- Clone repo and install dependencies:
npm install- Generate Prisma client:
cd mcp-server
npx prisma generate
cd ..- Copy
.env.exampleto.envand adjust. - (Optional) Start ML service:
cd ml-service
python -m venv .venv
. .venv/Scripts/Activate.ps1
pip install -r requirements.txt
uvicorn app.main:app --port 8000- Run server + ML concurrently:
npm run dev- Build extension:
npm -w extension run buildnpm run dev:mcp– MCP server (ts-node-dev)npm run dev:ml– FastAPI ML servicenpm run dev– Concurrent server + MLnpm run build– Build shared libs, server, extensionnpm test– Unit/integration testsnpm run test:e2e– Playwright tests
- TypeScript strictness maintained; avoid unnecessary any.
- Keep functions small and purposeful.
- Do not commit generated
dist/except for packaged extension release branches.
- Never commit real secrets.
.envis ignored. - Rotate
JWT_SECRETfor production deployments.
- Include description & rationale.
- Ensure
npm testpasses locally. - Run
npm -w shared run buildif modifying shared models.
Follow conventional style (recommended):
feat: add Xfix: correct Ydocs: update READMErefactor: simplify Z
- Add unit tests for new utilities.
- Extend Playwright E2E if UI flows change.
Include reproduction steps, expected vs actual behavior, and environment (OS, Node, Python versions).
Happy hacking!