Search insights
Authentication
Bearer authentication of the form Bearer <token>, where token is your auth token.
Request
Natural-language query, up to 512 characters. Results are ranked by relevance to it. Whitespace-only is treated as absent. For long research questions, use /v1/insights/ask.
Filter by claim type: thesis = a forward-looking argument, observation = a qualitative read on the present, data_point = a concrete number. Multiple values are ORed (any-of). Omit for all three.
Company-specific sentiment pairs. A hit matches when at least one requested pair holds on a SINGLE mention — “bearish about NVDA”, not “mentions NVDA and is bearish about something”. Multiple pairs are ORed (any-of), like every other list filter.
Filter by what the speaker does professionally: operator runs a business, investor deploys capital, researcher produces research, commentator is a journalist or host, government is a public official. unknown when the profession can’t be determined. Multiple values are ORed (any-of).
Earliest published_at to include, inclusive. Accepts a bare date (2026-07-18) or a full timestamp.
Latest published_at to include, inclusive — a bare date covers the whole day.
Fetch specific insights by id (format in-...) — e.g. refreshing previously saved results. Ids that no longer resolve are omitted — compare the result count to the ids you sent.
Scope the search to specific episodes — episode_id values.
Scope the search to specific shows — id values from the podcast directory. “Everything this show has said about X” is one query.
relevance (the default when query is set) or recency (newest first — the default when query is absent).
Results per page, up to 25. Further results are reachable via next_cursor.
Opaque pagination token from a previous response’s next_cursor. Send it ALONE — the original query’s parameters are replayed server-side, and combining cursor with any other field is a 400. Cursors expire after ~15 minutes.
Attach the verbatim transcript passage to each hit as quote. A hit whose transcript is unavailable carries quote: null and the search succeeds regardless.
Response headers
Response
Ranked insights, most-relevant first — newest first when no query is set.
Server-side time for this request, in milliseconds.
Pass back as cursor to fetch the next page. null on the last page — read pagination_end for why paging stopped.
Why paging stopped — null while more pages exist. end_of_results: every match was returned. ranking_exhausted: the relevance ranking reached its ranked window (~150 results; ~500 with a mentions filter) — narrow the query for more depth.