Documentation Index
Fetch the complete documentation index at: https://docs.secapi.ai/llms.txt
Use this file to discover all available pages before exploring further.
Freshness and trust
Trust is not a slogan here. It is the combination of source-backed answers, explicit freshness, stable contracts, and enough metadata to keep the trail intact after the data leaves the API.
Measurement policy
Source publish time
Use the upstream SEC publish timestamp, such as
acceptanceDateTime, as the source-side clock.API availability
Measure the first successful Datastream resolution of the workflow being benchmarked.
Current trust posture
- freshness dashboard status is currently
monitor, with form-level telemetry and alert counts published in the checked-in artifact - benchmark validation passes for the current dated investor-agent suite
- FinanceBench evaluation: 150/150 (100%) on the full PatronusAI FinanceBench corpus as of 2026-04-04
- FIGI-aware company-resolution inputs are live, while canonical-store completeness proof is tracked separately from the public claim set
- page-level filing traceability is live for filing-derived datapoints, with HTML-first exact fact-region anchors now available on supported segmented revenue traces
- non-filing trace hydration now extends to SEC enforcement source records, market price snapshots, and analyst estimate datapoints through the same
/v1/tracessurface - sentiment remains intentionally gated and is not part of the public 3.0 surface
What makes a response trustworthy
Provenance
If OMNI gives you a structured answer, it should also tell you which filing or source object it came from.
Freshness
If OMNI gives you a current answer, it should also tell you when the source was published and when OMNI last refreshed it.
Materialization
If a parser or normalization pass shaped the output, the version should be visible.
Failure posture
If a workflow is stale or degraded, the response should say so instead of pretending nothing is wrong.
Current health posture
- freshness dashboard status:
monitor - freshness claim posture: use the dashboard artifact and form-level telemetry, not generic real-time language
- benchmark scorecard posture:
18 wins,0 losses,2 tieson the current dated suite - traceability posture: filing-derived datapoints can resolve to the correct filing document and page, supported segmented revenue traces can preserve exact HTML fact ranges, SEC enforcement releases expose shared non-filing trace records, and market prices and analyst estimates carry full source-record traces with provider attribution
- remaining advisory caveat: some freshness signals still publish warnings instead of silently disappearing from the trust story
XBRL facts architecture
Financial data queries (facts and statements) use a DB-first architecture. Over 95 million pre-parsed XBRL facts across 9,600+ SEC-registered companies are stored in our own PostgreSQL database and refreshed daily via an automated Dagster pipeline at 07:00 UTC. The query path:- Check if the company has data in the
xbrl_factstable - If data exists, query directly from the database (fast path)
- If no data exists, fall back to live SEC EDGAR fetch (slow path)
source field in its provenance or completeness object: "xbrl_facts" for database-served responses, "company_facts" for SEC-served responses.
Source artifacts
ops/freshness-telemetry/latest.jsonops/freshness-dashboard/latest.jsonops/external-reference-validation/latest.jsonbenchmarks/competitive/results/scorecard.jsondocs/strategy/omni-datastream-v3-evidence-matrix.md