The
evaluators functions are currently in BETA. The API may change without notice. A one-time warning is emitted on first use.List Evaluators
Create a Template (LLM-as-Judge) Evaluator
Create a Code Evaluator
Use a built-inMANAGED code check (e.g. JSON_PARSEABLE, MATCHES_REGEX, CONTAINS_ANY_KEYWORD, CONTAINS_ALL_KEYWORDS, EXACT_MATCH) or a CUSTOM user-supplied Python evaluator.
Create a Remote Evaluator
Remote evaluators call a customer-hosted HTTP endpoint on each evaluation run. The endpoint must already exist as anEVALUATOR integration created via createIntegration (or POST /v2/integrations). The enableRemoteEvalTasks feature flag must be enabled on the account.
Get an Evaluator
Get an Evaluator Version
Update an Evaluator
At least one ofname or description must be provided.
Delete an Evaluator
This operation is irreversible — it deletes the evaluator and all its versions.Manage Versions
List Versions
Create a New Template Version
Versions are immutable once created. To change configuration, create a new version.Create a New Code Version
Create a New Remote Version
The new version becomes the latest immediately. The integration is shared across evaluator versions — updating it affects all versions that reference it.Delete Versions
Deletes a batch of evaluator versions by their IDs (1-100 per call). This operation is irreversible but partial-tolerant: versions that belong to the evaluator are deleted, requested IDs that were not deleted are reported back, and re-submitting already-deleted IDs is safe.DeleteEvaluatorVersionsResult with completed (boolean), deletedVersionIds (string[]), and notDeletedVersionIds (string[]).