Embeddings API

Embed text only after an actual executor is selected, configured and qualified. The registered endpoint by itself does not establish that embeddings are available.

1. Open API workbench (/rust/api) and Base URLs, then Authentication to obtain the Panel URL and an Access-issued key.
2. Inspect the configured model inventory at GET /api/llama/v060/models and the capability contract at GET /api/llama/v060/capabilities. Only choose a suitable embedding model and upstream executor after validating their advertised capabilities.
3. In Endpoint catalog, choose POST /v1/embeddings, then open it under Browse every endpoint (generated from route registration) or use Generate examples as an administrator. Replace MODEL_ID and the sample input. The Python, JavaScript and command-line examples use the Panel URL and a bearer token.
4. Optionally validate admission before inference using POST /api/llama/v060/admission with path set to /v1/embeddings and a nested request. Admission is a request-shape and capacity check, not proof of a live embedding executor.
5. Authorized administrators may select a safe endpoint in In-page tester, supply JSON body for POST and press Send test request. A missing upstream executor must return an explicit error.

Example embedding request:

{"model":"MODEL_ID","input":"Sample text"}

Example admission request:

{"path":"/v1/embeddings","request":{"model":"MODEL_ID","input":"Sample text"}}

A successful embedding response comes from the selected executor, not from a synthetic Panel calculation. See Panel API and Compatible clients.

Control reference