QA-Pilot

Introduction to QA-Pilot

QA-Pilot is a pioneering project designed to facilitate an interactive chat experience with GitHub code repositories using advanced local and online language learning models (LLMs). It provides users with a rapid and intuitive way to understand and navigate code, aiming to streamline the coding process and enhance user interaction.

Key Features

• Interactive GitHub Chat: Users can chat with public GitHub repositories by simply cloning them, making it easier to understand and interact with complex codebases.

• Chat History Storage: QA-Pilot allows users to store chat histories, enabling them to revisit previous conversations for reference and continued learning.

• Easy Configuration: The project offers straightforward setup options, making it accessible and user-friendly for diverse audiences.

• Multiple Chat Sessions: Users can manage multiple chat sessions simultaneously, enhancing productivity and allowing for better project management.

• Quick Session Search: With a robust search function, users can locate their sessions efficiently, saving time and effort.

• CodeGraph Integration: QA-Pilot integrates with CodeGraph, enabling users to view Python files within the interface for a holistic coding experience.

• Diverse LLM Support: The project supports various LLM models, including:

  • Ollama models like llama3.1, phi3
  • OpenAI models including gpt-4 and gpt-3.5-turbo
  • MistralAI models such as mistral-tiny and codestral
  • LocalAI and Nvidia models
  • ZhipuAI, Anthropic, Llamacpp, Tongyi, and Moonshot

Recent Updates and Enhancements

QA-Pilot continuously evolves, incorporating the latest techniques and tools:

• July 3, 2024: Integrated the Moonshot API and updated the Langchain to version 0.2.6.
• June 2024: Added Go CodeGraph support, included Nvidia/Tongyi API, and introduced Llamacpp features.
• May-June 2024: Emphasized API support for various models, refactored functions, and transitioned from Flask to FastAPI for better performance.

Deployment Guide

For deploying QA-Pilot, follow these steps:

1. Clone the Repository: Use git to clone the QA-Pilot repository and navigate into it.

2. Set Up Environment: Install conda, create a virtual environment, and activate it.

3. Install Dependencies: Use pip to install required packages and set up PyTorch with CUDA.

4. Configure LLM Setups:

   • Ollama: Follow guidelines on their website and GitHub to manage local LLM models.
   • LocalAI: Use Docker to manage local LLM with appropriate base URLs.
   • Llamacpp: Upload models to specified directories and configure them.

5. API Key Setup: Place API keys for various models in the .env file.

6. Database Configuration: Set up a PostgreSQL environment for storing chat sessions, configure connections, and test them.

7. Frontend Setup: Install Node.js and set the frontend environment in a separate terminal.

8. Run the QA-Pilot Backend: Execute the backend program to begin using QA-Pilot.

Usage Tips

• Avoid using URLs and uploads simultaneously.
• Manual removal of local chromadb is required upon stopping its use.
• Use the New Source Button to add new projects.
• Specific commands such as rsd: and rr: facilitate advanced search and document retrieval.
• Click on Open Code Graph for viewing code in supported languages like Python and Go.

QA-Pilot represents an exciting venture into the world of interactive coding assistance, harnessed by cutting-edge LLMs and comprehensive code analysis tools. It's a powerful resource for developers looking to enhance their coding capabilities and streamline their interactions with code repositories.