Data Coverage
The corpus is built from US public-company earnings-call transcripts — semantically indexed and continuously updated.
Coverage is a single, system-wide corpus of public market data — roughly 4,600 documents across about 303,000 indexed segments today, and growing as new calls land. It’s shared across all accounts; there is no per-customer or private data in the service.
Earnings-call transcripts#
Quarterly earnings calls, speaker-segmented (prepared remarks and Q&A), so retrieval can return exactly what management or an analyst said. Coverage spans the major US indices and an expanding watchlist universe:
- S&P 500
- Nasdaq 100
- Dow Jones Industrial Average
- Plus additional tickers added by demand
Filings#
source_types filter and the documentType / filingType fields already exist for this — so adding filings won’t change the contract. Tell us if filings coverage would unblock your use case and we’ll prioritize it.sourceUrl back to that document, so your users can verify the original.Filterable metadata#
Every document is tagged with metadata you can scope retrieval by (via the Retrieve endpoint’s filters) and that comes back on every chunk’s source:
ticker— company symbol (e.g.NVDA)yearandquarter— fiscal perioddocumentType— currentlyearnings_call; you can scope by it with thesource_typesarray, e.g.["earnings_call"]filingType— reserved for future filing types (e.g.10-K);nullfor transcriptspageNumbers— where the passage appears, for paginated source types (omitted for transcripts)
Freshness#
New transcripts are ingested on an ongoing basis. If you query a period that hasn’t been ingested yet, the API serves the nearest prior period and flags it in meta.periodMismatch rather than returning nothing — see Period fallback.