AI Codebase Reviewer

An AI-powered codebase review system that analyzes uploaded project archives (.zip) using tree-sitter for parsing and Groq (Llama) for generating structured feedback. The system uses an asynchronous architecture with Celery and Redis to handle large projects efficiently.

Features

• Asynchronous Processing: Non-blocking API that handles project analysis in the background via Celery.
• Project Uploads: Accepts .zip archives containing project source code.
• Multi-language Support: Parses Python, JavaScript, TypeScript, Java, C, and C++ files using tree-sitter.
• Intelligent Chunking: Chunks code by parsing functions and classes to stay within LLM context limits.
• Intelligent Caching: Uses Redis to cache analysis results for code chunks (SHA-256), preventing redundant LLM calls.
• Job Tracking: Persistent job status and results stored in SQLite (or PostgreSQL).
• AI-Powered Analysis: Utilizes Groq (Llama 3.1) to identify bugs, security vulnerabilities, performance bottlenecks, and structural warnings.

Tech Stack

• Backend: FastAPI, Python 3.12+
• Task Queue: Celery + Redis
• Database: SQLAlchemy (SQLite default)
• Parsing: tree-sitter
• AI Integration: groq Python SDK
• Dependency Management: Poetry

Prerequisites

• Python 3.12 or higher
• Poetry
• Redis (for Celery and Caching)
• Groq API Key

Installation

1. Clone the repository:

   git clone https://github.com/falakrana/AIReviewBot.git
   cd AIReviewBot

2. Install dependencies:

   poetry install

3. Set up environment variables (.env):

   GROQ_API_KEY=your_groq_api_key_here
   REDIS_URL=redis://localhost:6379/0
   DATABASE_URL=sqlite:///./storage/aireviewbot.db

Usage

You need to run two processes:

1. Start the API Server:

   poetry run python -m backend.app.main

2. Start the Celery Worker:

   # Windows
   poetry run celery -A backend.app.workers.celery_app.celery_app worker --loglevel=info --pool=solo
   
   # Linux/Mac
   poetry run celery -A backend.app.workers.celery_app.celery_app worker --loglevel=info

API Reference

• POST /api/v1/analyze: Submit a ZIP file for analysis (returns job_id).
• GET /api/v1/status/{job_id}: Check analysis progress.
• GET /api/v1/result/{job_id}: Fetch the final analysis report.
• GET /api/v1/jobs: List recent analysis jobs.

Documentation

For a detailed walkthrough of the architecture and setup, see walkthrough.md.