Prompts API reference - LLM Pulse

Prompts

The questions we run against the AI models every week, and the execution records each run produces.

GET /dimensions/prompts

Lists prompts in a project.

Params:

GET /dimensions/prompt_executions

Lists prompt executions in a project.

Params:

GET /dimensions/query_fan_outs

The sub-queries a model actually issued when answering your tracked prompts, useful for finding the phrasing models search for rather than the phrasing you wrote. view=query (default) returns one row per distinct sub-query with its count and share of all occurrences; view=prompt returns one row per prompt with how many distinct sub-queries it produced. Fan-out is reported mainly by ChatGPT, so an empty result usually means the models in scope do not expose it.

Params:

POST /prompts

Create up to 100 prompts per request. Existing prompts (same text + locale) are skipped.

Body (JSON):

DELETE /prompts/:id

Delete a prompt. Irreversible; the prompt disappears immediately, frees a prompt slot, and its history (executions, mentions, citations, sentiment) is purged by a background job. Requires a writable API key.

Params:

Returns ERR_NOT_FOUND for prompts outside the project.