Skip to content

README.md

Latest commit

 

History

History
101 lines (72 loc) · 4.07 KB

README.md

File metadata and controls

101 lines (72 loc) · 4.07 KB

English / 简体中文

AutoGen Enhanced Project with 9 Agents

Project Introduction

This project, Codegen, built on Microsoft AutoGen, implements 9 efficiently collaborating agents to support complex code requirement processing, code self-execution, and iteration. The project stands out for its unique combination of features, particularly in advanced code processing and iteration. The frontend is built with the Flask framework, and the backend uses React, ensuring user-friendly interfaces and efficient backend processing.In the future, we will focus on program level multi program file generation and interpertation. codegen

Key Features

  • Code Team Collaboration: Effective handling of complex code requirements through the collaboration of 9 agents.

  • EDA: Explore Data Analyse EDA

  • Code Self-Execution and Iteration: Implements automatic code execution and feedback-based iterative improvement. fix bug

  • Document Reading and Understanding: Utilizes RAG (Retrieval-Augmented Generation) functionality to effectively process and understand a large volume of documents. rag

  • website scraping website_scraping

  • Image Processing Capability: Capable of reading and interpreting images, supporting various image analysis scenarios.

  • Support for Large Language Model (LLM) URLs Large language models can utilize the GPT-4 API (recommended) or the free LLM URLs (described later).

  • Code Download and Retrieval (In Progress)

  • Batch Code File Interpretation (In Progress)

  • Cross-File Code Debugging and Search (In Progress)

  • Image Generation Interface (In Progress)

For more details check the demo videos in our file.

Technology Stack

  • Frontend: Flask
  • Backend: React

Getting Started

Follow these steps to set up and run the project:

Clone the repository:

git clone https://github.com/Xin-Ray/CodeGen.git

Installing Dependencies:

  • **Frontend: Backend:
cd path/to/backend/autogen_modifi
pip install .

Starting the Server:

  • Frontend:
npm start
  • Backend:
Copy code
python flask_websockeGPT.py

How to Use LLM's URL

Method 1: You can download a large language model (LLM) locally using tools like LM Studio, start a server, and then copy the server link. Paste this link into our page and you can leave the model_name field blank.

Method 2: Run Mistral7B.ipynb in Colab to obtain the link. Copy and paste this link into our page, and you can also leave the model_name field blank.

Note: The functionality of Mistral7B is not stable, as it has a smaller number of parameters and weaker inference capabilities, making effective team collaboration difficult.

**Motivation：**The initial purpose of creating the LLM URL feature was driven by the high cost of GPT-4. We hope that in the future, there will be open-source and free large language models with fewer parameters that can run on local computers and have inference capabilities surpassing humans. Such models could be directly integrated into our framework to achieve better scalability.

Contribution Guide

We welcome all forms of contributions, whether suggestions for new features, code submissions, or issue reports. Please follow these steps:

  1. Fork the project repository.
  2. Create a new feature branch (git checkout -b my-new-feature).
  3. Commit your changes (git commit -am 'Add some feature').
  4. Push to the branch (git push origin my-new-feature).
  5. Create a new Pull Request.

License

This project follows the MIT License.

References

AutoGen official website: https://microsoft.github.io/autogen/docs/Contribute AutoGen GitHub: https://github.com/microsoft/autogen

Contact

Currently seeking job opportunities For any questions or suggestions, please contact [xxiang@mail.yu.edu].

Development Team

Yeshiva University Codegen Team lead by xinxiang

Instructor：Professor Honggang Wang

group member:Deepa, Manish, Pinxue Lin, Xin Xiang