Evidence Pipeline Enrichment
Expand the fact validation evidence pipeline with authoritative external data APIs. Currently facts are verified only against Wikipedia text and Wikidata triples — these integrations add domain-specific authoritative sources.
API Sources
| Source | Domain | Free Tier | Status |
|---|---|---|---|
| Alpha Vantage | Financial fundamentals, stock prices, economic indicators | 25 req/day | Draft |
| FRED + Finnhub + FMP | Federal Reserve data, ESG, congressional trading, capacity relief | Varies | Draft |
| API-Sports | Sports scores, standings, player stats | Free key active | Draft |
| OpenAlex | Academic papers, citations, institution data | Unlimited (free) | Draft |
| DBpedia | General structured Wikipedia data | Unlimited (free) | Draft |
| Wikimedia Enterprise | Structured Wikipedia with real-time updates | Paid (enterprise) | Draft |
What This Solves
The evidence pipeline currently has zero financial data APIs, no sports data, and limited academic/scientific coverage. These integrations let the validation tier confirm specific claims like:
- "Apple is worth $3 trillion" (Alpha Vantage -> MarketCapitalization)
- "US GDP grew 5.7% in 2021" (FRED -> actual GDP values)
- "Liverpool won the Premier League" (API-Sports -> standings)
- "The study was cited 10,000 times" (OpenAlex -> citation count)
Architecture
Each API gets a validation tier client following the existing pattern in packages/ai/src/validation/. The client normalizes responses into the standard evidence format used by the validation pipeline.
Reference Documents
- alphavantage-evidence-enrichment.md — Company fundamentals, economic indicators, commodities
- financial-evidence-expansion.md — FRED, Finnhub, FMP (extends Alpha Vantage)
- api-sports-evidence-integration.md — Sports data verification
- openalex-evidence-enrichment.md — Academic/scientific evidence
- dbpedia-evidence-enrichment.md — General structured Wikipedia lookups
- wikimedia-enterprise-evidence.md — Enterprise Wikipedia with real-time updates