Skip to main content
The workflow that makes Prometheus different: never quote a number you can’t cite. This guide traces NVIDIA’s quarterly revenue to its source filing.

1. Resolve the company

Take the ticker (NVDA) or, better, the CIK (1045810) — the CIK survives ticker changes. For a non-US issuer take the platform key the payload returns as cik (XSAU000001), which is the only unambiguous handle. See Coverage.

2. Pull the statements

One call returns the standardized income statement, balance sheet and cash flow for up to 200 periods. Note two things about the cell you want: its row key (std:revenue) and its column’s period endperiod_meta["Q3 2025"].end_date is 2024-10-27.

3. Trace the line

The response carries the XBRL concept the value was standardized from, the accession number of the filing it was read from, and the raw fact with its context, unit and precision. Notice the filing: the value for the quarter ended 2024-10-27 comes from the 10-Q filed on 2025-11-19 — the following year’s Q3 report, whose comparative column is the latest restated statement of that quarter. That is the default vintage, and it is what the citation must say.

4. Cite it

Q3 FY2025 revenue: $35.08B — us-gaap:Revenues, 10-Q 0001045810-25-000230 (filed 2025-11-19), quarter ended 2024-10-27, latest-restated basis.
That’s the whole discipline.

Variations

  • A derived quarter. Ask for end_date=2025-01-26 (NVIDIA’s Q4 FY2025) and provenance is derived_q4, with composition showing FY − Q1 − Q2 − Q3 and each input cell you can trace in turn.
  • What the company originally filed. Pull the same statement with GET /companies/NVDA/financials/as-filed?statement=income_statement&period_type=quarterly&vintage=original — each period from the filing the company made for it, cell by cell with its docRef.
  • A segment or KPI. The row keys in financials.segments_kpis.data (seg::…, kpi::…) are fact-source keys too, with statement=segments_kpis. See Segments & KPIs.
  • A non-US issuer. For Almarai (XSAU000001) the same call returns the PDF statement’s page and printed text in place of an XBRL context, and filing.document_ref in place of an accession.
Building an agent? skill.md encodes this exact loop as a playbook, and every rule in agents.md assumes it.