Groundedness Checks

Groundedness checks score an LLM response against a set of source passages to estimate how well the response is grounded in the provided context. The default scorer uses deterministic token-overlap coverage, so it runs without calling an external embedding model or LLM. All groundedness endpoints require the caller to be an organization owner or admin.

Base URL

For local development:

Endpoints

Methods

Run a check

Request body

Response

List historical checks

Response

Error codes