import CodeBasedExtensionService
+from services.api_based_extension_service import APIBasedExtensionService
+
+
+class CodeBasedExtensionAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def get(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument('module', type=str, required=True, location='args')
+        args = parser.parse_args()
+
+        return {
+            'module': args['module'],
+            'data': CodeBasedExtensionService.get_code_based_extension(args['module'])
+        }
+
+
+class APIBasedExtensionAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def get(self):
+        tenant_id = current_user.current_tenant_id
+        return APIBasedExtensionService.get_all_by_tenant_id(tenant_id)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def post(self):
+        parser = reqparse.RequestParser()
+        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('api_endpoint', type=str, required=True, location='json')
+        parser.add_argument('api_key', type=str, required=True, location='json')
+        args = parser.parse_args()
+
+        extension_data = APIBasedExtension(
+            tenant_id=current_user.current_tenant_id,
+            name=args['name'],
+            api_endpoint=args['api_endpoint'],
+            api_key=args['api_key']
+        )
+
+        return APIBasedExtensionService.save(extension_data)
+
+
+class APIBasedExtensionDetailAPI(Resource):
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def get(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        return APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    @marshal_with(api_based_extension_fields)
+    def post(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        extension_data_from_db = APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+        parser = reqparse.RequestParser()
+        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('api_endpoint', type=str, required=True, location='json')
+        parser.add_argument('api_key', type=str, required=True, location='json')
+        args = parser.parse_args()
+
+        extension_data_from_db.name = args['name']
+        extension_data_from_db.api_endpoint = args['api_endpoint']
+
+        if args['api_key'] != '[__HIDDEN__]':
+            extension_data_from_db.api_key = args['api_key']
+
+        return APIBasedExtensionService.save(extension_data_from_db)
+
+    @setup_required
+    @login_required
+    @account_initialization_required
+    def delete(self, id):
+        api_based_extension_id = str(id)
+        tenant_id = current_user.current_tenant_id
+
+        extension_data_from_db = APIBasedExtensionService.get_with_tenant_id(tenant_id, api_based_extension_id)
+
+        APIBasedExtensionService.delete(extension_data_from_db)
+
+        return {'result': 'success'}
+
+
+api.add_resource(CodeBasedExtensionAPI, '/code-based-extension')
+
+api.add_resource(APIBasedExtensionAPI, '/api-based-extension')
+api.add_resource(APIBasedExtensionDetailAPI, '/api-based-extension/<uuid:id>')

api/controllers/console/universal_chat/chat.py

@@ -25,6 +25,7 @@ class UniversalChatApi(UniversalChatResource):
 
         parser = reqparse.RequestParser()
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('provider', type=str, required=True, location='json')
         parser.add_argument('model', type=str, required=True, location='json')
@@ -60,6 +61,8 @@ class UniversalChatApi(UniversalChatResource):
         del args['model']
         del args['tools']
 
+        args['auto_generate_name'] = False
+
         try:
             response = CompletionService.completion(
                 app_model=app_model,

api/controllers/console/universal_chat/conversation.py

@@ -65,11 +65,18 @@ class UniversalChatConversationRenameApi(UniversalChatResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, current_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                current_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/console/universal_chat/wraps.py

@@ -2,7 +2,7 @@ import json
 from functools import wraps
 
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource
 from controllers.console.setup import setup_required
 from controllers.console.wraps import account_initialization_required

api/controllers/console/workspace/account.py

@@ -4,7 +4,7 @@ from datetime import datetime
 import pytz
 from flask import current_app, request
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse, fields, marshal_with
 
 from services.errors.account import CurrentPasswordIncorrectError as ServiceCurrentPasswordIncorrectError

api/controllers/console/workspace/members.py

@@ -1,7 +1,7 @@
 # -*- coding:utf-8 -*-
 from flask import current_app
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse, marshal_with, abort, fields, marshal
 
 import services

api/controllers/console/workspace/model_providers.py

@@ -1,5 +1,5 @@
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse
 from werkzeug.exceptions import Forbidden
 
@@ -21,8 +21,12 @@ class ModelProviderListApi(Resource):
     def get(self):
         tenant_id = current_user.current_tenant_id
 
+        parser = reqparse.RequestParser()
+        parser.add_argument('model_type', type=str, required=False, nullable=True, location='args')
+        args = parser.parse_args()
+
         provider_service = ProviderService()
-        provider_list = provider_service.get_provider_list(tenant_id)
+        provider_list = provider_service.get_provider_list(tenant_id=tenant_id, model_type=args.get('model_type'))
 
         return provider_list
 

api/controllers/console/workspace/models.py

@@ -1,5 +1,7 @@
+import logging
+
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse
 
 from controllers.console import api
@@ -19,7 +21,7 @@ class DefaultModelApi(Resource):
     def get(self):
         parser = reqparse.RequestParser()
         parser.add_argument('model_type', type=str, required=True, nullable=False,
-                            choices=['text-generation', 'embeddings', 'speech2text'], location='args')
+                            choices=['text-generation', 'embeddings', 'speech2text', 'reranking'], location='args')
         args = parser.parse_args()
 
         tenant_id = current_user.current_tenant_id
@@ -71,19 +73,21 @@ class DefaultModelApi(Resource):
     @account_initialization_required
     def post(self):
         parser = reqparse.RequestParser()
-        parser.add_argument('model_name', type=str, required=True, nullable=False, location='json')
-        parser.add_argument('model_type', type=str, required=True, nullable=False,
-                            choices=['text-generation', 'embeddings', 'speech2text'], location='json')
-        parser.add_argument('provider_name', type=str, required=True, nullable=False, location='json')
+        parser.add_argument('model_settings', type=list, required=True, nullable=False, location='json')
         args = parser.parse_args()
 
         provider_service = ProviderService()
-        provider_service.update_default_model_of_model_type(
-            tenant_id=current_user.current_tenant_id,
-            model_type=args['model_type'],
-            provider_name=args['provider_name'],
-            model_name=args['model_name']
-        )
+        model_settings = args['model_settings']
+        for model_setting in model_settings:
+            try:
+                provider_service.update_default_model_of_model_type(
+                    tenant_id=current_user.current_tenant_id,
+                    model_type=model_setting['model_type'],
+                    provider_name=model_setting['provider_name'],
+                    model_name=model_setting['model_name']
+                )
+            except Exception:
+                logging.warning(f"{model_setting['model_type']} save error")
 
         return {'result': 'success'}
 

api/controllers/console/workspace/providers.py

@@ -1,6 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, reqparse
 from werkzeug.exceptions import Forbidden
 

api/controllers/console/workspace/tool_providers.py

@@ -1,7 +1,7 @@
 import json
 
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, abort, reqparse
 from werkzeug.exceptions import Forbidden
 

api/controllers/console/workspace/workspace.py

@@ -3,9 +3,8 @@ import logging
 
 from flask import request
 from flask_login import current_user
-from core.login.login import login_required
+from libs.login import login_required
 from flask_restful import Resource, fields, marshal_with, reqparse, marshal, inputs
-from flask_restful.inputs import int_range
 
 from controllers.console import api
 from controllers.console.admin import admin_required

api/controllers/files/__init__.py

@@ -0,0 +1,10 @@
+# -*- coding:utf-8 -*-
+from flask import Blueprint
+
+from libs.external_api import ExternalApi
+
+bp = Blueprint('files', __name__)
+api = ExternalApi(bp)
+
+
+from . import image_preview

api/controllers/files/image_preview.py

@@ -0,0 +1,40 @@
+from flask import request, Response
+from flask_restful import Resource
+
+import services
+from controllers.files import api
+from libs.exception import BaseHTTPException
+from services.file_service import FileService
+
+
+class ImagePreviewApi(Resource):
+    def get(self, file_id):
+        file_id = str(file_id)
+
+        timestamp = request.args.get('timestamp')
+        nonce = request.args.get('nonce')
+        sign = request.args.get('sign')
+
+        if not timestamp or not nonce or not sign:
+            return {'content': 'Invalid request.'}, 400
+
+        try:
+            generator, mimetype = FileService.get_image_preview(
+                file_id,
+                timestamp,
+                nonce,
+                sign
+            )
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return Response(generator, mimetype=mimetype)
+
+
+api.add_resource(ImagePreviewApi, '/files/<uuid:file_id>/image-preview')
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/service_api/__init__.py

@@ -7,6 +7,6 @@ bp = Blueprint('service_api', __name__, url_prefix='/v1')
 api = ExternalApi(bp)
 
 
-from .app import completion, app, conversation, message, audio
+from .app import completion, app, conversation, message, audio, file
 
 from .dataset import document, segment, dataset

api/controllers/service_api/app/app.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import fields, marshal_with
+from flask import current_app
 
 from controllers.service_api import api
 from controllers.service_api.wraps import AppApiResource
@@ -20,6 +21,10 @@ class AppParameterApi(AppApiResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -28,6 +33,9 @@ class AppParameterApi(AppApiResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -42,7 +50,12 @@ class AppParameterApi(AppApiResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/service_api/app/completion.py

@@ -28,6 +28,7 @@ class CompletionApi(AppApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('user', type=str, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
@@ -39,13 +40,15 @@ class CompletionApi(AppApiResource):
         if end_user is None and args['user'] is not None:
             end_user = create_or_update_end_user_for_user_id(app_model, args['user'])
 
+        args['auto_generate_name'] = False
+
         try:
             response = CompletionService.completion(
                 app_model=app_model,
                 user=end_user,
                 args=args,
                 from_source='api',
-                streaming=streaming
+                streaming=streaming,
             )
 
             return compact_response(response)
@@ -90,10 +93,12 @@ class ChatApi(AppApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('user', type=str, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='dev', location='json')
+        parser.add_argument('auto_generate_name', type=bool, required=False, default='True', location='json')
 
         args = parser.parse_args()
 
@@ -183,4 +188,3 @@ api.add_resource(CompletionApi, '/completion-messages')
 api.add_resource(CompletionStopApi, '/completion-messages/<string:task_id>/stop')
 api.add_resource(ChatApi, '/chat-messages')
 api.add_resource(ChatStopApi, '/chat-messages/<string:task_id>/stop')
-

api/controllers/service_api/app/conversation.py

@@ -54,6 +54,7 @@ class ConversationDetailApi(AppApiResource):
             raise NotFound("Conversation Not Exists.")
         return {"result": "success"}, 204
 
+
 class ConversationRenameApi(AppApiResource):
 
     @marshal_with(simple_conversation_fields)
@@ -64,15 +65,22 @@ class ConversationRenameApi(AppApiResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
         parser.add_argument('user', type=str, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default=False, location='json')
         args = parser.parse_args()
 
         if end_user is None and args['user'] is not None:
             end_user = create_or_update_end_user_for_user_id(app_model, args['user'])
 
         try:
-            return ConversationService.rename(app_model, conversation_id, end_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                end_user,
+                args['name'],
+                args['auto_generate']
+            )
         except services.errors.conversation.ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/service_api/app/error.py

@@ -75,3 +75,26 @@ class ProviderNotSupportSpeechToTextError(BaseHTTPException):
     description = "Provider not support speech to text."
     code = 400
 
+
+class NoFileUploadedError(BaseHTTPException):
+    error_code = 'no_file_uploaded'
+    description = "Please upload your file."
+    code = 400
+
+
+class TooManyFilesError(BaseHTTPException):
+    error_code = 'too_many_files'
+    description = "Only one file is allowed."
+    code = 400
+
+
+class FileTooLargeError(BaseHTTPException):
+    error_code = 'file_too_large'
+    description = "File size exceeded. {message}"
+    code = 413
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/service_api/app/file.py

@@ -0,0 +1,45 @@
+from flask import request
+from flask_restful import marshal_with
+
+from controllers.service_api import api
+from controllers.service_api.wraps import AppApiResource
+from controllers.service_api.app import create_or_update_end_user_for_user_id
+from controllers.service_api.app.error import NoFileUploadedError, TooManyFilesError, FileTooLargeError, \
+    UnsupportedFileTypeError
+import services
+from services.file_service import FileService
+from fields.file_fields import file_fields
+
+
+class FileApi(AppApiResource):
+
+    @marshal_with(file_fields)
+    def post(self, app_model, end_user):
+
+        file = request.files['file']
+        user_args = request.form.get('user')
+
+        if end_user is None and user_args is not None:
+            end_user = create_or_update_end_user_for_user_id(app_model, user_args)
+
+        # check file
+        if 'file' not in request.files:
+            raise NoFileUploadedError()
+
+        if not file.mimetype:
+            raise UnsupportedFileTypeError()
+
+        if len(request.files) > 1:
+            raise TooManyFilesError()
+
+        try:
+            upload_file = FileService.upload_file(file, end_user)
+        except services.errors.file.FileTooLargeError as file_too_large_error:
+            raise FileTooLargeError(file_too_large_error.description)
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return upload_file, 201
+
+
+api.add_resource(FileApi, '/files/upload')

api/controllers/service_api/app/message.py

@@ -10,7 +10,9 @@ from controllers.service_api.app.error import NotChatAppError
 from controllers.service_api.wraps import AppApiResource
 from libs.helper import TimestampField, uuid_value
 from services.message_service import MessageService
-
+from extensions.ext_database import db
+from models.model import Message, EndUser
+from fields.conversation_fields import message_file_fields
 
 class MessageListApi(AppApiResource):
     feedback_fields = {
@@ -41,6 +43,7 @@ class MessageListApi(AppApiResource):
         'inputs': fields.Raw,
         'query': fields.String,
         'answer': fields.String,
+        'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
         'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
         'retriever_resources': fields.List(fields.Nested(retriever_resource_fields)),
         'created_at': TimestampField
@@ -96,5 +99,38 @@ class MessageFeedbackApi(AppApiResource):
         return {'result': 'success'}
 
 
+class MessageSuggestedApi(AppApiResource):
+    def get(self, app_model, end_user, message_id):
+        message_id = str(message_id)
+        if app_model.mode != 'chat':
+            raise NotChatAppError()
+        try:
+            message = db.session.query(Message).filter(
+                Message.id == message_id,
+                Message.app_id == app_model.id,
+            ).first()
+
+            if end_user is None and message.from_end_user_id is not None:
+                user = db.session.query(EndUser) \
+                    .filter(
+                        EndUser.tenant_id == app_model.tenant_id,
+                        EndUser.id == message.from_end_user_id,
+                        EndUser.type == 'service_api'
+                    ).first()
+            else:
+                user = end_user
+            questions = MessageService.get_suggested_questions_after_answer(
+                app_model=app_model,
+                user=user,
+                message_id=message_id,
+                check_enabled=False
+            )
+        except services.errors.message.MessageNotExistsError:
+            raise NotFound("Message Not Exists.")
+
+        return {'result': 'success', 'data': questions}
+
+
 api.add_resource(MessageListApi, '/messages')
 api.add_resource(MessageFeedbackApi, '/messages/<uuid:message_id>/feedbacks')
+api.add_resource(MessageSuggestedApi, '/messages/<uuid:message_id>/suggested')

api/controllers/service_api/dataset/dataset.py

@@ -4,12 +4,9 @@ import services.dataset_service
 from controllers.service_api import api
 from controllers.service_api.dataset.error import DatasetNameDuplicateError
 from controllers.service_api.wraps import DatasetApiResource
-from core.login.login import current_user
+from libs.login import current_user
 from core.model_providers.models.entity.model_params import ModelType
-from extensions.ext_database import db
 from fields.dataset_fields import dataset_detail_fields
-from models.account import Account, TenantAccountJoin
-from models.dataset import Dataset
 from services.dataset_service import DatasetService
 from services.provider_service import ProviderService
 

api/controllers/service_api/dataset/document.py

@@ -1,9 +1,8 @@
-import datetime
 import json
-import uuid
 
-from flask import current_app, request
+from flask import request
 from flask_restful import reqparse, marshal
+from flask_login import current_user
 from sqlalchemy import desc
 from werkzeug.exceptions import NotFound
 
@@ -13,13 +12,11 @@ from controllers.service_api.app.error import ProviderNotInitializeError
 from controllers.service_api.dataset.error import ArchivedDocumentImmutableError, DocumentIndexingError, \
     NoFileUploadedError, TooManyFilesError
 from controllers.service_api.wraps import DatasetApiResource
-from core.login.login import current_user
+from libs.login import current_user
 from core.model_providers.error import ProviderTokenNotInitError
 from extensions.ext_database import db
-from extensions.ext_storage import storage
 from fields.document_fields import document_fields, document_status_fields
 from models.dataset import Dataset, Document, DocumentSegment
-from models.model import UploadFile
 from services.dataset_service import DocumentService
 from services.file_service import FileService
 
@@ -39,6 +36,8 @@ class DocumentAddByTextApi(DatasetApiResource):
                             location='json')
         parser.add_argument('indexing_technique', type=str, choices=Dataset.INDEXING_TECHNIQUE_LIST, nullable=False,
                             location='json')
+        parser.add_argument('retrieval_model', type=dict, required=False, nullable=False,
+                            location='json')
         args = parser.parse_args()
         dataset_id = str(dataset_id)
         tenant_id = str(tenant_id)
@@ -98,6 +97,8 @@ class DocumentUpdateByTextApi(DatasetApiResource):
         parser.add_argument('doc_form', type=str, default='text_model', required=False, nullable=False, location='json')
         parser.add_argument('doc_language', type=str, default='English', required=False, nullable=False,
                             location='json')
+        parser.add_argument('retrieval_model', type=dict, required=False, nullable=False,
+                            location='json')
         args = parser.parse_args()
         dataset_id = str(dataset_id)
         tenant_id = str(tenant_id)
@@ -177,7 +178,7 @@ class DocumentAddByFileApi(DatasetApiResource):
         if len(request.files) > 1:
             raise TooManyFilesError()
 
-        upload_file = FileService.upload_file(file)
+        upload_file = FileService.upload_file(file, current_user)
         data_source = {
             'type': 'upload_file',
             'info_list': {
@@ -239,7 +240,7 @@ class DocumentUpdateByFileApi(DatasetApiResource):
             if len(request.files) > 1:
                 raise TooManyFilesError()
 
-            upload_file = FileService.upload_file(file)
+            upload_file = FileService.upload_file(file, current_user)
             data_source = {
                 'type': 'upload_file',
                 'info_list': {

api/controllers/service_api/dataset/segment.py

@@ -1,7 +1,6 @@
 from flask_login import current_user
 from flask_restful import reqparse, marshal
 from werkzeug.exceptions import NotFound
-
 from controllers.service_api import api
 from controllers.service_api.app.error import ProviderNotInitializeError
 from controllers.service_api.wraps import DatasetApiResource
@@ -9,8 +8,8 @@ from core.model_providers.error import ProviderTokenNotInitError, LLMBadRequestE
 from core.model_providers.model_factory import ModelFactory
 from extensions.ext_database import db
 from fields.segment_fields import segment_fields
-from models.dataset import Dataset
-from services.dataset_service import DocumentService, SegmentService
+from models.dataset import Dataset, DocumentSegment
+from services.dataset_service import DatasetService, DocumentService, SegmentService
 
 
 class SegmentApi(DatasetApiResource):
@@ -24,6 +23,8 @@ class SegmentApi(DatasetApiResource):
             Dataset.tenant_id == tenant_id,
             Dataset.id == dataset_id
         ).first()
+        if not dataset:
+            raise NotFound('Dataset not found.')
         # check document
         document_id = str(document_id)
         document = DocumentService.get_document(dataset.id, document_id)
@@ -55,5 +56,146 @@ class SegmentApi(DatasetApiResource):
             'doc_form': document.doc_form
         }, 200
 
+    def get(self, tenant_id, dataset_id, document_id):
+        """Create single segment."""
+        # check dataset
+        dataset_id = str(dataset_id)
+        tenant_id = str(tenant_id)
+        dataset = db.session.query(Dataset).filter(
+            Dataset.tenant_id == tenant_id,
+            Dataset.id == dataset_id
+        ).first()
+        if not dataset:
+            raise NotFound('Dataset not found.')
+        # check document
+        document_id = str(document_id)
+        document = DocumentService.get_document(dataset.id, document_id)
+        if not document:
+            raise NotFound('Document not found.')
+        # check embedding model setting
+        if dataset.indexing_technique == 'high_quality':
+            try:
+                ModelFactory.get_embedding_model(
+                    tenant_id=current_user.current_tenant_id,
+                    model_provider_name=dataset.embedding_model_provider,
+                    model_name=dataset.embedding_model
+                )
+            except LLMBadRequestError:
+                raise ProviderNotInitializeError(
+                    f"No Embedding Model available. Please configure a valid provider "
+                    f"in the Settings -> Model Provider.")
+            except ProviderTokenNotInitError as ex:
+                raise ProviderNotInitializeError(ex.description)
+
+        parser = reqparse.RequestParser()
+        parser.add_argument('status', type=str,
+                            action='append', default=[], location='args')
+        parser.add_argument('keyword', type=str, default=None, location='args')
+        args = parser.parse_args()
+
+        status_list = args['status']
+        keyword = args['keyword']
+
+        query = DocumentSegment.query.filter(
+            DocumentSegment.document_id == str(document_id),
+            DocumentSegment.tenant_id == current_user.current_tenant_id
+        )
+
+        if status_list:
+            query = query.filter(DocumentSegment.status.in_(status_list))
+
+        if keyword:
+            query = query.where(DocumentSegment.content.ilike(f'%{keyword}%'))
+
+        total = query.count()
+        segments = query.order_by(DocumentSegment.position).all()
+        return {
+            'data': marshal(segments, segment_fields),
+            'doc_form': document.doc_form,
+            'total': total
+        }, 200
+
+
+class DatasetSegmentApi(DatasetApiResource):
+    def delete(self, tenant_id, dataset_id, document_id, segment_id):
+        # check dataset
+        dataset_id = str(dataset_id)
+        tenant_id = str(tenant_id)
+        dataset = db.session.query(Dataset).filter(
+            Dataset.tenant_id == tenant_id,
+            Dataset.id == dataset_id
+        ).first()
+        if not dataset:
+            raise NotFound('Dataset not found.')
+        # check user's model setting
+        DatasetService.check_dataset_model_setting(dataset)
+        # check document
+        document_id = str(document_id)
+        document = DocumentService.get_document(dataset_id, document_id)
+        if not document:
+            raise NotFound('Document not found.')
+        # check segment
+        segment = DocumentSegment.query.filter(
+            DocumentSegment.id == str(segment_id),
+            DocumentSegment.tenant_id == current_user.current_tenant_id
+        ).first()
+        if not segment:
+            raise NotFound('Segment not found.')
+        SegmentService.delete_segment(segment, document, dataset)
+        return {'result': 'success'}, 200
+
+    def post(self, tenant_id, dataset_id, document_id, segment_id):
+        # check dataset
+        dataset_id = str(dataset_id)
+        tenant_id = str(tenant_id)
+        dataset = db.session.query(Dataset).filter(
+            Dataset.tenant_id == tenant_id,
+            Dataset.id == dataset_id
+        ).first()
+        if not dataset:
+            raise NotFound('Dataset not found.')
+        # check user's model setting
+        DatasetService.check_dataset_model_setting(dataset)
+        # check document
+        document_id = str(document_id)
+        document = DocumentService.get_document(dataset_id, document_id)
+        if not document:
+            raise NotFound('Document not found.')
+        if dataset.indexing_technique == 'high_quality':
+            # check embedding model setting
+            try:
+                ModelFactory.get_embedding_model(
+                    tenant_id=current_user.current_tenant_id,
+                    model_provider_name=dataset.embedding_model_provider,
+                    model_name=dataset.embedding_model
+                )
+            except LLMBadRequestError:
+                raise ProviderNotInitializeError(
+                    f"No Embedding Model available. Please configure a valid provider "
+                    f"in the Settings -> Model Provider.")
+            except ProviderTokenNotInitError as ex:
+                raise ProviderNotInitializeError(ex.description)
+            # check segment
+        segment_id = str(segment_id)
+        segment = DocumentSegment.query.filter(
+            DocumentSegment.id == str(segment_id),
+            DocumentSegment.tenant_id == current_user.current_tenant_id
+        ).first()
+        if not segment:
+            raise NotFound('Segment not found.')
+
+        # validate args
+        parser = reqparse.RequestParser()
+        parser.add_argument('segments', type=dict, required=False, nullable=True, location='json')
+        args = parser.parse_args()
+
+        SegmentService.segment_create_args_validate(args['segments'], document)
+        segment = SegmentService.update_segment(args['segments'], segment, document, dataset)
+        return {
+            'data': marshal(segment, segment_fields),
+            'doc_form': document.doc_form
+        }, 200
+
 
 api.add_resource(SegmentApi, '/datasets/<uuid:dataset_id>/documents/<uuid:document_id>/segments')
+api.add_resource(DatasetSegmentApi, '/datasets/<uuid:dataset_id>/documents/<uuid:document_id>/segments/<uuid:segment_id>')

api/controllers/service_api/wraps.py

@@ -7,10 +7,9 @@ from flask_login import user_logged_in
 from flask_restful import Resource
 from werkzeug.exceptions import NotFound, Unauthorized
 
-from core.login.login import _get_user
+from libs.login import _get_user
 from extensions.ext_database import db
 from models.account import Tenant, TenantAccountJoin, Account
-from models.dataset import Dataset
 from models.model import ApiToken, App
 
 

api/controllers/web/__init__.py

@@ -7,4 +7,4 @@ bp = Blueprint('web', __name__, url_prefix='/api')
 api = ExternalApi(bp)
 
 
-from . import completion, app, conversation, message, site, saved_message, audio, passport
+from . import completion, app, conversation, message, site, saved_message, audio, passport, file

api/controllers/web/app.py

@@ -1,5 +1,6 @@
 # -*- coding:utf-8 -*-
 from flask_restful import marshal_with, fields
+from flask import current_app
 
 from controllers.web import api
 from controllers.web.wraps import WebApiResource
@@ -19,6 +20,10 @@ class AppParameterApi(WebApiResource):
         'options': fields.List(fields.String)
     }
 
+    system_parameters_fields = {
+        'image_file_size_limit': fields.String
+    }
+
     parameters_fields = {
         'opening_statement': fields.String,
         'suggested_questions': fields.Raw,
@@ -27,6 +32,9 @@ class AppParameterApi(WebApiResource):
         'retriever_resource': fields.Raw,
         'more_like_this': fields.Raw,
         'user_input_form': fields.Raw,
+        'sensitive_word_avoidance': fields.Raw,
+        'file_upload': fields.Raw,
+        'system_parameters': fields.Nested(system_parameters_fields)
     }
 
     @marshal_with(parameters_fields)
@@ -41,7 +49,12 @@ class AppParameterApi(WebApiResource):
             'speech_to_text': app_model_config.speech_to_text_dict,
             'retriever_resource': app_model_config.retriever_resource_dict,
             'more_like_this': app_model_config.more_like_this_dict,
-            'user_input_form': app_model_config.user_input_form_list
+            'user_input_form': app_model_config.user_input_form_list,
+            'sensitive_word_avoidance': app_model_config.sensitive_word_avoidance_dict,
+            'file_upload': app_model_config.file_upload_dict,
+            'system_parameters': {
+                'image_file_size_limit': current_app.config.get('UPLOAD_IMAGE_FILE_SIZE_LIMIT')
+            }
         }
 
 

api/controllers/web/completion.py

@@ -30,12 +30,14 @@ class CompletionApi(WebApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, location='json', default='')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='web_app', location='json')
 
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
 
         try:
             response = CompletionService.completion(
@@ -88,6 +90,7 @@ class ChatApi(WebApiResource):
         parser = reqparse.RequestParser()
         parser.add_argument('inputs', type=dict, required=True, location='json')
         parser.add_argument('query', type=str, required=True, location='json')
+        parser.add_argument('files', type=list, required=False, location='json')
         parser.add_argument('response_mode', type=str, choices=['blocking', 'streaming'], location='json')
         parser.add_argument('conversation_id', type=uuid_value, location='json')
         parser.add_argument('retriever_from', type=str, required=False, default='web_app', location='json')
@@ -95,6 +98,7 @@ class ChatApi(WebApiResource):
         args = parser.parse_args()
 
         streaming = args['response_mode'] == 'streaming'
+        args['auto_generate_name'] = False
 
         try:
             response = CompletionService.completion(
@@ -139,7 +143,7 @@ class ChatStopApi(WebApiResource):
         return {'result': 'success'}, 200
 
 
-def compact_response(response: Union[dict | Generator]) -> Response:
+def compact_response(response: Union[dict, Generator]) -> Response:
     if isinstance(response, dict):
         return Response(response=json.dumps(response), status=200, mimetype='application/json')
     else:

api/controllers/web/conversation.py

@@ -67,11 +67,18 @@ class ConversationRenameApi(WebApiResource):
         conversation_id = str(c_id)
 
         parser = reqparse.RequestParser()
-        parser.add_argument('name', type=str, required=True, location='json')
+        parser.add_argument('name', type=str, required=False, location='json')
+        parser.add_argument('auto_generate', type=bool, required=False, default='False', location='json')
         args = parser.parse_args()
 
         try:
-            return ConversationService.rename(app_model, conversation_id, end_user, args['name'])
+            return ConversationService.rename(
+                app_model,
+                conversation_id,
+                end_user,
+                args['name'],
+                args['auto_generate']
+            )
         except ConversationNotExistsError:
             raise NotFound("Conversation Not Exists.")
 

api/controllers/web/error.py

@@ -85,4 +85,28 @@ class UnsupportedAudioTypeError(BaseHTTPException):
 class ProviderNotSupportSpeechToTextError(BaseHTTPException):
     error_code = 'provider_not_support_speech_to_text'
     description = "Provider not support speech to text."
-    code = 400
+    code = 400
+
+
+class NoFileUploadedError(BaseHTTPException):
+    error_code = 'no_file_uploaded'
+    description = "Please upload your file."
+    code = 400
+
+
+class TooManyFilesError(BaseHTTPException):
+    error_code = 'too_many_files'
+    description = "Only one file is allowed."
+    code = 400
+
+
+class FileTooLargeError(BaseHTTPException):
+    error_code = 'file_too_large'
+    description = "File size exceeded. {message}"
+    code = 413
+
+
+class UnsupportedFileTypeError(BaseHTTPException):
+    error_code = 'unsupported_file_type'
+    description = "File type not allowed."
+    code = 415

api/controllers/web/file.py

@@ -0,0 +1,36 @@
+from flask import request
+from flask_restful import marshal_with
+
+from controllers.web import api
+from controllers.web.wraps import WebApiResource
+from controllers.web.error import NoFileUploadedError, TooManyFilesError, FileTooLargeError, \
+    UnsupportedFileTypeError
+import services
+from services.file_service import FileService
+from fields.file_fields import file_fields
+
+
+class FileApi(WebApiResource):
+
+    @marshal_with(file_fields)
+    def post(self, app_model, end_user):
+        # get file from request
+        file = request.files['file']
+
+        # check file
+        if 'file' not in request.files:
+            raise NoFileUploadedError()
+
+        if len(request.files) > 1:
+            raise TooManyFilesError()
+        try:
+            upload_file = FileService.upload_file(file, end_user)
+        except services.errors.file.FileTooLargeError as file_too_large_error:
+            raise FileTooLargeError(file_too_large_error.description)
+        except services.errors.file.UnsupportedFileTypeError:
+            raise UnsupportedFileTypeError()
+
+        return upload_file, 201
+
+
+api.add_resource(FileApi, '/files/upload')

api/controllers/web/message.py

@@ -22,6 +22,7 @@ from services.errors.app import MoreLikeThisDisabledError
 from services.errors.conversation import ConversationNotExistsError
 from services.errors.message import MessageNotExistsError, SuggestedQuestionsAfterAnswerDisabledError
 from services.message_service import MessageService
+from fields.conversation_fields import message_file_fields
 
 
 class MessageListApi(WebApiResource):
@@ -54,6 +55,7 @@ class MessageListApi(WebApiResource):
         'inputs': fields.Raw,
         'query': fields.String,
         'answer': fields.String,
+        'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
         'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
         'retriever_resources': fields.List(fields.Nested(retriever_resource_fields)),
         'created_at': TimestampField
@@ -115,7 +117,7 @@ class MessageMoreLikeThisApi(WebApiResource):
         streaming = args['response_mode'] == 'streaming'
 
         try:
-            response = CompletionService.generate_more_like_this(app_model, end_user, message_id, streaming)
+            response = CompletionService.generate_more_like_this(app_model, end_user, message_id, streaming, 'web_app')
             return compact_response(response)
         except MessageNotExistsError:
             raise NotFound("Message Not Exists.")

api/controllers/web/saved_message.py

@@ -8,6 +8,8 @@ from controllers.web.wraps import WebApiResource
 from libs.helper import uuid_value, TimestampField
 from services.errors.message import MessageNotExistsError
 from services.saved_message_service import SavedMessageService
+from fields.conversation_fields import message_file_fields
+
 
 feedback_fields = {
     'rating': fields.String
@@ -18,6 +20,7 @@ message_fields = {
     'inputs': fields.Raw,
     'query': fields.String,
     'answer': fields.String,
+    'message_files': fields.List(fields.Nested(message_file_fields), attribute='files'),
     'feedback': fields.Nested(feedback_fields, attribute='user_feedback', allow_null=True),
     'created_at': TimestampField
 }

api/core/__init__.py

@@ -0,0 +1 @@
+import core.moderation.base

api/core/agent/agent/multi_dataset_router_agent.py

@@ -2,15 +2,18 @@ import json
 from typing import Tuple, List, Any, Union, Sequence, Optional, cast
 
 from langchain.agents import OpenAIFunctionsAgent, BaseSingleActionAgent
+from langchain.agents.openai_functions_agent.base import _format_intermediate_steps, _parse_ai_message
 from langchain.callbacks.base import BaseCallbackManager
 from langchain.callbacks.manager import Callbacks
 from langchain.prompts.chat import BaseMessagePromptTemplate
-from langchain.schema import AgentAction, AgentFinish, SystemMessage
+from langchain.schema import AgentAction, AgentFinish, SystemMessage, Generation, LLMResult, AIMessage
 from langchain.schema.language_model import BaseLanguageModel
 from langchain.tools import BaseTool
+from pydantic import root_validator
 
+from core.model_providers.models.entity.message import to_prompt_messages
 from core.model_providers.models.llm.base import BaseLLM
-from core.tool.dataset_retriever_tool import DatasetRetrieverTool
+from core.third_party.langchain.llms.fake import FakeLLM
 
 
 class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
@@ -24,6 +27,10 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
 
         arbitrary_types_allowed = True
 
+    @root_validator
+    def validate_llm(cls, values: dict) -> dict:
+        return values
+
     def should_use_agent(self, query: str):
         """
         return should use agent
@@ -52,7 +59,6 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
             return AgentFinish(return_values={"output": ''}, log='')
         elif len(self.tools) == 1:
             tool = next(iter(self.tools))
-            tool = cast(DatasetRetrieverTool, tool)
             rst = tool.run(tool_input={'query': kwargs['input']})
             # output = ''
             # rst_json = json.loads(rst)
@@ -65,17 +71,57 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
             return AgentFinish(return_values={"output": observation}, log=observation)
 
         try:
-            agent_decision = super().plan(intermediate_steps, callbacks, **kwargs)
+            agent_decision = self.real_plan(intermediate_steps, callbacks, **kwargs)
             if isinstance(agent_decision, AgentAction):
                 tool_inputs = agent_decision.tool_input
-                if isinstance(tool_inputs, dict) and 'query' in tool_inputs:
+                if isinstance(tool_inputs, dict) and 'query' in tool_inputs and 'chat_history' not in kwargs:
                     tool_inputs['query'] = kwargs['input']
                     agent_decision.tool_input = tool_inputs
+            else:
+                agent_decision.return_values['output'] = ''
             return agent_decision
         except Exception as e:
             new_exception = self.model_instance.handle_exceptions(e)
             raise new_exception
 
+    def real_plan(
+        self,
+        intermediate_steps: List[Tuple[AgentAction, str]],
+        callbacks: Callbacks = None,
+        **kwargs: Any,
+    ) -> Union[AgentAction, AgentFinish]:
+        """Given input, decided what to do.
+
+        Args:
+            intermediate_steps: Steps the LLM has taken to date, along with observations
+            **kwargs: User inputs.
+
+        Returns:
+            Action specifying what tool to use.
+        """
+        agent_scratchpad = _format_intermediate_steps(intermediate_steps)
+        selected_inputs = {
+            k: kwargs[k] for k in self.prompt.input_variables if k != "agent_scratchpad"
+        }
+        full_inputs = dict(**selected_inputs, agent_scratchpad=agent_scratchpad)
+        prompt = self.prompt.format_prompt(**full_inputs)
+        messages = prompt.to_messages()
+        prompt_messages = to_prompt_messages(messages)
+        result = self.model_instance.run(
+            messages=prompt_messages,
+            functions=self.functions,
+        )
+
+        ai_message = AIMessage(
+            content=result.content,
+            additional_kwargs={
+                'function_call': result.function_call
+            }
+        )
+
+        agent_decision = _parse_ai_message(ai_message)
+        return agent_decision
+
     async def aplan(
             self,
             intermediate_steps: List[Tuple[AgentAction, str]],
@@ -87,7 +133,7 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
     @classmethod
     def from_llm_and_tools(
             cls,
-            llm: BaseLanguageModel,
+            model_instance: BaseLLM,
             tools: Sequence[BaseTool],
             callback_manager: Optional[BaseCallbackManager] = None,
             extra_prompt_messages: Optional[List[BaseMessagePromptTemplate]] = None,
@@ -96,11 +142,15 @@ class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
             ),
             **kwargs: Any,
     ) -> BaseSingleActionAgent:
-        return super().from_llm_and_tools(
-            llm=llm,
-            tools=tools,
-            callback_manager=callback_manager,
+        prompt = cls.create_prompt(
             extra_prompt_messages=extra_prompt_messages,
             system_message=system_message,
+        )
+        return cls(
+            model_instance=model_instance,
+            llm=FakeLLM(response=''),
+            prompt=prompt,
+            tools=tools,
+            callback_manager=callback_manager,
             **kwargs,
         )

api/core/agent/agent/openai_function_call.py

@@ -5,21 +5,40 @@ from langchain.agents.openai_functions_agent.base import _parse_ai_message, \
     _format_intermediate_steps
 from langchain.callbacks.base import BaseCallbackManager
 from langchain.callbacks.manager import Callbacks
+from langchain.chat_models.openai import _convert_message_to_dict, _import_tiktoken
+from langchain.memory.prompt import SUMMARY_PROMPT
 from langchain.prompts.chat import BaseMessagePromptTemplate
-from langchain.schema import AgentAction, AgentFinish, SystemMessage
-from langchain.schema.language_model import BaseLanguageModel
+from langchain.schema import AgentAction, AgentFinish, SystemMessage, AIMessage, HumanMessage, BaseMessage, \
+    get_buffer_string
 from langchain.tools import BaseTool
+from pydantic import root_validator
 
-from core.agent.agent.calc_token_mixin import ExceededLLMTokensLimitError
-from core.agent.agent.openai_function_call_summarize_mixin import OpenAIFunctionCallSummarizeMixin
+from core.agent.agent.calc_token_mixin import ExceededLLMTokensLimitError, CalcTokenMixin
+from core.chain.llm_chain import LLMChain
+from core.model_providers.models.entity.message import to_prompt_messages
+from core.model_providers.models.llm.base import BaseLLM
+from core.third_party.langchain.llms.fake import FakeLLM
 
 
-class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, OpenAIFunctionCallSummarizeMixin):
+class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, CalcTokenMixin):
+    moving_summary_buffer: str = ""
+    moving_summary_index: int = 0
+    summary_model_instance: BaseLLM = None
+    model_instance: BaseLLM
+
+    class Config:
+        """Configuration for this pydantic object."""
+
+        arbitrary_types_allowed = True
+
+    @root_validator
+    def validate_llm(cls, values: dict) -> dict:
+        return values
 
     @classmethod
     def from_llm_and_tools(
             cls,
-            llm: BaseLanguageModel,
+            model_instance: BaseLLM,
             tools: Sequence[BaseTool],
             callback_manager: Optional[BaseCallbackManager] = None,
             extra_prompt_messages: Optional[List[BaseMessagePromptTemplate]] = None,
@@ -28,12 +47,16 @@ class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, OpenAIFunctio
             ),
             **kwargs: Any,
     ) -> BaseSingleActionAgent:
-        return super().from_llm_and_tools(
-            llm=llm,
+        prompt = cls.create_prompt(
+            extra_prompt_messages=extra_prompt_messages,
+            system_message=system_message,
+        )
+        return cls(
+            model_instance=model_instance,
+            llm=FakeLLM(response=''),
+            prompt=prompt,
             tools=tools,
             callback_manager=callback_manager,
-            extra_prompt_messages=extra_prompt_messages,
-            system_message=cls.get_system_message(),
             **kwargs,
         )
 
@@ -44,23 +67,26 @@ class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, OpenAIFunctio
         :param query:
         :return:
         """
-        original_max_tokens = self.llm.max_tokens
-        self.llm.max_tokens = 40
+        original_max_tokens = self.model_instance.model_kwargs.max_tokens
+        self.model_instance.model_kwargs.max_tokens = 40
 
         prompt = self.prompt.format_prompt(input=query, agent_scratchpad=[])
         messages = prompt.to_messages()
 
         try:
-            predicted_message = self.llm.predict_messages(
-                messages, functions=self.functions, callbacks=None
+            prompt_messages = to_prompt_messages(messages)
+            result = self.model_instance.run(
+                messages=prompt_messages,
+                functions=self.functions,
+                callbacks=None
             )
         except Exception as e:
             new_exception = self.model_instance.handle_exceptions(e)
             raise new_exception
 
-        function_call = predicted_message.additional_kwargs.get("function_call", {})
+        function_call = result.function_call
 
-        self.llm.max_tokens = original_max_tokens
+        self.model_instance.model_kwargs.max_tokens = original_max_tokens
 
         return True if function_call else False
 
@@ -93,10 +119,19 @@ class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, OpenAIFunctio
         except ExceededLLMTokensLimitError as e:
             return AgentFinish(return_values={"output": str(e)}, log=str(e))
 
-        predicted_message = self.llm.predict_messages(
-            messages, functions=self.functions, callbacks=callbacks
+        prompt_messages = to_prompt_messages(messages)
+        result = self.model_instance.run(
+            messages=prompt_messages,
+            functions=self.functions,
         )
-        agent_decision = _parse_ai_message(predicted_message)
+
+        ai_message = AIMessage(
+            content=result.content,
+            additional_kwargs={
+                'function_call': result.function_call
+            }
+        )
+        agent_decision = _parse_ai_message(ai_message)
 
         if isinstance(agent_decision, AgentAction) and agent_decision.tool == 'dataset':
             tool_inputs = agent_decision.tool_input
@@ -122,3 +157,142 @@ class AutoSummarizingOpenAIFunctionCallAgent(OpenAIFunctionsAgent, OpenAIFunctio
             return super().return_stopped_response(early_stopping_method, intermediate_steps, **kwargs)
         except ValueError:
             return AgentFinish({"output": "I'm sorry, I don't know how to respond to that."}, "")
+
+    def summarize_messages_if_needed(self, messages: List[BaseMessage], **kwargs) -> List[BaseMessage]:
+        # calculate rest tokens and summarize previous function observation messages if rest_tokens < 0
+        rest_tokens = self.get_message_rest_tokens(self.model_instance, messages, **kwargs)
+        rest_tokens = rest_tokens - 20  # to deal with the inaccuracy of rest_tokens
+        if rest_tokens >= 0:
+            return messages
+
+        system_message = None
+        human_message = None
+        should_summary_messages = []
+        for message in messages:
+            if isinstance(message, SystemMessage):
+                system_message = message
+            elif isinstance(message, HumanMessage):
+                human_message = message
+            else:
+                should_summary_messages.append(message)
+
+        if len(should_summary_messages) > 2:
+            ai_message = should_summary_messages[-2]
+            function_message = should_summary_messages[-1]
+            should_summary_messages = should_summary_messages[self.moving_summary_index:-2]
+            self.moving_summary_index = len(should_summary_messages)
+        else:
+            error_msg = "Exceeded LLM tokens limit, stopped."
+            raise ExceededLLMTokensLimitError(error_msg)
+
+        new_messages = [system_message, human_message]
+
+        if self.moving_summary_index == 0:
+            should_summary_messages.insert(0, human_message)
+
+        self.moving_summary_buffer = self.predict_new_summary(
+            messages=should_summary_messages,
+            existing_summary=self.moving_summary_buffer
+        )
+
+        new_messages.append(AIMessage(content=self.moving_summary_buffer))
+        new_messages.append(ai_message)
+        new_messages.append(function_message)
+
+        return new_messages
+
+    def predict_new_summary(
+        self, messages: List[BaseMessage], existing_summary: str
+    ) -> str:
+        new_lines = get_buffer_string(
+            messages,
+            human_prefix="Human",
+            ai_prefix="AI",
+        )
+
+        chain = LLMChain(model_instance=self.summary_model_instance, prompt=SUMMARY_PROMPT)
+        return chain.predict(summary=existing_summary, new_lines=new_lines)
+
+    def get_num_tokens_from_messages(self, model_instance: BaseLLM, messages: List[BaseMessage], **kwargs) -> int:
+        """Calculate num tokens for gpt-3.5-turbo and gpt-4 with tiktoken package.
+
+        Official documentation: https://github.com/openai/openai-cookbook/blob/
+        main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb"""
+        if model_instance.model_provider.provider_name == 'azure_openai':
+            model = model_instance.base_model_name
+            model = model.replace("gpt-35", "gpt-3.5")
+        else:
+            model = model_instance.base_model_name
+
+        tiktoken_ = _import_tiktoken()
+        try:
+            encoding = tiktoken_.encoding_for_model(model)
+        except KeyError:
+            model = "cl100k_base"
+            encoding = tiktoken_.get_encoding(model)
+
+        if model.startswith("gpt-3.5-turbo"):
+            # every message follows <im_start>{role/name}\n{content}<im_end>\n
+            tokens_per_message = 4
+            # if there's a name, the role is omitted
+            tokens_per_name = -1
+        elif model.startswith("gpt-4"):
+            tokens_per_message = 3
+            tokens_per_name = 1
+        else:
+            raise NotImplementedError(
+                f"get_num_tokens_from_messages() is not presently implemented "
+                f"for model {model}."
+                "See https://github.com/openai/openai-python/blob/main/chatml.md for "
+                "information on how messages are converted to tokens."
+            )
+        num_tokens = 0
+        for m in messages:
+            message = _convert_message_to_dict(m)
+            num_tokens += tokens_per_message
+            for key, value in message.items():
+                if key == "function_call":
+                    for f_key, f_value in value.items():
+                        num_tokens += len(encoding.encode(f_key))
+                        num_tokens += len(encoding.encode(f_value))
+                else:
+                    num_tokens += len(encoding.encode(value))
+
+                if key == "name":
+                    num_tokens += tokens_per_name
+        # every reply is primed with <im_start>assistant
+        num_tokens += 3
+
+        if kwargs.get('functions'):
+            for function in kwargs.get('functions'):
+                num_tokens += len(encoding.encode('name'))
+                num_tokens += len(encoding.encode(function.get("name")))
+                num_tokens += len(encoding.encode('description'))
+                num_tokens += len(encoding.encode(function.get("description")))
+                parameters = function.get("parameters")
+                num_tokens += len(encoding.encode('parameters'))
+                if 'title' in parameters:
+                    num_tokens += len(encoding.encode('title'))
+                    num_tokens += len(encoding.encode(parameters.get("title")))
+                num_tokens += len(encoding.encode('type'))
+                num_tokens += len(encoding.encode(parameters.get("type")))
+                if 'properties' in parameters:
+                    num_tokens += len(encoding.encode('properties'))
+                    for key, value in parameters.get('properties').items():
+                        num_tokens += len(encoding.encode(key))
+                        for field_key, field_value in value.items():
+                            num_tokens += len(encoding.encode(field_key))
+                            if field_key == 'enum':
+                                for enum_field in field_value:
+                                    num_tokens += 3
+                                    num_tokens += len(encoding.encode(enum_field))
+                            else:
+                                num_tokens += len(encoding.encode(field_key))
+                                num_tokens += len(encoding.encode(str(field_value)))
+                if 'required' in parameters:
+                    num_tokens += len(encoding.encode('required'))
+                    for required_field in parameters['required']:
+                        num_tokens += 3
+                        num_tokens += len(encoding.encode(required_field))
+
+        return num_tokens

api/core/agent/agent/openai_function_call_summarize_mixin.py

@@ -1,140 +0,0 @@
-from typing import cast, List
-
-from langchain.chat_models import ChatOpenAI
-from langchain.chat_models.openai import _convert_message_to_dict
-from langchain.memory.summary import SummarizerMixin
-from langchain.schema import SystemMessage, HumanMessage, BaseMessage, AIMessage
-from langchain.schema.language_model import BaseLanguageModel
-from pydantic import BaseModel
-
-from core.agent.agent.calc_token_mixin import ExceededLLMTokensLimitError, CalcTokenMixin
-from core.model_providers.models.llm.base import BaseLLM
-
-
-class OpenAIFunctionCallSummarizeMixin(BaseModel, CalcTokenMixin):
-    moving_summary_buffer: str = ""
-    moving_summary_index: int = 0
-    summary_llm: BaseLanguageModel = None
-    model_instance: BaseLLM
-
-    class Config:
-        """Configuration for this pydantic object."""
-
-        arbitrary_types_allowed = True
-
-    def summarize_messages_if_needed(self, messages: List[BaseMessage], **kwargs) -> List[BaseMessage]:
-        # calculate rest tokens and summarize previous function observation messages if rest_tokens < 0
-        rest_tokens = self.get_message_rest_tokens(self.model_instance, messages, **kwargs)
-        rest_tokens = rest_tokens - 20  # to deal with the inaccuracy of rest_tokens
-        if rest_tokens >= 0:
-            return messages
-
-        system_message = None
-        human_message = None
-        should_summary_messages = []
-        for message in messages:
-            if isinstance(message, SystemMessage):
-                system_message = message
-            elif isinstance(message, HumanMessage):
-                human_message = message
-            else:
-                should_summary_messages.append(message)
-
-        if len(should_summary_messages) > 2:
-            ai_message = should_summary_messages[-2]
-            function_message = should_summary_messages[-1]
-            should_summary_messages = should_summary_messages[self.moving_summary_index:-2]
-            self.moving_summary_index = len(should_summary_messages)
-        else:
-            error_msg = "Exceeded LLM tokens limit, stopped."
-            raise ExceededLLMTokensLimitError(error_msg)
-
-        new_messages = [system_message, human_message]
-
-        if self.moving_summary_index == 0:
-            should_summary_messages.insert(0, human_message)
-
-        summary_handler = SummarizerMixin(llm=self.summary_llm)
-        self.moving_summary_buffer = summary_handler.predict_new_summary(
-            messages=should_summary_messages,
-            existing_summary=self.moving_summary_buffer
-        )
-
-        new_messages.append(AIMessage(content=self.moving_summary_buffer))
-        new_messages.append(ai_message)
-        new_messages.append(function_message)
-
-        return new_messages
-
-    def get_num_tokens_from_messages(self, model_instance: BaseLLM, messages: List[BaseMessage], **kwargs) -> int:
-        """Calculate num tokens for gpt-3.5-turbo and gpt-4 with tiktoken package.
-
-        Official documentation: https://github.com/openai/openai-cookbook/blob/
-        main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb"""
-        llm = cast(ChatOpenAI, model_instance.client)
-        model, encoding = llm._get_encoding_model()
-        if model.startswith("gpt-3.5-turbo"):
-            # every message follows <im_start>{role/name}\n{content}<im_end>\n
-            tokens_per_message = 4
-            # if there's a name, the role is omitted
-            tokens_per_name = -1
-        elif model.startswith("gpt-4"):
-            tokens_per_message = 3
-            tokens_per_name = 1
-        else:
-            raise NotImplementedError(
-                f"get_num_tokens_from_messages() is not presently implemented "
-                f"for model {model}."
-                "See https://github.com/openai/openai-python/blob/main/chatml.md for "
-                "information on how messages are converted to tokens."
-            )
-        num_tokens = 0
-        for m in messages:
-            message = _convert_message_to_dict(m)
-            num_tokens += tokens_per_message
-            for key, value in message.items():
-                if key == "function_call":
-                    for f_key, f_value in value.items():
-                        num_tokens += len(encoding.encode(f_key))
-                        num_tokens += len(encoding.encode(f_value))
-                else:
-                    num_tokens += len(encoding.encode(value))
-
-                if key == "name":
-                    num_tokens += tokens_per_name
-        # every reply is primed with <im_start>assistant
-        num_tokens += 3
-
-        if kwargs.get('functions'):
-            for function in kwargs.get('functions'):
-                num_tokens += len(encoding.encode('name'))
-                num_tokens += len(encoding.encode(function.get("name")))
-                num_tokens += len(encoding.encode('description'))
-                num_tokens += len(encoding.encode(function.get("description")))
-                parameters = function.get("parameters")
-                num_tokens += len(encoding.encode('parameters'))
-                if 'title' in parameters:
-                    num_tokens += len(encoding.encode('title'))
-                    num_tokens += len(encoding.encode(parameters.get("title")))
-                num_tokens += len(encoding.encode('type'))
-                num_tokens += len(encoding.encode(parameters.get("type")))
-                if 'properties' in parameters:
-                    num_tokens += len(encoding.encode('properties'))
-                    for key, value in parameters.get('properties').items():
-                        num_tokens += len(encoding.encode(key))
-                        for field_key, field_value in value.items():
-                            num_tokens += len(encoding.encode(field_key))
-                            if field_key == 'enum':
-                                for enum_field in field_value:
-                                    num_tokens += 3
-                                    num_tokens += len(encoding.encode(enum_field))
-                            else:
-                                num_tokens += len(encoding.encode(field_key))
-                                num_tokens += len(encoding.encode(str(field_value)))
-                if 'required' in parameters:
-                    num_tokens += len(encoding.encode('required'))
-                    for required_field in parameters['required']:
-                        num_tokens += 3
-                        num_tokens += len(encoding.encode(required_field))
-
-        return num_tokens

api/core/agent/agent/openai_multi_function_call.py

@@ -1,107 +0,0 @@
-from typing import List, Tuple, Any, Union, Sequence, Optional
-
-from langchain.agents import BaseMultiActionAgent
-from langchain.agents.openai_functions_multi_agent.base import OpenAIMultiFunctionsAgent, _format_intermediate_steps, \
-    _parse_ai_message
-from langchain.callbacks.base import BaseCallbackManager
-from langchain.callbacks.manager import Callbacks
-from langchain.prompts.chat import BaseMessagePromptTemplate
-from langchain.schema import AgentAction, AgentFinish, SystemMessage
-from langchain.schema.language_model import BaseLanguageModel
-from langchain.tools import BaseTool
-
-from core.agent.agent.calc_token_mixin import ExceededLLMTokensLimitError
-from core.agent.agent.openai_function_call_summarize_mixin import OpenAIFunctionCallSummarizeMixin
-
-
-class AutoSummarizingOpenMultiAIFunctionCallAgent(OpenAIMultiFunctionsAgent, OpenAIFunctionCallSummarizeMixin):
-
-    @classmethod
-    def from_llm_and_tools(
-            cls,
-            llm: BaseLanguageModel,
-            tools: Sequence[BaseTool],
-            callback_manager: Optional[BaseCallbackManager] = None,
-            extra_prompt_messages: Optional[List[BaseMessagePromptTemplate]] = None,
-            system_message: Optional[SystemMessage] = SystemMessage(
-                content="You are a helpful AI assistant."
-            ),
-            **kwargs: Any,
-    ) -> BaseMultiActionAgent:
-        return super().from_llm_and_tools(
-            llm=llm,
-            tools=tools,
-            callback_manager=callback_manager,
-            extra_prompt_messages=extra_prompt_messages,
-            system_message=cls.get_system_message(),
-            **kwargs,
-        )
-
-    def should_use_agent(self, query: str):
-        """
-        return should use agent
-
-        :param query:
-        :return:
-        """
-        original_max_tokens = self.llm.max_tokens
-        self.llm.max_tokens = 15
-
-        prompt = self.prompt.format_prompt(input=query, agent_scratchpad=[])
-        messages = prompt.to_messages()
-
-        try:
-            predicted_message = self.llm.predict_messages(
-                messages, functions=self.functions, callbacks=None
-            )
-        except Exception as e:
-            new_exception = self.model_instance.handle_exceptions(e)
-            raise new_exception
-
-        function_call = predicted_message.additional_kwargs.get("function_call", {})
-
-        self.llm.max_tokens = original_max_tokens
-
-        return True if function_call else False
-
-    def plan(
-            self,
-            intermediate_steps: List[Tuple[AgentAction, str]],
-            callbacks: Callbacks = None,
-            **kwargs: Any,
-    ) -> Union[AgentAction, AgentFinish]:
-        """Given input, decided what to do.
-
-        Args:
-            intermediate_steps: Steps the LLM has taken to date, along with observations
-            **kwargs: User inputs.
-
-        Returns:
-            Action specifying what tool to use.
-        """
-        agent_scratchpad = _format_intermediate_steps(intermediate_steps)
-        selected_inputs = {
-            k: kwargs[k] for k in self.prompt.input_variables if k != "agent_scratchpad"
-        }
-        full_inputs = dict(**selected_inputs, agent_scratchpad=agent_scratchpad)
-        prompt = self.prompt.format_prompt(**full_inputs)
-        messages = prompt.to_messages()
-
-        # summarize messages if rest_tokens < 0
-        try:
-            messages = self.summarize_messages_if_needed(messages, functions=self.functions)
-        except ExceededLLMTokensLimitError as e:
-            return AgentFinish(return_values={"output": str(e)}, log=str(e))
-
-        predicted_message = self.llm.predict_messages(
-            messages, functions=self.functions, callbacks=callbacks
-        )
-        agent_decision = _parse_ai_message(predicted_message)
-        return agent_decision
-
-    @classmethod
-    def get_system_message(cls):
-        # get current time
-        return SystemMessage(content="You are a helpful AI assistant.\n"
-                                     "The current date or current time you know is wrong.\n"
-                                     "Respond directly if appropriate.")

api/core/agent/agent/output_parser/retirver_dataset_agent.py

@@ -0,0 +1,158 @@
+import json
+from typing import Tuple, List, Any, Union, Sequence, Optional, cast
+
+from langchain.agents import OpenAIFunctionsAgent, BaseSingleActionAgent
+from langchain.agents.openai_functions_agent.base import _format_intermediate_steps, _parse_ai_message
+from langchain.callbacks.base import BaseCallbackManager
+from langchain.callbacks.manager import Callbacks
+from langchain.prompts.chat import BaseMessagePromptTemplate
+from langchain.schema import AgentAction, AgentFinish, SystemMessage, Generation, LLMResult, AIMessage
+from langchain.schema.language_model import BaseLanguageModel
+from langchain.tools import BaseTool
+from pydantic import root_validator
+
+from core.model_providers.models.entity.message import to_prompt_messages
+from core.model_providers.models.llm.base import BaseLLM
+from core.third_party.langchain.llms.fake import FakeLLM
+from core.tool.dataset_retriever_tool import DatasetRetrieverTool
+
+
+class MultiDatasetRouterAgent(OpenAIFunctionsAgent):
+    """
+    An Multi Dataset Retrieve Agent driven by Router.
+    """
+    model_instance: BaseLLM
+
+    class Config:
+        """Configuration for this pydantic object."""
+
+        arbitrary_types_allowed = True
+
+    @root_validator
+    def validate_llm(cls, values: dict) -> dict:
+        return values
+
+    def should_use_agent(self, query: str):
+        """
+        return should use agent
+
+        :param query:
+        :return:
+        """
+        return True
+
+    def plan(
+        self,
+        intermediate_steps: List[Tuple[AgentAction, str]],
+        callbacks: Callbacks = None,
+        **kwargs: Any,
+    ) -> Union[AgentAction, AgentFinish]:
+        """Given input, decided what to do.
+
+        Args:
+            intermediate_steps: Steps the LLM has taken to date, along with observations
+            **kwargs: User inputs.
+
+        Returns:
+            Action specifying what tool to use.
+        """
+        if len(self.tools) == 0:
+            return AgentFinish(return_values={"output": ''}, log='')
+        elif len(self.tools) == 1:
+            tool = next(iter(self.tools))
+            tool = cast(DatasetRetrieverTool, tool)
+            rst = tool.run(tool_input={'query': kwargs['input']})
+            # output = ''
+            # rst_json = json.loads(rst)
+            # for item in rst_json:
+            #     output += f'{item["content"]}\n'
+            return AgentFinish(return_values={"output": rst}, log=rst)
+
+        if intermediate_steps:
+            _, observation = intermediate_steps[-1]
+            return AgentFinish(return_values={"output": observation}, log=observation)
+
+        try:
+            agent_decision = self.real_plan(intermediate_steps, callbacks, **kwargs)
+            if isinstance(agent_decision, AgentAction):
+                tool_inputs = agent_decision.tool_input
+                if isinstance(tool_inputs, dict) and 'query' in tool_inputs and 'chat_history' not in kwargs:
+                    tool_inputs['query'] = kwargs['input']
+                    agent_decision.tool_input = tool_inputs
+            else:
+                agent_decision.return_values['output'] = ''
+            return agent_decision
+        except Exception as e:
+            new_exception = self.model_instance.handle_exceptions(e)
+            raise new_exception
+
+    def real_plan(
+        self,
+        intermediate_steps: List[Tuple[AgentAction, str]],
+        callbacks: Callbacks = None,
+        **kwargs: Any,
+    ) -> Union[AgentAction, AgentFinish]:
+        """Given input, decided what to do.
+
+        Args:
+            intermediate_steps: Steps the LLM has taken to date, along with observations
+            **kwargs: User inputs.
+
+        Returns:
+            Action specifying what tool to use.
+        """
+        agent_scratchpad = _format_intermediate_steps(intermediate_steps)
+        selected_inputs = {
+            k: kwargs[k] for k in self.prompt.input_variables if k != "agent_scratchpad"
+        }
+        full_inputs = dict(**selected_inputs, agent_scratchpad=agent_scratchpad)
+        prompt = self.prompt.format_prompt(**full_inputs)
+        messages = prompt.to_messages()
+        prompt_messages = to_prompt_messages(messages)
+        result = self.model_instance.run(
+            messages=prompt_messages,
+            functions=self.functions,
+        )
+
+        ai_message = AIMessage(
+            content=result.content,
+            additional_kwargs={
+                'function_call': result.function_call
+            }
+        )
+
+        agent_decision = _parse_ai_message(ai_message)
+        return agent_decision
+
+    async def aplan(
+            self,
+            intermediate_steps: List[Tuple[AgentAction, str]],
+            callbacks: Callbacks = None,
+            **kwargs: Any,
+    ) -> Union[AgentAction, AgentFinish]:
+        raise NotImplementedError()
+
+    @classmethod
+    def from_llm_and_tools(
+            cls,
+            model_instance: BaseLLM,
+            tools: Sequence[BaseTool],
+            callback_manager: Optional[BaseCallbackManager] = None,
+            extra_prompt_messages: Optional[List[BaseMessagePromptTemplate]] = None,
+            system_message: Optional[SystemMessage] = SystemMessage(
+                content="You are a helpful AI assistant."
+            ),
+            **kwargs: Any,
+    ) -> BaseSingleActionAgent:
+        prompt = cls.create_prompt(
+            extra_prompt_messages=extra_prompt_messages,
+            system_message=system_message,
+        )
+        return cls(
+            model_instance=model_instance,
+            llm=FakeLLM(response=''),
+            prompt=prompt,
+            tools=tools,
+            callback_manager=callback_manager,
+            **kwargs,
+        )

api/core/agent/agent/structed_multi_dataset_router_agent.py

@@ -1,10 +1,9 @@
 import re
 from typing import List, Tuple, Any, Union, Sequence, Optional, cast
 
-from langchain import BasePromptTemplate
+from langchain import BasePromptTemplate, PromptTemplate
 from langchain.agents import StructuredChatAgent, AgentOutputParser, Agent
 from langchain.agents.structured_chat.base import HUMAN_MESSAGE_TEMPLATE
-from langchain.base_language import BaseLanguageModel
 from langchain.callbacks.base import BaseCallbackManager
 from langchain.callbacks.manager import Callbacks
 from langchain.prompts import SystemMessagePromptTemplate, HumanMessagePromptTemplate, ChatPromptTemplate
@@ -12,6 +11,8 @@ from langchain.schema import AgentAction, AgentFinish, OutputParserException
 from langchain.tools import BaseTool
 from langchain.agents.structured_chat.prompt import PREFIX, SUFFIX
 
+from core.chain.llm_chain import LLMChain
+from core.model_providers.models.entity.model_params import ModelMode
 from core.model_providers.models.llm.base import BaseLLM
 from core.tool.dataset_retriever_tool import DatasetRetrieverTool
 
@@ -49,7 +50,6 @@ Action:
 
 
 class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
-    model_instance: BaseLLM
     dataset_tools: Sequence[BaseTool]
 
     class Config:
@@ -89,16 +89,19 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
             return AgentFinish(return_values={"output": ''}, log='')
         elif len(self.dataset_tools) == 1:
             tool = next(iter(self.dataset_tools))
-            tool = cast(DatasetRetrieverTool, tool)
             rst = tool.run(tool_input={'query': kwargs['input']})
             return AgentFinish(return_values={"output": rst}, log=rst)
 
+        if intermediate_steps:
+            _, observation = intermediate_steps[-1]
+            return AgentFinish(return_values={"output": observation}, log=observation)
+
         full_inputs = self.get_full_inputs(intermediate_steps, **kwargs)
 
         try:
             full_output = self.llm_chain.predict(callbacks=callbacks, **full_inputs)
         except Exception as e:
-            new_exception = self.model_instance.handle_exceptions(e)
+            new_exception = self.llm_chain.model_instance.handle_exceptions(e)
             raise new_exception
 
         try:
@@ -108,6 +111,10 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
                 if isinstance(tool_inputs, dict) and 'query' in tool_inputs:
                     tool_inputs['query'] = kwargs['input']
                     agent_decision.tool_input = tool_inputs
+                elif isinstance(tool_inputs, str):
+                    agent_decision.tool_input = kwargs['input']
+            else:
+                agent_decision.return_values['output'] = ''
             return agent_decision
         except OutputParserException:
             return AgentFinish({"output": "I'm sorry, the answer of model is invalid, "
@@ -142,10 +149,65 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
         ]
         return ChatPromptTemplate(input_variables=input_variables, messages=messages)
 
+    @classmethod
+    def create_completion_prompt(
+            cls,
+            tools: Sequence[BaseTool],
+            prefix: str = PREFIX,
+            format_instructions: str = FORMAT_INSTRUCTIONS,
+            input_variables: Optional[List[str]] = None,
+    ) -> PromptTemplate:
+        """Create prompt in the style of the zero shot agent.
+
+        Args:
+            tools: List of tools the agent will have access to, used to format the
+                prompt.
+            prefix: String to put before the list of tools.
+            input_variables: List of input variables the final prompt will expect.
+
+        Returns:
+            A PromptTemplate with the template assembled from the pieces here.
+        """
+        suffix = """Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
+Question: {input}
+Thought: {agent_scratchpad}
+"""
+
+        tool_strings = "\n".join([f"{tool.name}: {tool.description}" for tool in tools])
+        tool_names = ", ".join([tool.name for tool in tools])
+        format_instructions = format_instructions.format(tool_names=tool_names)
+        template = "\n\n".join([prefix, tool_strings, format_instructions, suffix])
+        if input_variables is None:
+            input_variables = ["input", "agent_scratchpad"]
+        return PromptTemplate(template=template, input_variables=input_variables)
+
+    def _construct_scratchpad(
+        self, intermediate_steps: List[Tuple[AgentAction, str]]
+    ) -> str:
+        agent_scratchpad = ""
+        for action, observation in intermediate_steps:
+            agent_scratchpad += action.log
+            agent_scratchpad += f"\n{self.observation_prefix}{observation}\n{self.llm_prefix}"
+
+        if not isinstance(agent_scratchpad, str):
+            raise ValueError("agent_scratchpad should be of type string.")
+        if agent_scratchpad:
+            llm_chain = cast(LLMChain, self.llm_chain)
+            if llm_chain.model_instance.model_mode == ModelMode.CHAT:
+                return (
+                    f"This was your previous work "
+                    f"(but I haven't seen any of it! I only see what "
+                    f"you return as final answer):\n{agent_scratchpad}"
+                )
+            else:
+                return agent_scratchpad
+        else:
+            return agent_scratchpad
+
     @classmethod
     def from_llm_and_tools(
             cls,
-            llm: BaseLanguageModel,
+            model_instance: BaseLLM,
             tools: Sequence[BaseTool],
             callback_manager: Optional[BaseCallbackManager] = None,
             output_parser: Optional[AgentOutputParser] = None,
@@ -157,17 +219,36 @@ class StructuredMultiDatasetRouterAgent(StructuredChatAgent):
             memory_prompts: Optional[List[BasePromptTemplate]] = None,
             **kwargs: Any,
     ) -> Agent:
-        return super().from_llm_and_tools(
-            llm=llm,
-            tools=tools,
+        """Construct an agent from an LLM and tools."""
+        cls._validate_tools(tools)
+        if model_instance.model_mode == ModelMode.CHAT:
+            prompt = cls.create_prompt(
+                tools,
+                prefix=prefix,
+                suffix=suffix,
+                human_message_template=human_message_template,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+                memory_prompts=memory_prompts,
+            )
+        else:
+            prompt = cls.create_completion_prompt(
+                tools,
+                prefix=prefix,
+                format_instructions=format_instructions,
+                input_variables=input_variables
+            )
+        llm_chain = LLMChain(
+            model_instance=model_instance,
+            prompt=prompt,
             callback_manager=callback_manager,
-            output_parser=output_parser,
-            prefix=prefix,
-            suffix=suffix,
-            human_message_template=human_message_template,
-            format_instructions=format_instructions,
-            input_variables=input_variables,
-            memory_prompts=memory_prompts,
+        )
+        tool_names = [tool.name for tool in tools]
+        _output_parser = output_parser
+        return cls(
+            llm_chain=llm_chain,
+            allowed_tools=tool_names,
+            output_parser=_output_parser,
             dataset_tools=tools,
             **kwargs,
         )

api/core/agent/agent/structured_chat.py

@@ -1,19 +1,21 @@
 import re
-from typing import List, Tuple, Any, Union, Sequence, Optional
+from typing import List, Tuple, Any, Union, Sequence, Optional, cast
 
-from langchain import BasePromptTemplate
+from langchain import BasePromptTemplate, PromptTemplate
 from langchain.agents import StructuredChatAgent, AgentOutputParser, Agent
 from langchain.agents.structured_chat.base import HUMAN_MESSAGE_TEMPLATE
-from langchain.base_language import BaseLanguageModel
 from langchain.callbacks.base import BaseCallbackManager
 from langchain.callbacks.manager import Callbacks
-from langchain.memory.summary import SummarizerMixin
+from langchain.memory.prompt import SUMMARY_PROMPT
 from langchain.prompts import SystemMessagePromptTemplate, HumanMessagePromptTemplate, ChatPromptTemplate
-from langchain.schema import AgentAction, AgentFinish, AIMessage, HumanMessage, OutputParserException
+from langchain.schema import AgentAction, AgentFinish, AIMessage, HumanMessage, OutputParserException, BaseMessage, \
+    get_buffer_string
 from langchain.tools import BaseTool
 from langchain.agents.structured_chat.prompt import PREFIX, SUFFIX
 
 from core.agent.agent.calc_token_mixin import CalcTokenMixin, ExceededLLMTokensLimitError
+from core.chain.llm_chain import LLMChain
+from core.model_providers.models.entity.model_params import ModelMode
 from core.model_providers.models.llm.base import BaseLLM
 
 FORMAT_INSTRUCTIONS = """Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).
@@ -52,8 +54,7 @@ Action:
 class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
     moving_summary_buffer: str = ""
     moving_summary_index: int = 0
-    summary_llm: BaseLanguageModel = None
-    model_instance: BaseLLM
+    summary_model_instance: BaseLLM = None
 
     class Config:
         """Configuration for this pydantic object."""
@@ -95,14 +96,14 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
         if prompts:
             messages = prompts[0].to_messages()
 
-        rest_tokens = self.get_message_rest_tokens(self.model_instance, messages)
+        rest_tokens = self.get_message_rest_tokens(self.llm_chain.model_instance, messages)
         if rest_tokens < 0:
             full_inputs = self.summarize_messages(intermediate_steps, **kwargs)
 
         try:
             full_output = self.llm_chain.predict(callbacks=callbacks, **full_inputs)
         except Exception as e:
-            new_exception = self.model_instance.handle_exceptions(e)
+            new_exception = self.llm_chain.model_instance.handle_exceptions(e)
             raise new_exception
 
         try:
@@ -118,7 +119,7 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
                                           "I don't know how to respond to that."}, "")
 
     def summarize_messages(self, intermediate_steps: List[Tuple[AgentAction, str]], **kwargs):
-        if len(intermediate_steps) >= 2 and self.summary_llm:
+        if len(intermediate_steps) >= 2 and self.summary_model_instance:
             should_summary_intermediate_steps = intermediate_steps[self.moving_summary_index:-1]
             should_summary_messages = [AIMessage(content=observation)
                                        for _, observation in should_summary_intermediate_steps]
@@ -130,11 +131,10 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
             error_msg = "Exceeded LLM tokens limit, stopped."
             raise ExceededLLMTokensLimitError(error_msg)
 
-        summary_handler = SummarizerMixin(llm=self.summary_llm)
         if self.moving_summary_buffer and 'chat_history' in kwargs:
             kwargs["chat_history"].pop()
 
-        self.moving_summary_buffer = summary_handler.predict_new_summary(
+        self.moving_summary_buffer = self.predict_new_summary(
             messages=should_summary_messages,
             existing_summary=self.moving_summary_buffer
         )
@@ -144,6 +144,18 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
 
         return self.get_full_inputs([intermediate_steps[-1]], **kwargs)
 
+    def predict_new_summary(
+        self, messages: List[BaseMessage], existing_summary: str
+    ) -> str:
+        new_lines = get_buffer_string(
+            messages,
+            human_prefix="Human",
+            ai_prefix="AI",
+        )
+
+        chain = LLMChain(model_instance=self.summary_model_instance, prompt=SUMMARY_PROMPT)
+        return chain.predict(summary=existing_summary, new_lines=new_lines)
+
     @classmethod
     def create_prompt(
             cls,
@@ -173,10 +185,65 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
         ]
         return ChatPromptTemplate(input_variables=input_variables, messages=messages)
 
+    @classmethod
+    def create_completion_prompt(
+            cls,
+            tools: Sequence[BaseTool],
+            prefix: str = PREFIX,
+            format_instructions: str = FORMAT_INSTRUCTIONS,
+            input_variables: Optional[List[str]] = None,
+    ) -> PromptTemplate:
+        """Create prompt in the style of the zero shot agent.
+
+        Args:
+            tools: List of tools the agent will have access to, used to format the
+                prompt.
+            prefix: String to put before the list of tools.
+            input_variables: List of input variables the final prompt will expect.
+
+        Returns:
+            A PromptTemplate with the template assembled from the pieces here.
+        """
+        suffix = """Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.
+Question: {input}
+Thought: {agent_scratchpad}
+"""
+
+        tool_strings = "\n".join([f"{tool.name}: {tool.description}" for tool in tools])
+        tool_names = ", ".join([tool.name for tool in tools])
+        format_instructions = format_instructions.format(tool_names=tool_names)
+        template = "\n\n".join([prefix, tool_strings, format_instructions, suffix])
+        if input_variables is None:
+            input_variables = ["input", "agent_scratchpad"]
+        return PromptTemplate(template=template, input_variables=input_variables)
+
+    def _construct_scratchpad(
+        self, intermediate_steps: List[Tuple[AgentAction, str]]
+    ) -> str:
+        agent_scratchpad = ""
+        for action, observation in intermediate_steps:
+            agent_scratchpad += action.log
+            agent_scratchpad += f"\n{self.observation_prefix}{observation}\n{self.llm_prefix}"
+
+        if not isinstance(agent_scratchpad, str):
+            raise ValueError("agent_scratchpad should be of type string.")
+        if agent_scratchpad:
+            llm_chain = cast(LLMChain, self.llm_chain)
+            if llm_chain.model_instance.model_mode == ModelMode.CHAT:
+                return (
+                    f"This was your previous work "
+                    f"(but I haven't seen any of it! I only see what "
+                    f"you return as final answer):\n{agent_scratchpad}"
+                )
+            else:
+                return agent_scratchpad
+        else:
+            return agent_scratchpad
+
     @classmethod
     def from_llm_and_tools(
             cls,
-            llm: BaseLanguageModel,
+            model_instance: BaseLLM,
             tools: Sequence[BaseTool],
             callback_manager: Optional[BaseCallbackManager] = None,
             output_parser: Optional[AgentOutputParser] = None,
@@ -188,16 +255,35 @@ class AutoSummarizingStructuredChatAgent(StructuredChatAgent, CalcTokenMixin):
             memory_prompts: Optional[List[BasePromptTemplate]] = None,
             **kwargs: Any,
     ) -> Agent:
-        return super().from_llm_and_tools(
-            llm=llm,
-            tools=tools,
+        """Construct an agent from an LLM and tools."""
+        cls._validate_tools(tools)
+        if model_instance.model_mode == ModelMode.CHAT:
+            prompt = cls.create_prompt(
+                tools,
+                prefix=prefix,
+                suffix=suffix,
+                human_message_template=human_message_template,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+                memory_prompts=memory_prompts,
+            )
+        else:
+            prompt = cls.create_completion_prompt(
+                tools,
+                prefix=prefix,
+                format_instructions=format_instructions,
+                input_variables=input_variables,
+            )
+        llm_chain = LLMChain(
+            model_instance=model_instance,
+            prompt=prompt,
             callback_manager=callback_manager,
-            output_parser=output_parser,
-            prefix=prefix,
-            suffix=suffix,
-            human_message_template=human_message_template,
-            format_instructions=format_instructions,
-            input_variables=input_variables,
-            memory_prompts=memory_prompts,
+        )
+        tool_names = [tool.name for tool in tools]
+        _output_parser = output_parser
+        return cls(
+            llm_chain=llm_chain,
+            allowed_tools=tool_names,
+            output_parser=_output_parser,
             **kwargs,
         )

api/core/agent/agent_executor.py

@@ -10,7 +10,6 @@ from pydantic import BaseModel, Extra
 
 from core.agent.agent.multi_dataset_router_agent import MultiDatasetRouterAgent
 from core.agent.agent.openai_function_call import AutoSummarizingOpenAIFunctionCallAgent
-from core.agent.agent.openai_multi_function_call import AutoSummarizingOpenMultiAIFunctionCallAgent
 from core.agent.agent.output_parser.structured_chat import StructuredChatOutputParser
 from core.agent.agent.structed_multi_dataset_router_agent import StructuredMultiDatasetRouterAgent
 from core.agent.agent.structured_chat import AutoSummarizingStructuredChatAgent
@@ -19,6 +18,7 @@ from langchain.agents import AgentExecutor as LCAgentExecutor
 from core.helper import moderation
 from core.model_providers.error import LLMError
 from core.model_providers.models.llm.base import BaseLLM
+from core.tool.dataset_multi_retriever_tool import DatasetMultiRetrieverTool
 from core.tool.dataset_retriever_tool import DatasetRetrieverTool
 
 
@@ -27,7 +27,6 @@ class PlanningStrategy(str, enum.Enum):
     REACT_ROUTER = 'react_router'
     REACT = 'react'
     FUNCTION_CALL = 'function_call'
-    MULTI_FUNCTION_CALL = 'multi_function_call'
 
 
 class AgentConfiguration(BaseModel):
@@ -64,47 +63,33 @@ class AgentExecutor:
         if self.configuration.strategy == PlanningStrategy.REACT:
             agent = AutoSummarizingStructuredChatAgent.from_llm_and_tools(
                 model_instance=self.configuration.model_instance,
-                llm=self.configuration.model_instance.client,
                 tools=self.configuration.tools,
                 output_parser=StructuredChatOutputParser(),
-                summary_llm=self.configuration.summary_model_instance.client
+                summary_model_instance=self.configuration.summary_model_instance
                 if self.configuration.summary_model_instance else None,
                 verbose=True
             )
         elif self.configuration.strategy == PlanningStrategy.FUNCTION_CALL:
             agent = AutoSummarizingOpenAIFunctionCallAgent.from_llm_and_tools(
                 model_instance=self.configuration.model_instance,
-                llm=self.configuration.model_instance.client,
                 tools=self.configuration.tools,
                 extra_prompt_messages=self.configuration.memory.buffer if self.configuration.memory else None,  # used for read chat histories memory
-                summary_llm=self.configuration.summary_model_instance.client
-                if self.configuration.summary_model_instance else None,
-                verbose=True
-            )
-        elif self.configuration.strategy == PlanningStrategy.MULTI_FUNCTION_CALL:
-            agent = AutoSummarizingOpenMultiAIFunctionCallAgent.from_llm_and_tools(
-                model_instance=self.configuration.model_instance,
-                llm=self.configuration.model_instance.client,
-                tools=self.configuration.tools,
-                extra_prompt_messages=self.configuration.memory.buffer if self.configuration.memory else None,  # used for read chat histories memory
-                summary_llm=self.configuration.summary_model_instance.client
+                summary_model_instance=self.configuration.summary_model_instance
                 if self.configuration.summary_model_instance else None,
                 verbose=True
             )
         elif self.configuration.strategy == PlanningStrategy.ROUTER:
-            self.configuration.tools = [t for t in self.configuration.tools if isinstance(t, DatasetRetrieverTool)]
+            self.configuration.tools = [t for t in self.configuration.tools if isinstance(t, DatasetRetrieverTool) or isinstance(t, DatasetMultiRetrieverTool)]
             agent = MultiDatasetRouterAgent.from_llm_and_tools(
                 model_instance=self.configuration.model_instance,
-                llm=self.configuration.model_instance.client,
                 tools=self.configuration.tools,
                 extra_prompt_messages=self.configuration.memory.buffer if self.configuration.memory else None,
                 verbose=True
             )
         elif self.configuration.strategy == PlanningStrategy.REACT_ROUTER:
-            self.configuration.tools = [t for t in self.configuration.tools if isinstance(t, DatasetRetrieverTool)]
+            self.configuration.tools = [t for t in self.configuration.tools if isinstance(t, DatasetRetrieverTool) or isinstance(t, DatasetMultiRetrieverTool)]
             agent = StructuredMultiDatasetRouterAgent.from_llm_and_tools(
                 model_instance=self.configuration.model_instance,
-                llm=self.configuration.model_instance.client,
                 tools=self.configuration.tools,
                 output_parser=StructuredChatOutputParser(),
                 verbose=True

api/core/callback_handler/index_tool_callback_handler.py

@@ -10,8 +10,7 @@ from models.dataset import DocumentSegment
 class DatasetIndexToolCallbackHandler:
     """Callback handler for dataset tool."""
 
-    def __init__(self, dataset_id: str, conversation_message_task: ConversationMessageTask) -> None:
-        self.dataset_id = dataset_id
+    def __init__(self, conversation_message_task: ConversationMessageTask) -> None:
         self.conversation_message_task = conversation_message_task
 
     def on_tool_end(self, documents: List[Document]) -> None:
@@ -21,7 +20,6 @@ class DatasetIndexToolCallbackHandler:
 
             # add hit count to document segment
             db.session.query(DocumentSegment).filter(
-                DocumentSegment.dataset_id == self.dataset_id,
                 DocumentSegment.index_node_id == doc_id
             ).update(
                 {DocumentSegment.hit_count: DocumentSegment.hit_count + 1},

api/core/callback_handler/llm_callback_handler.py

@@ -1,13 +1,26 @@
 import logging
-from typing import Any, Dict, List, Union
+import threading
+import time
+from typing import Any, Dict, List, Union, Optional
 
+from flask import Flask, current_app
 from langchain.callbacks.base import BaseCallbackHandler
 from langchain.schema import LLMResult, BaseMessage
+from pydantic import BaseModel
 
 from core.callback_handler.entity.llm_message import LLMMessage
-from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException
-from core.model_providers.models.entity.message import to_prompt_messages, PromptMessage
+from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException, \
+    ConversationTaskInterruptException
+from core.model_providers.models.entity.message import to_prompt_messages, PromptMessage, LCHumanMessageWithFiles, \
+    ImagePromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
+from core.moderation.base import ModerationOutputsResult, ModerationAction
+from core.moderation.factory import ModerationFactory
+
+
+class ModerationRule(BaseModel):
+    type: str
+    config: Dict[str, Any]
 
 
 class LLMCallbackHandler(BaseCallbackHandler):
@@ -20,6 +33,24 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.start_at = None
         self.conversation_message_task = conversation_message_task
 
+        self.output_moderation_handler = None
+        self.init_output_moderation()
+
+    def init_output_moderation(self):
+        app_model_config = self.conversation_message_task.app_model_config
+        sensitive_word_avoidance_dict = app_model_config.sensitive_word_avoidance_dict
+
+        if sensitive_word_avoidance_dict and sensitive_word_avoidance_dict.get("enabled"):
+            self.output_moderation_handler = OutputModerationHandler(
+                tenant_id=self.conversation_message_task.tenant_id,
+                app_id=self.conversation_message_task.app.id,
+                rule=ModerationRule(
+                    type=sensitive_word_avoidance_dict.get("type"),
+                    config=sensitive_word_avoidance_dict.get("config")
+                ),
+                on_message_replace_func=self.conversation_message_task.on_message_replace
+            )
+
     @property
     def always_verbose(self) -> bool:
         """Whether to call verbose callbacks even if verbose is False."""
@@ -42,7 +73,12 @@ class LLMCallbackHandler(BaseCallbackHandler):
 
             real_prompts.append({
                 "role": role,
-                "text": message.content
+                "text": message.content,
+                "files": [{
+                    "type": file.type.value,
+                    "data": file.data[:10] + '...[TRUNCATED]...' + file.data[-10:],
+                    "detail": file.detail.value if isinstance(file, ImagePromptMessageFile) else None,
+                } for file in (message.files if isinstance(message, LCHumanMessageWithFiles) else [])]
             })
 
         self.llm_message.prompt = real_prompts
@@ -59,10 +95,19 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.llm_message.prompt_tokens = self.model_instance.get_num_tokens([PromptMessage(content=prompts[0])])
 
     def on_llm_end(self, response: LLMResult, **kwargs: Any) -> None:
-        if not self.conversation_message_task.streaming:
-            self.conversation_message_task.append_message_text(response.generations[0][0].text)
+        if self.output_moderation_handler:
+            self.output_moderation_handler.stop_thread()
+
+            self.llm_message.completion = self.output_moderation_handler.moderation_completion(
+                completion=response.generations[0][0].text,
+                public_event=True if self.conversation_message_task.streaming else False
+            )
+        else:
             self.llm_message.completion = response.generations[0][0].text
 
+        if not self.conversation_message_task.streaming:
+            self.conversation_message_task.append_message_text(self.llm_message.completion)
+
         if response.llm_output and 'token_usage' in response.llm_output:
             if 'prompt_tokens' in response.llm_output['token_usage']:
                 self.llm_message.prompt_tokens = response.llm_output['token_usage']['prompt_tokens']
@@ -79,23 +124,161 @@ class LLMCallbackHandler(BaseCallbackHandler):
         self.conversation_message_task.save_message(self.llm_message)
 
     def on_llm_new_token(self, token: str, **kwargs: Any) -> None:
-        try:
-            self.conversation_message_task.append_message_text(token)
-        except ConversationTaskStoppedException as ex:
+        if self.output_moderation_handler and self.output_moderation_handler.should_direct_output():
+            # stop subscribe new token when output moderation should direct output
+            ex = ConversationTaskInterruptException()
             self.on_llm_error(error=ex)
             raise ex
 
-        self.llm_message.completion += token
+        try:
+            self.conversation_message_task.append_message_text(token)
+            self.llm_message.completion += token
+
+            if self.output_moderation_handler:
+                self.output_moderation_handler.append_new_token(token)
+        except ConversationTaskStoppedException as ex:
+            self.on_llm_error(error=ex)
+            raise ex
 
     def on_llm_error(
             self, error: Union[Exception, KeyboardInterrupt], **kwargs: Any
     ) -> None:
         """Do nothing."""
+        if self.output_moderation_handler:
+            self.output_moderation_handler.stop_thread()
+
         if isinstance(error, ConversationTaskStoppedException):
             if self.conversation_message_task.streaming:
                 self.llm_message.completion_tokens = self.model_instance.get_num_tokens(
                     [PromptMessage(content=self.llm_message.completion)]
                 )
                 self.conversation_message_task.save_message(llm_message=self.llm_message, by_stopped=True)
+        if isinstance(error, ConversationTaskInterruptException):
+            self.llm_message.completion = self.output_moderation_handler.get_final_output()
+            self.llm_message.completion_tokens = self.model_instance.get_num_tokens(
+                [PromptMessage(content=self.llm_message.completion)]
+            )
+            self.conversation_message_task.save_message(llm_message=self.llm_message)
         else:
             logging.debug("on_llm_error: %s", error)
+
+
+class OutputModerationHandler(BaseModel):
+    DEFAULT_BUFFER_SIZE: int = 300
+
+    tenant_id: str
+    app_id: str
+
+    rule: ModerationRule
+    on_message_replace_func: Any
+
+    thread: Optional[threading.Thread] = None
+    thread_running: bool = True
+    buffer: str = ''
+    is_final_chunk: bool = False
+    final_output: Optional[str] = None
+
+    class Config:
+        arbitrary_types_allowed = True
+
+    def should_direct_output(self):
+        return self.final_output is not None
+
+    def get_final_output(self):
+        return self.final_output
+
+    def append_new_token(self, token: str):
+        self.buffer += token
+
+        if not self.thread:
+            self.thread = self.start_thread()
+
+    def moderation_completion(self, completion: str, public_event: bool = False) -> str:
+        self.buffer = completion
+        self.is_final_chunk = True
+
+        result = self.moderation(
+            tenant_id=self.tenant_id,
+            app_id=self.app_id,
+            moderation_buffer=completion
+        )
+
+        if not result or not result.flagged:
+            return completion
+
+        if result.action == ModerationAction.DIRECT_OUTPUT:
+            final_output = result.preset_response
+        else:
+            final_output = result.text
+
+        if public_event:
+            self.on_message_replace_func(final_output)
+
+        return final_output
+
+    def start_thread(self) -> threading.Thread:
+        buffer_size = int(current_app.config.get('MODERATION_BUFFER_SIZE', self.DEFAULT_BUFFER_SIZE))
+        thread = threading.Thread(target=self.worker, kwargs={
+            'flask_app': current_app._get_current_object(),
+            'buffer_size': buffer_size if buffer_size > 0 else self.DEFAULT_BUFFER_SIZE
+        })
+
+        thread.start()
+
+        return thread
+
+    def stop_thread(self):
+        if self.thread and self.thread.is_alive():
+            self.thread_running = False
+
+    def worker(self, flask_app: Flask, buffer_size: int):
+        with flask_app.app_context():
+            current_length = 0
+            while self.thread_running:
+                moderation_buffer = self.buffer
+                buffer_length = len(moderation_buffer)
+                if not self.is_final_chunk:
+                    chunk_length = buffer_length - current_length
+                    if 0 <= chunk_length < buffer_size:
+                        time.sleep(1)
+                        continue
+
+                current_length = buffer_length
+
+                result = self.moderation(
+                    tenant_id=self.tenant_id,
+                    app_id=self.app_id,
+                    moderation_buffer=moderation_buffer
+                )
+
+                if not result or not result.flagged:
+                    continue
+
+                if result.action == ModerationAction.DIRECT_OUTPUT:
+                    final_output = result.preset_response
+                    self.final_output = final_output
+                else:
+                    final_output = result.text + self.buffer[len(moderation_buffer):]
+
+                # trigger replace event
+                if self.thread_running:
+                    self.on_message_replace_func(final_output)
+
+                if result.action == ModerationAction.DIRECT_OUTPUT:
+                    break
+
+    def moderation(self, tenant_id: str, app_id: str, moderation_buffer: str) -> Optional[ModerationOutputsResult]:
+        try:
+            moderation_factory = ModerationFactory(
+                name=self.rule.type,
+                app_id=app_id,
+                tenant_id=tenant_id,
+                config=self.rule.config
+            )
+
+            result: ModerationOutputsResult = moderation_factory.moderation_for_outputs(moderation_buffer)
+            return result
+        except Exception as e:
+            logging.error("Moderation Output error: %s", e)
+
+        return None

api/core/chain/llm_chain.py

@@ -0,0 +1,36 @@
+from typing import List, Dict, Any, Optional
+
+from langchain import LLMChain as LCLLMChain
+from langchain.callbacks.manager import CallbackManagerForChainRun
+from langchain.schema import LLMResult, Generation
+from langchain.schema.language_model import BaseLanguageModel
+
+from core.model_providers.models.entity.message import to_prompt_messages
+from core.model_providers.models.llm.base import BaseLLM
+from core.third_party.langchain.llms.fake import FakeLLM
+
+
+class LLMChain(LCLLMChain):
+    model_instance: BaseLLM
+    """The language model instance to use."""
+    llm: BaseLanguageModel = FakeLLM(response="")
+
+    def generate(
+        self,
+        input_list: List[Dict[str, Any]],
+        run_manager: Optional[CallbackManagerForChainRun] = None,
+    ) -> LLMResult:
+        """Generate LLM result from inputs."""
+        prompts, stop = self.prep_prompts(input_list, run_manager=run_manager)
+        messages = prompts[0].to_messages()
+        prompt_messages = to_prompt_messages(messages)
+        result = self.model_instance.run(
+            messages=prompt_messages,
+            stop=stop
+        )
+
+        generations = [
+            [Generation(text=result.content)]
+        ]
+
+        return LLMResult(generations=generations)

api/core/chain/sensitive_word_avoidance_chain.py

@@ -1,92 +0,0 @@
-import enum
-import logging
-from typing import List, Dict, Optional, Any
-
-from langchain.callbacks.manager import CallbackManagerForChainRun
-from langchain.chains.base import Chain
-from pydantic import BaseModel
-
-from core.model_providers.error import LLMBadRequestError
-from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.llm.base import BaseLLM
-from core.model_providers.models.moderation import openai_moderation
-
-
-class SensitiveWordAvoidanceRule(BaseModel):
-    class Type(enum.Enum):
-        MODERATION = "moderation"
-        KEYWORDS = "keywords"
-
-    type: Type
-    canned_response: str = 'Your content violates our usage policy. Please revise and try again.'
-    extra_params: dict = {}
-
-
-class SensitiveWordAvoidanceChain(Chain):
-    input_key: str = "input"  #: :meta private:
-    output_key: str = "output"  #: :meta private:
-
-    model_instance: BaseLLM
-    sensitive_word_avoidance_rule: SensitiveWordAvoidanceRule
-
-    @property
-    def _chain_type(self) -> str:
-        return "sensitive_word_avoidance_chain"
-
-    @property
-    def input_keys(self) -> List[str]:
-        """Expect input key.
-
-        :meta private:
-        """
-        return [self.input_key]
-
-    @property
-    def output_keys(self) -> List[str]:
-        """Return output key.
-
-        :meta private:
-        """
-        return [self.output_key]
-
-    def _check_sensitive_word(self, text: str) -> bool:
-        for word in self.sensitive_word_avoidance_rule.extra_params.get('sensitive_words', []):
-            if word in text:
-                return False
-        return True
-
-    def _check_moderation(self, text: str) -> bool:
-        moderation_model_instance = ModelFactory.get_moderation_model(
-            tenant_id=self.model_instance.model_provider.provider.tenant_id,
-            model_provider_name='openai',
-            model_name=openai_moderation.DEFAULT_MODEL
-        )
-
-        try:
-            return moderation_model_instance.run(text=text)
-        except Exception as ex:
-            logging.exception(ex)
-            raise LLMBadRequestError('Rate limit exceeded, please try again later.')
-
-    def _call(
-            self,
-            inputs: Dict[str, Any],
-            run_manager: Optional[CallbackManagerForChainRun] = None,
-    ) -> Dict[str, Any]:
-        text = inputs[self.input_key]
-
-        if self.sensitive_word_avoidance_rule.type == SensitiveWordAvoidanceRule.Type.KEYWORDS:
-            result = self._check_sensitive_word(text)
-        else:
-            result = self._check_moderation(text)
-
-        if not result:
-            raise SensitiveWordAvoidanceError(self.sensitive_word_avoidance_rule.canned_response)
-
-        return {self.output_key: text}
-
-
-class SensitiveWordAvoidanceError(Exception):
-    def __init__(self, message):
-        super().__init__(message)
-        self.message = message

api/core/completion.py

@@ -1,36 +1,43 @@
+import concurrent
 import json
 import logging
-from typing import Optional, List, Union
+from concurrent.futures import ThreadPoolExecutor
+from typing import Optional, List, Union, Tuple
 
+from flask import current_app, Flask
 from requests.exceptions import ChunkedEncodingError
 
 from core.agent.agent_executor import AgentExecuteResult, PlanningStrategy
 from core.callback_handler.main_chain_gather_callback_handler import MainChainGatherCallbackHandler
 from core.callback_handler.llm_callback_handler import LLMCallbackHandler
-from core.chain.sensitive_word_avoidance_chain import SensitiveWordAvoidanceError
-from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException
+from core.conversation_message_task import ConversationMessageTask, ConversationTaskStoppedException, \
+    ConversationTaskInterruptException
+from core.external_data_tool.factory import ExternalDataToolFactory
+from core.file.file_obj import FileObj
 from core.model_providers.error import LLMBadRequestError
 from core.memory.read_only_conversation_token_db_buffer_shared_memory import \
     ReadOnlyConversationTokenDBBufferSharedMemory
 from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.entity.message import PromptMessage
+from core.model_providers.models.entity.message import PromptMessage, PromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
 from core.orchestrator_rule_parser import OrchestratorRuleParser
-from core.prompt.prompt_builder import PromptBuilder
-from core.prompt.prompts import MORE_LIKE_THIS_GENERATE_PROMPT
-from models.dataset import DocumentSegment, Dataset, Document
-from models.model import App, AppModelConfig, Account, Conversation, Message, EndUser
+from core.prompt.prompt_template import PromptTemplateParser
+from core.prompt.prompt_transform import PromptTransform
+from models.model import App, AppModelConfig, Account, Conversation, EndUser
+from core.moderation.base import ModerationException, ModerationAction
+from core.moderation.factory import ModerationFactory
 
 
 class Completion:
     @classmethod
     def generate(cls, task_id: str, app: App, app_model_config: AppModelConfig, query: str, inputs: dict,
-                 user: Union[Account, EndUser], conversation: Optional[Conversation], streaming: bool,
-                 is_override: bool = False, retriever_from: str = 'dev'):
+                 files: List[FileObj], user: Union[Account, EndUser], conversation: Optional[Conversation],
+                 streaming: bool, is_override: bool = False, retriever_from: str = 'dev',
+                 auto_generate_name: bool = True):
         """
         errors: ProviderTokenNotInitError
         """
-        query = PromptBuilder.process_template(query)
+        query = PromptTemplateParser.remove_template_variables(query)
 
         memory = None
         if conversation:
@@ -59,16 +66,21 @@ class Completion:
             is_override=is_override,
             inputs=inputs,
             query=query,
+            files=files,
             streaming=streaming,
-            model_instance=final_model_instance
+            model_instance=final_model_instance,
+            auto_generate_name=auto_generate_name
         )
 
+        prompt_message_files = [file.prompt_message_file for file in files]
+
         rest_tokens_for_context_and_memory = cls.get_validate_rest_tokens(
             mode=app.mode,
             model_instance=final_model_instance,
             app_model_config=app_model_config,
             query=query,
-            inputs=inputs
+            inputs=inputs,
+            files=prompt_message_files
         )
 
         # init orchestrator rule parser
@@ -78,26 +90,36 @@ class Completion:
         )
 
         try:
-            # parse sensitive_word_avoidance_chain
             chain_callback = MainChainGatherCallbackHandler(conversation_message_task)
-            sensitive_word_avoidance_chain = orchestrator_rule_parser.to_sensitive_word_avoidance_chain(
-                final_model_instance, [chain_callback])
-            if sensitive_word_avoidance_chain:
-                try:
-                    query = sensitive_word_avoidance_chain.run(query)
-                except SensitiveWordAvoidanceError as ex:
-                    cls.run_final_llm(
-                        model_instance=final_model_instance,
-                        mode=app.mode,
-                        app_model_config=app_model_config,
-                        query=query,
-                        inputs=inputs,
-                        agent_execute_result=None,
-                        conversation_message_task=conversation_message_task,
-                        memory=memory,
-                        fake_response=ex.message
-                    )
-                    return
+
+            try:
+                # process sensitive_word_avoidance
+                inputs, query = cls.moderation_for_inputs(app.id, app.tenant_id, app_model_config, inputs, query)
+            except ModerationException as e:
+                cls.run_final_llm(
+                    model_instance=final_model_instance,
+                    mode=app.mode,
+                    app_model_config=app_model_config,
+                    query=query,
+                    inputs=inputs,
+                    files=prompt_message_files,
+                    agent_execute_result=None,
+                    conversation_message_task=conversation_message_task,
+                    memory=memory,
+                    fake_response=str(e)
+                )
+                return
+
+            # fill in variable inputs from external data tools if exists
+            external_data_tools = app_model_config.external_data_tools_list
+            if external_data_tools:
+                inputs = cls.fill_in_inputs_from_external_data_tools(
+                    tenant_id=app.tenant_id,
+                    app_id=app.id,
+                    external_data_tools=external_data_tools,
+                    inputs=inputs,
+                    query=query
+                )
 
             # get agent executor
             agent_executor = orchestrator_rule_parser.to_agent_executor(
@@ -105,6 +127,7 @@ class Completion:
                 memory=memory,
                 rest_tokens=rest_tokens_for_context_and_memory,
                 chain_callback=chain_callback,
+                tenant_id=app.tenant_id,
                 retriever_from=retriever_from
             )
 
@@ -132,42 +155,152 @@ class Completion:
                 app_model_config=app_model_config,
                 query=query,
                 inputs=inputs,
+                files=prompt_message_files,
                 agent_execute_result=agent_execute_result,
                 conversation_message_task=conversation_message_task,
                 memory=memory,
                 fake_response=fake_response
             )
-        except ConversationTaskStoppedException:
+        except (ConversationTaskInterruptException, ConversationTaskStoppedException):
             return
         except ChunkedEncodingError as e:
             # Interrupt by LLM (like OpenAI), handle it.
             logging.warning(f'ChunkedEncodingError: {e}')
             conversation_message_task.end()
             return
-        
+
+    @classmethod
+    def moderation_for_inputs(cls, app_id: str, tenant_id: str, app_model_config: AppModelConfig, inputs: dict, query: str):
+        if not app_model_config.sensitive_word_avoidance_dict['enabled']:
+            return inputs, query
+
+        type = app_model_config.sensitive_word_avoidance_dict['type']
+
+        moderation = ModerationFactory(type, app_id, tenant_id, app_model_config.sensitive_word_avoidance_dict['config'])
+        moderation_result = moderation.moderation_for_inputs(inputs, query)
+
+        if not moderation_result.flagged:
+            return inputs, query
+
+        if moderation_result.action == ModerationAction.DIRECT_OUTPUT:
+            raise ModerationException(moderation_result.preset_response)
+        elif moderation_result.action == ModerationAction.OVERRIDED:
+            inputs = moderation_result.inputs
+            query = moderation_result.query
+
+        return inputs, query
+
+    @classmethod
+    def fill_in_inputs_from_external_data_tools(cls, tenant_id: str, app_id: str, external_data_tools: list[dict],
+                                                inputs: dict, query: str) -> dict:
+        """
+        Fill in variable inputs from external data tools if exists.
+
+        :param tenant_id: workspace id
+        :param app_id: app id
+        :param external_data_tools: external data tools configs
+        :param inputs: the inputs
+        :param query: the query
+        :return: the filled inputs
+        """
+        # Group tools by type and config
+        grouped_tools = {}
+        for tool in external_data_tools:
+            if not tool.get("enabled"):
+                continue
+
+            tool_key = (tool.get("type"), json.dumps(tool.get("config"), sort_keys=True))
+            grouped_tools.setdefault(tool_key, []).append(tool)
+
+        results = {}
+        with ThreadPoolExecutor() as executor:
+            futures = {}
+            for tool in external_data_tools:
+                if not tool.get("enabled"):
+                    continue
+
+                future = executor.submit(
+                    cls.query_external_data_tool, current_app._get_current_object(), tenant_id, app_id, tool,
+                    inputs, query
+                )
+
+                futures[future] = tool
+
+            for future in concurrent.futures.as_completed(futures):
+                tool_variable, result = future.result()
+                results[tool_variable] = result
+
+        inputs.update(results)
+        return inputs
+
+    @classmethod
+    def query_external_data_tool(cls, flask_app: Flask, tenant_id: str, app_id: str, external_data_tool: dict,
+                                 inputs: dict, query: str) -> Tuple[Optional[str], Optional[str]]:
+        with flask_app.app_context():
+            tool_variable = external_data_tool.get("variable")
+            tool_type = external_data_tool.get("type")
+            tool_config = external_data_tool.get("config")
+
+            external_data_tool_factory = ExternalDataToolFactory(
+                name=tool_type,
+                tenant_id=tenant_id,
+                app_id=app_id,
+                variable=tool_variable,
+                config=tool_config
+            )
+
+            # query external data tool
+            result = external_data_tool_factory.query(
+                inputs=inputs,
+                query=query
+            )
+
+            return tool_variable, result
+
     @classmethod
     def get_query_for_agent(cls, app: App, app_model_config: AppModelConfig, query: str, inputs: dict) -> str:
         if app.mode != 'completion':
             return query
-        
+
         return inputs.get(app_model_config.dataset_query_variable, "")
 
     @classmethod
     def run_final_llm(cls, model_instance: BaseLLM, mode: str, app_model_config: AppModelConfig, query: str,
                       inputs: dict,
+                      files: List[PromptMessageFile],
                       agent_execute_result: Optional[AgentExecuteResult],
                       conversation_message_task: ConversationMessageTask,
                       memory: Optional[ReadOnlyConversationTokenDBBufferSharedMemory],
                       fake_response: Optional[str]):
+        prompt_transform = PromptTransform()
+
         # get llm prompt
-        prompt_messages, stop_words = model_instance.get_prompt(
-            mode=mode,
-            pre_prompt=app_model_config.pre_prompt,
-            inputs=inputs,
-            query=query,
-            context=agent_execute_result.output if agent_execute_result else None,
-            memory=memory
-        )
+        if app_model_config.prompt_type == 'simple':
+            prompt_messages, stop_words = prompt_transform.get_prompt(
+                app_mode=mode,
+                pre_prompt=app_model_config.pre_prompt,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=agent_execute_result.output if agent_execute_result else None,
+                memory=memory,
+                model_instance=model_instance
+            )
+        else:
+            prompt_messages = prompt_transform.get_advanced_prompt(
+                app_mode=mode,
+                app_model_config=app_model_config,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=agent_execute_result.output if agent_execute_result else None,
+                memory=memory,
+                model_instance=model_instance
+            )
+
+            model_config = app_model_config.model_dict
+            completion_params = model_config.get("completion_params", {})
+            stop_words = completion_params.get("stop", [])
 
         cls.recale_llm_max_tokens(
             model_instance=model_instance,
@@ -176,7 +309,7 @@ class Completion:
 
         response = model_instance.run(
             messages=prompt_messages,
-            stop=stop_words,
+            stop=stop_words if stop_words else None,
             callbacks=[LLMCallbackHandler(model_instance, conversation_message_task)],
             fake_response=fake_response
         )
@@ -217,7 +350,7 @@ class Completion:
 
     @classmethod
     def get_validate_rest_tokens(cls, mode: str, model_instance: BaseLLM, app_model_config: AppModelConfig,
-                                 query: str, inputs: dict) -> int:
+                                 query: str, inputs: dict, files: List[PromptMessageFile]) -> int:
         model_limited_tokens = model_instance.model_rules.max_tokens.max
         max_tokens = model_instance.get_model_kwargs().max_tokens
 
@@ -227,15 +360,31 @@ class Completion:
         if max_tokens is None:
             max_tokens = 0
 
+        prompt_transform = PromptTransform()
+
         # get prompt without memory and context
-        prompt_messages, _ = model_instance.get_prompt(
-            mode=mode,
-            pre_prompt=app_model_config.pre_prompt,
-            inputs=inputs,
-            query=query,
-            context=None,
-            memory=None
-        )
+        if app_model_config.prompt_type == 'simple':
+            prompt_messages, _ = prompt_transform.get_prompt(
+                app_mode=mode,
+                pre_prompt=app_model_config.pre_prompt,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=None,
+                memory=None,
+                model_instance=model_instance
+            )
+        else:
+            prompt_messages = prompt_transform.get_advanced_prompt(
+                app_mode=mode,
+                app_model_config=app_model_config,
+                inputs=inputs,
+                query=query,
+                files=files,
+                context=None,
+                memory=None,
+                model_instance=model_instance
+            )
 
         prompt_tokens = model_instance.get_num_tokens(prompt_messages)
         rest_tokens = model_limited_tokens - max_tokens - prompt_tokens
@@ -266,52 +415,3 @@ class Completion:
             model_kwargs = model_instance.get_model_kwargs()
             model_kwargs.max_tokens = max_tokens
             model_instance.set_model_kwargs(model_kwargs)
-
-    @classmethod
-    def generate_more_like_this(cls, task_id: str, app: App, message: Message, pre_prompt: str,
-                                app_model_config: AppModelConfig, user: Account, streaming: bool):
-
-        final_model_instance = ModelFactory.get_text_generation_model_from_model_config(
-            tenant_id=app.tenant_id,
-            model_config=app_model_config.model_dict,
-            streaming=streaming
-        )
-
-        # get llm prompt
-        old_prompt_messages, _ = final_model_instance.get_prompt(
-            mode='completion',
-            pre_prompt=pre_prompt,
-            inputs=message.inputs,
-            query=message.query,
-            context=None,
-            memory=None
-        )
-
-        original_completion = message.answer.strip()
-
-        prompt = MORE_LIKE_THIS_GENERATE_PROMPT
-        prompt = prompt.format(prompt=old_prompt_messages[0].content, original_completion=original_completion)
-
-        prompt_messages = [PromptMessage(content=prompt)]
-
-        conversation_message_task = ConversationMessageTask(
-            task_id=task_id,
-            app=app,
-            app_model_config=app_model_config,
-            user=user,
-            inputs=message.inputs,
-            query=message.query,
-            is_override=True if message.override_model_configs else False,
-            streaming=streaming,
-            model_instance=final_model_instance
-        )
-
-        cls.recale_llm_max_tokens(
-            model_instance=final_model_instance,
-            prompt_messages=prompt_messages
-        )
-
-        final_model_instance.run(
-            messages=prompt_messages,
-            callbacks=[LLMCallbackHandler(final_model_instance, conversation_message_task)]
-        )

api/core/conversation_message_task.py

@@ -6,23 +6,25 @@ from core.callback_handler.entity.agent_loop import AgentLoop
 from core.callback_handler.entity.dataset_query import DatasetQueryObj
 from core.callback_handler.entity.llm_message import LLMMessage
 from core.callback_handler.entity.chain_result import ChainResult
+from core.file.file_obj import FileObj
 from core.model_providers.model_factory import ModelFactory
-from core.model_providers.models.entity.message import to_prompt_messages, MessageType
+from core.model_providers.models.entity.message import to_prompt_messages, MessageType, PromptMessageFile
 from core.model_providers.models.llm.base import BaseLLM
 from core.prompt.prompt_builder import PromptBuilder
-from core.prompt.prompt_template import JinjaPromptTemplate
+from core.prompt.prompt_template import PromptTemplateParser
 from events.message_event import message_was_created
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from models.dataset import DatasetQuery
 from models.model import AppModelConfig, Conversation, Account, Message, EndUser, App, MessageAgentThought, \
-    MessageChain, DatasetRetrieverResource
+    MessageChain, DatasetRetrieverResource, MessageFile
 
 
 class ConversationMessageTask:
     def __init__(self, task_id: str, app: App, app_model_config: AppModelConfig, user: Account,
-                 inputs: dict, query: str, streaming: bool, model_instance: BaseLLM,
-                 conversation: Optional[Conversation] = None, is_override: bool = False):
+                 inputs: dict, query: str, files: List[FileObj], streaming: bool,
+                 model_instance: BaseLLM, conversation: Optional[Conversation] = None, is_override: bool = False,
+                 auto_generate_name: bool = True):
         self.start_at = time.perf_counter()
 
         self.task_id = task_id
@@ -35,6 +37,7 @@ class ConversationMessageTask:
         self.user = user
         self.inputs = inputs
         self.query = query
+        self.files = files
         self.streaming = streaming
 
         self.conversation = conversation
@@ -45,6 +48,7 @@ class ConversationMessageTask:
         self.message = None
 
         self.retriever_resource = None
+        self.auto_generate_name = auto_generate_name
 
         self.model_dict = self.app_model_config.model_dict
         self.provider_name = self.model_dict.get('provider')
@@ -74,10 +78,10 @@ class ConversationMessageTask:
         if self.mode == 'chat':
             introduction = self.app_model_config.opening_statement
             if introduction:
-                prompt_template = JinjaPromptTemplate.from_template(template=introduction)
-                prompt_inputs = {k: self.inputs[k] for k in prompt_template.input_variables if k in self.inputs}
+                prompt_template = PromptTemplateParser(template=introduction)
+                prompt_inputs = {k: self.inputs[k] for k in prompt_template.variable_keys if k in self.inputs}
                 try:
-                    introduction = prompt_template.format(**prompt_inputs)
+                    introduction = prompt_template.format(prompt_inputs)
                 except KeyError:
                     pass
 
@@ -94,13 +98,13 @@ class ConversationMessageTask:
         if not self.conversation:
             self.is_new_conversation = True
             self.conversation = Conversation(
-                app_id=self.app_model_config.app_id,
+                app_id=self.app.id,
                 app_model_config_id=self.app_model_config.id,
                 model_provider=self.provider_name,
                 model_id=self.model_name,
                 override_model_configs=json.dumps(override_model_configs) if override_model_configs else None,
                 mode=self.mode,
-                name='',
+                name='New conversation',
                 inputs=self.inputs,
                 introduction=introduction,
                 system_instruction=system_instruction,
@@ -115,7 +119,7 @@ class ConversationMessageTask:
             db.session.commit()
 
         self.message = Message(
-            app_id=self.app_model_config.app_id,
+            app_id=self.app.id,
             model_provider=self.provider_name,
             model_id=self.model_name,
             override_model_configs=json.dumps(override_model_configs) if override_model_configs else None,
@@ -142,6 +146,19 @@ class ConversationMessageTask:
         db.session.add(self.message)
         db.session.commit()
 
+        for file in self.files:
+            message_file = MessageFile(
+                message_id=self.message.id,
+                type=file.type.value,
+                transfer_method=file.transfer_method.value,
+                url=file.url,
+                upload_file_id=file.upload_file_id,
+                created_by_role=('account' if isinstance(self.user, Account) else 'end_user'),
+                created_by=self.user.id
+            )
+            db.session.add(message_file)
+            db.session.commit()
+
     def append_message_text(self, text: str):
         if text is not None:
             self._pub_handler.pub_text(text)
@@ -150,12 +167,12 @@ class ConversationMessageTask:
         message_tokens = llm_message.prompt_tokens
         answer_tokens = llm_message.completion_tokens
 
-        message_unit_price = self.model_instance.get_tokens_unit_price(MessageType.HUMAN)
-        message_price_unit = self.model_instance.get_price_unit(MessageType.HUMAN)
+        message_unit_price = self.model_instance.get_tokens_unit_price(MessageType.USER)
+        message_price_unit = self.model_instance.get_price_unit(MessageType.USER)
         answer_unit_price = self.model_instance.get_tokens_unit_price(MessageType.ASSISTANT)
         answer_price_unit = self.model_instance.get_price_unit(MessageType.ASSISTANT)
 
-        message_total_price = self.model_instance.calc_tokens_price(message_tokens, MessageType.HUMAN)
+        message_total_price = self.model_instance.calc_tokens_price(message_tokens, MessageType.USER)
         answer_total_price = self.model_instance.calc_tokens_price(answer_tokens, MessageType.ASSISTANT)
         total_price = message_total_price + answer_total_price
 
@@ -163,7 +180,7 @@ class ConversationMessageTask:
         self.message.message_tokens = message_tokens
         self.message.message_unit_price = message_unit_price
         self.message.message_price_unit = message_price_unit
-        self.message.answer = PromptBuilder.process_template(
+        self.message.answer = PromptTemplateParser.remove_template_variables(
             llm_message.completion.strip()) if llm_message.completion else ''
         self.message.answer_tokens = answer_tokens
         self.message.answer_unit_price = answer_unit_price
@@ -176,7 +193,8 @@ class ConversationMessageTask:
         message_was_created.send(
             self.message,
             conversation=self.conversation,
-            is_first_message=self.is_new_conversation
+            is_first_message=self.is_new_conversation,
+            auto_generate_name=self.auto_generate_name
         )
 
         if not by_stopped:
@@ -226,15 +244,15 @@ class ConversationMessageTask:
 
     def on_agent_end(self, message_agent_thought: MessageAgentThought, agent_model_instance: BaseLLM,
                      agent_loop: AgentLoop):
-        agent_message_unit_price = agent_model_instance.get_tokens_unit_price(MessageType.HUMAN)
-        agent_message_price_unit = agent_model_instance.get_price_unit(MessageType.HUMAN)
+        agent_message_unit_price = agent_model_instance.get_tokens_unit_price(MessageType.USER)
+        agent_message_price_unit = agent_model_instance.get_price_unit(MessageType.USER)
         agent_answer_unit_price = agent_model_instance.get_tokens_unit_price(MessageType.ASSISTANT)
         agent_answer_price_unit = agent_model_instance.get_price_unit(MessageType.ASSISTANT)
 
         loop_message_tokens = agent_loop.prompt_tokens
         loop_answer_tokens = agent_loop.completion_tokens
 
-        loop_message_total_price = agent_model_instance.calc_tokens_price(loop_message_tokens, MessageType.HUMAN)
+        loop_message_total_price = agent_model_instance.calc_tokens_price(loop_message_tokens, MessageType.USER)
         loop_answer_total_price = agent_model_instance.calc_tokens_price(loop_answer_tokens, MessageType.ASSISTANT)
         loop_total_price = loop_message_total_price + loop_answer_total_price
 
@@ -290,6 +308,10 @@ class ConversationMessageTask:
                 db.session.commit()
             self.retriever_resource = resource
 
+    def on_message_replace(self, text: str):
+        if text is not None:
+            self._pub_handler.pub_message_replace(text)
+
     def message_end(self):
         self._pub_handler.pub_message_end(self.retriever_resource)
 
@@ -342,6 +364,24 @@ class PubHandler:
             self.pub_end()
             raise ConversationTaskStoppedException()
 
+    def pub_message_replace(self, text: str):
+        content = {
+            'event': 'message_replace',
+            'data': {
+                'task_id': self._task_id,
+                'message_id': str(self._message.id),
+                'text': text,
+                'mode': self._conversation.mode,
+                'conversation_id': str(self._conversation.id)
+            }
+        }
+
+        redis_client.publish(self._channel, json.dumps(content))
+
+        if self._is_stopped():
+            self.pub_end()
+            raise ConversationTaskStoppedException()
+
     def pub_chain(self, message_chain: MessageChain):
         if self._chain_pub:
             content = {
@@ -443,3 +483,7 @@ class PubHandler:
 
 class ConversationTaskStoppedException(Exception):
     pass
+
+
+class ConversationTaskInterruptException(Exception):
+    pass

api/core/data_loader/file_extractor.py

@@ -3,7 +3,7 @@ from pathlib import Path
 from typing import List, Union, Optional
 
 import requests
-from langchain.document_loaders import TextLoader, Docx2txtLoader
+from langchain.document_loaders import TextLoader, Docx2txtLoader, UnstructuredFileLoader, UnstructuredAPIFileLoader
 from langchain.schema import Document
 
 from core.data_loader.loader.csv_loader import CSVLoader
@@ -20,13 +20,13 @@ USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTM
 
 class FileExtractor:
     @classmethod
-    def load(cls, upload_file: UploadFile, return_text: bool = False) -> Union[List[Document] | str]:
+    def load(cls, upload_file: UploadFile, return_text: bool = False, is_automatic: bool = False) -> Union[List[Document] | str]:
         with tempfile.TemporaryDirectory() as temp_dir:
             suffix = Path(upload_file.key).suffix
             file_path = f"{temp_dir}/{next(tempfile._get_candidate_names())}{suffix}"
             storage.download(upload_file.key, file_path)
 
-            return cls.load_from_file(file_path, return_text, upload_file)
+            return cls.load_from_file(file_path, return_text, upload_file, is_automatic)
 
     @classmethod
     def load_from_url(cls, url: str, return_text: bool = False) -> Union[List[Document] | str]:
@@ -44,24 +44,34 @@ class FileExtractor:
 
     @classmethod
     def load_from_file(cls, file_path: str, return_text: bool = False,
-                       upload_file: Optional[UploadFile] = None) -> Union[List[Document] | str]:
+                       upload_file: Optional[UploadFile] = None,
+                       is_automatic: bool = False) -> Union[List[Document] | str]:
         input_file = Path(file_path)
         delimiter = '\n'
         file_extension = input_file.suffix.lower()
-        if file_extension == '.xlsx':
-            loader = ExcelLoader(file_path)
-        elif file_extension == '.pdf':
-            loader = PdfLoader(file_path, upload_file=upload_file)
-        elif file_extension in ['.md', '.markdown']:
-            loader = MarkdownLoader(file_path, autodetect_encoding=True)
-        elif file_extension in ['.htm', '.html']:
-            loader = HTMLLoader(file_path)
-        elif file_extension == '.docx':
-            loader = Docx2txtLoader(file_path)
-        elif file_extension == '.csv':
-            loader = CSVLoader(file_path, autodetect_encoding=True)
+        if is_automatic:
+            loader = UnstructuredFileLoader(
+                file_path, strategy="hi_res", mode="elements"
+            )
+            # loader = UnstructuredAPIFileLoader(
+            #     file_path=filenames[0],
+            #     api_key="FAKE_API_KEY",
+            # )
         else:
-            # txt
-            loader = TextLoader(file_path, autodetect_encoding=True)
+            if file_extension == '.xlsx':
+                loader = ExcelLoader(file_path)
+            elif file_extension == '.pdf':
+                loader = PdfLoader(file_path, upload_file=upload_file)
+            elif file_extension in ['.md', '.markdown']:
+                loader = MarkdownLoader(file_path, autodetect_encoding=True)
+            elif file_extension in ['.htm', '.html']:
+                loader = HTMLLoader(file_path)
+            elif file_extension == '.docx':
+                loader = Docx2txtLoader(file_path)
+            elif file_extension == '.csv':
+                loader = CSVLoader(file_path, autodetect_encoding=True)
+            else:
+                # txt
+                loader = TextLoader(file_path, autodetect_encoding=True)
 
         return delimiter.join([document.page_content for document in loader.load()]) if return_text else loader.load()