Multi-agent GitHub PR reviews with Tiger Cloud memory, ARQ workers, and a human approval gate.
- Specialists in parallel — security, quality, tests, and docs agents review the diff together, then merge findings.
- Grounded context — retrieves repo code, past reviews, and conventions from Tiger Cloud before judging.
- Safer posting — low-confidence or critical findings go to a HITL queue instead of auto-commenting on GitHub.
- Visible cost — every run is traced with spend and LLM calls so reviews stay budget-aware.
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
python -m src.memory.bootstrapFill in REDIS_URL, GITHUB_TOKEN, OPENAI_API_KEY, and TIMESCALE_SERVICE_URL in .env.
# API
python -m uvicorn src.api.app:app --host 0.0.0.0 --port 8000 --reload
# Worker
arq src.arq_worker.WorkerSettings
# Dashboard
cd dashboard
npm install
npm run dev- API: http://127.0.0.1:8000
- Dashboard: http://localhost:3000
Enqueue a review:
python -m src.enqueue owner/repo 42MIT — see LICENSE.
