Quick start¶
1. Initialize¶
All built-in provider templates begin disabled. In configuration, enable one provider, point its
credential reference at a keyring or environment variable, add a model (or use discovery), and
map default to provider/model.
2. Start¶
The default listener is http://127.0.0.1:8742. Put the generated local token in the client
environment as DIVINE_API_TOKEN; do not copy it into source code.
3. Verify¶
4. Make a request¶
curl http://127.0.0.1:8742/v1/chat/completions \
-H "Authorization: Bearer $DIVINE_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"model":"default","messages":[{"role":"user","content":"Reply with OK"}]}'
The response headers identify the chosen provider, model, route, fallback count, and request ID.
Use an explicit provider/model while validating a new setup.
Background operation¶
Background lifecycle records a PID creation fingerprint and verifies command identity before sending a stop signal, protecting unrelated processes from stale PID reuse.