AI Code Reviewer

Agentic code review system that fetches GitHub PR diffs, runs static analysis tools (Ruff, Bandit, Semgrep), synthesizes structured reviews via LLM, and learns codebase patterns across PRs using pgvector.

Phase 1 — Foundation (current)

Infrastructure and API skeleton are in place. Review pipeline comes in later phases.

Prerequisites

• Python 3.11+
• Docker Desktop

Quick start

# 1. Start PostgreSQL (pgvector) + Redis
docker compose up -d

# 2. Create virtual environment and install dependencies
python -m venv .venv
.venv\Scripts\activate          # Windows
# source .venv/bin/activate   # macOS/Linux
pip install -e ".[dev]"

# 3. Copy environment file
copy .env.example .env          # Windows
# cp .env.example .env          # macOS/Linux

# 4. Run database migrations
alembic upgrade head

# 5. Start the API
uvicorn app.main:app --reload

Endpoints

Method	Path	Description
GET	/health	Liveness — process is running
GET	/ready	Readiness — DB + Redis check
GET	/docs	Swagger UI

Verify

curl http://localhost:8000/health
curl http://localhost:8000/ready

Project structure

See docs/PROJECT_GUIDE.md for the full build plan.