> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vendo.run/llms.txt
> Use this file to discover all available pages before exploring further.

# Deploying Vendo to production: a wiring checklist

> Checklist for taking Vendo to production: model, principal resolution, Postgres store, mounted handler, policy or judge, tick schedule, webhooks, sandbox.

Before production:

1. Pass an AI SDK `LanguageModel` to `createVendo`.
2. Resolve every host request to a stable `Principal.subject`.
3. Point `createStore({ url })` at Postgres and run `ensureSchema()`.
4. Mount the fetch handler under one base path and preserve SSE streaming.
5. Configure policy or a judge if the default auto-run posture is not intended.
6. Schedule `/tick` or start the long-lived timer.
7. Register and verify every webhook source.
8. Add a sandbox adapter only if server-backed apps are needed.

Present tool execution forwards the signed-in request's auth material on a
same-origin fetch. Away host API execution requires `actAs`; without it, those
calls fail cleanly with `not-implemented`.

Generated components run in a no-egress iframe. App machines enforce the app's
domain allowlist and receive secret handles, never secret values.

Use `/status` for a live composition probe. Use the public Postgres tables for
backup and retention.