PR Agent

Autonomous code review agent system that uses AI to analyze GitHub pull requests.

Project Setup

1. clone git clone https://github.com/rohit1kumar/pr-agent.git && cd pr-agent

2. create .env using cp .env.example .env and update the values

   OPENAI_API_KEY=your_openai_token

3. run docker-compose up --build to start the services

API Documentation

API docs are available at http://localhost:8000/docs

1. Create task

   curl --request POST \
   --url http://localhost:8000/analyze-pr \
   --header 'content-type: application/json' \
   --data '{
       "repo_url": "<repo_url>",
       "pr_number": "<pr_number>",
   }'
   

2. Get status (use task_id from step 1)

   curl --request GET \
   --url http://127.0.0.1:8000/status/<task_id> \
   --header 'content-type: application/json'

3. Get result (use task_id from step 1)

   curl --request GET \
   --url http://127.0.0.1:8000/results/<task_id> \
   --header 'content-type: application/json'

Design Decisions

1. FastAPI for the API server with redis based rate limiting
2. Celery for background tasks and Redis as a message broker & backend
3. OpenAI API for the AI model
4. Docker for containerization

Future Improvements

1. Adding better AI models for better code analysis
2. Creating GitHub App & responding to PRs using webhooks

Bonus Points Todos

• Docker configuration
• Result caching system
• Structured logging
• Multi-language support
• Rate limiting
• GitHub webhook support
• Live deployment (e.g., Railway, Render, etc.)