Backing services run in Docker; the API runs locally against them.

git clone https://github.com/shepdogg6t7-glitch/atlasops-fullstack.git
cd atlasops-fullstack
cp .env.example .env             # edit values if needed
docker compose up -d             # Postgres, Redis, MinIO, Redpanda (ports bound to localhost)
python -m venv .venv && source .venv/bin/activate
pip install -r apps/api/requirements.txt
uvicorn apps.api.main:app --reload

The API is then available at http://localhost:8000. On first run, the SentenceTransformer model downloads automatically (~90MB).

To run the web app in a second terminal:

cd apps/web
npm install
npm run dev

Open http://localhost:3000 to register/sign in, create a project, and upload documents. Registration is enabled for local development by .env.example; set ALLOW_REGISTRATION=false in .env to disable new account creation. The API defaults to denying registration when the setting is unset. For a new deployment, temporarily enable registration to create the first account, then disable it and restart the API. Registration is a public on/off switch, not an invite-code system.

Deploy the interactive demo on Vercel

Import this GitHub repository in Vercel and set the project Root Directory to apps/web; Vercel will detect Next.js and use the existing build settings. The deployed page opens in demo mode and searches bundled sample passages without an account or backend. Live sign-in and PDF uploads require a publicly reachable AtlasOps API: set NEXT_PUBLIC_API_BASE_URL in the Vercel project environment variables and configure that API's CORS_ORIGIN for the deployed Vercel domain. Do not point a public deployment at localhost; that address refers to each visitor's own device.

Uploads must be PDFs and are limited to 25 MB by default (MAX_UPLOAD_MB). Since the API checks the file after multipart parsing, also configure a request-body limit on any reverse proxy. Rate limits are enabled by default; the in-memory counter is per API process. Set RATE_LIMIT_STORAGE_URI to a shared Redis URL when running multiple workers (for the local quickstart, redis://localhost:6379/0). Rate limiting fails closed: if Redis is unavailable, limited requests receive 503 until it recovers. Configure Uvicorn's proxy headers with only your trusted proxy addresses when deployed behind a proxy.