codefuse chatbot

Other source code

1.0.0

Download

Chinese | English

CodeFuse-ChatBot: Development by Private Knowledge Augmentation

CodeFuse-ChatBot is an open source AI intelligent assistant developed by the Ant CodeFuse team, dedicated to simplifying and optimizing all aspects of the software development life cycle. This project combines the collaborative scheduling mechanism of Multi-Agent and integrates a rich tool library, code library, knowledge base and sandbox environment, enabling the LLM model to effectively execute and handle complex tasks in the DevOps field.

? renew

[2024.01.29] Open and configurable multi-agent framework: codefuse-muAgent, please see the instructions for use for details
[2023.12.26] The ability to access open source privatized large models and large model interfaces based on FastChat is open
[2023.12.14] Qubit public account special report: article link
[2023.12.01] Multi-Agent and code base search functions are open
[2023.11.15] Added Q&A enhanced mode based on local code base
[2023.09.15] The sandbox function of the local/isolated environment is opened, and the specified URL knowledge retrieval is implemented based on the crawler.

? introduce

This project aims to build an AI intelligent assistant for the full life cycle of software development through Retrieval Augmented Generation (RAG), Tool Learning and a sandbox environment, covering the stages of design, coding, testing, deployment and operation and maintenance. . Gradually transform from the traditional development and operation model of data query everywhere and independent and decentralized platform operation to the intelligent development and operation model of large model question and answer, changing people's development and operation habits.

The core differentiating technologies and functional points of this project are:

? Intelligent dispatching core: A dispatching core with complete system links has been built to support multi-mode one-click configuration and simplify the operation process. Instructions for use
Whole code library analysis: Achieve in-depth understanding of warehouse-level code, as well as project file-level code writing and generation, improving development efficiency.
? Document analysis enhancement: It integrates the document knowledge base and knowledge graph, and provides deeper support for document analysis through retrieval and reasoning enhancement.
? Exclusive knowledge for vertical categories: An exclusive knowledge base customized for the DevOps field, supporting self-service one-click construction of vertical knowledge bases, which is convenient and practical.
? Vertical model compatibility: Aiming at small models in the DevOps field, it ensures compatibility with DevOps-related platforms and promotes the integration of the technology ecosystem.

? Relying on the open source LLM and Embedding models, this project can realize offline private deployment based on the open source model. In addition, this project also supports OpenAI API calls. Access Demo

The core R&D team has long been focusing on research in the field of AIOps + NLP. We launched the Codefuse-ai project and hope that everyone will widely contribute high-quality development and operation and maintenance documents to jointly improve this solution to achieve the goal of "making development easy in the world."

? Demo video

In order to help you understand the functions and usage of Codefuse-ChatBot more intuitively, we have recorded a series of demonstration videos. You can quickly understand the main features and operating procedures of this project by watching these videos.

Knowledge Base Import and Q&A: Demonstration Video
Local repository import and Q&A: Demo video

?Technical route

? Multi-Agent Schedule Core: Multi-agent scheduling core, which can be easily configured to create interactive agents.
?️ Multi Source Web Crawl: A multi-source web crawler that provides the function of crawling specified URLs to collect required information.
Data Processor: Data processor can easily complete document loading, data cleaning, and text segmentation, and integrate data from different sources.
? Text Embedding & Index:: Text Embedding Index, users can easily upload files for document retrieval, optimizing the document analysis process.
?️ Vector Database & Graph Database: Vector and graph database, providing flexible and powerful data management solutions.
Prompt Control & Management: Prompt control and management accurately defines the context of the agent.
? SandBox:: A sandbox environment to safely perform code compilation and actions.
LLM:: Agent brain, supporting multiple open source models and LLM interfaces.
API Management:: API management tool to achieve rapid integration of open source components and operation and maintenance platforms.

For specific implementation details, please see: Technical route details and project plan follow-up: Projects

Model access

If you need to integrate a specific model, please let us know your needs by submitting an issue.

model_name	model_size	gpu_memory	quantify	HFhub	ModelScope
chatgpt	-	-	-	-	-
codellama-34b-int4	34b	20g	int4	coming soon	link

Quick to use

muagent-py

For complete documentation, see: CodeFuse-muAgent

 pip install codefuse-muagent

Using ChatBot

Please install the nvidia driver yourself. This project has been tested in Python 3.9.18, CUDA 11.7 environment, Windows, and X86 architecture macOS systems.

For Docker installation, privatized LLM access and related startup issues, please see: Quick usage details

For Apple Silicon, you may need to brew install qpdf first.

1. python environment preparation

It is recommended to use conda to manage the python environment (optional)

 # 准备 conda 环境
conda create --name devopsgpt python=3.9
conda activate devopsgpt

Install related dependencies

 cd codefuse-chatbot
# python=3.9，notebook用最新即可，python=3.8用notebook=6.5.6
pip install -r requirements.txt

2. Start the service

 # 完成server_config.py配置后，可一键启动
cd examples
bash start.sh
# 开始在页面进行相关配置，然后打开`启动对话服务`即可

Or start the old version through start.py For more LLM access methods, see more details...

Contribution Guide

Thank you very much for your interest in the Codefuse project. We very much welcome your suggestions, opinions (including criticism), comments and contributions to the Codefuse project.

Your various suggestions, opinions, and comments about Codefuse can be submitted directly through Issues on GitHub.

There are many ways to participate in and contribute to the Codefuse project: code implementation, test writing, process tool improvements, documentation improvements, and more. Any contributions are very welcome and you will be added to the contributors list. See Contribution Guide for details...