evrz hackathon

Telegram Review Bot

A Python Telegram bot that processes archive files, extracts project structures, parses code, and generates review report based on the provided knowledge base.

Features

• File Support: Accepts various file formats.
• Archive Support: Accepts RAR, ZIP, and 7z archive formats.
• Review Preview: Sends paginated preview of the review results.
• Review Download: Provides a button to download the full review report as a PDF file.
• RAG: Uses RAG to provide context for the model.
• Small LLM: Uses Mistral-nemo 12B model for delivering comprehensive review requiring less than 20GB of VRAM.

Setup

Prerequisites

• Python 3.11+
• Poetry for dependency management. Install from Poetry's official site
• MinIO Server with DNS record for pre-assigned URLs (credentials will be provided separately)
• System Dependencies:
  • For RAR extraction, ensure that unrar is installed on your system:
    • Ubuntu: sudo apt-get install unrar
    • macOS (with Homebrew): brew install unrar
    • Windows: Is not supported due to the lack of supported libraries.

Important Note for Local Development

If running the bot locally (outside Docker), you'll need to handle PyTorch installation separately:

• Either run: pip install torch==2.1.2 --index-url https://download.pytorch.org/whl/cpu
• Or uncomment the torch dependency in pyproject.toml and try to install it, but it may fail.

Installation

1. Clone the Repository

   git clone https://github.com/isofinly/telegram_review_bot.git
   cd telegram_review_bot

2. Configure Environment Variables Create a .env file with the following required variables:

   BOT_TOKEN=your_telegram_bot_token
   MINIO_ENDPOINT=your_minio_endpoint
   MINIO_ACCESS_KEY=your_minio_access_key
   MINIO_SECRET_KEY=your_minio_secret_key
   MINIO_SECURE=true_or_false
   MODEL_API_KEY=your_model_api_key
   

   Note: MinIO credentials will be provided via manager and cloud drive document.

3. Choose Your Installation Method

   Docker (Recommended):

   docker-compose up -d

   This will start both the bot and MinIO server containers.

   Local Installation:

   poetry install
   poetry run telegram-review-bot

Usage

1. Start the Bot

   Open Telegram and search for your bot. Start a conversation by sending /start.

2. 2.1 Send a File

   Send a file to the bot.

   2.2 Send an Archive

   Or you can send an archive file (RAR, ZIP, 7z) to the bot.

3. Receive Outputs

   • Review Preview: The bot will reply with paginated preview of the review results.
   • Review Download: Use the provided button to download the full review report as a PDF file.

Review pipeline

1. RAG: Create a vector database from the knowledge base to align code chunks with relevant review examples.
2. Review:
   • Parse project structure
   • Analyze the structure of individual files.
   • Divide files into manageable chunks.
   • Use RAG to generate prompts and contextual information for the model.
   • Generate and format the review

Packages structure

• src/bot — Telegram bot code.
• src/review — Code review logic.
• • src/review/parsers - Code chunking logic

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

MIT