Reference · How-to · ~15 min

How to deploy a small AI app

Ship a minimal API or chat UI — secrets on the server, logs on, scope small.

Ship a minimal API or chat UI — secrets on the server, logs on, scope small.

Steps

1. **Working locally** — API route calls LLM with env var key

2. **Choose host** — Vercel, Railway, Fly.io, Render, etc.

3. **Set env vars** in dashboard (`OPENAI_API_KEY`, etc.)

4. **Add health check** route (`/health`)

5. **Deploy** from git; smoke-test one prompt

6. **Monitor** errors, latency, token spend

Minimum checklist

  • API keys not in frontend bundle
  • Rate limiting or auth if public
  • Timeouts on LLM calls
  • User-facing error message (not raw stack trace)

Pseudo deploy flow

git push origin main
curl https://your-app.example/health

Level up later

Caching frequent queries, streaming responses, staging environment.

**Try the lessons:** `model-deployment` · `capstone-ship-it`