Accept a proposed improvement (updates the agent's fields)
Applies the change set to the AGENT — exactly like editing it in the Skills/Advanced editor — and this changes how it answers real customers. It does NOT write a raw prompt. Specifically: changes.fields overwrite the agent columns (role, objective→purpose, company_service→company offer, interest_of_product→interacting with, tone→writing style, length_detail→response length, negative_response→action if no answer); changes.guidelines_add create new Response Guideline records; changes.notes_add create Note records; changes.training_add create Training Data records. Then the agent’s prompt is regenerated from those fields (same builder as Skills) and any custom_prompt override is cleared. Works from a ‘proposed’ attempt (no trial needed) or a ‘ready’ one. Reversible with /optimize/revert.
Authorizations
Use a Bearer token to access these API endpoints. Example: "Bearer {your-token}"
Path Parameters
Body
Required. The id of the optimization attempt from GET /{uuid} data.optimizations[].id.
Optional edited change set to apply instead of the proposed one. Same shape as data.optimizations[].changes: fields (object of agent-column → new value), guidelines_add (string[]), notes_add (string[]), training_add (string[]). Omit to apply the proposal as-is; drop a field key or a list item to skip it. Only these keys and known columns are honoured.
Response
Agent fields updated and prompt rebuilt