vaaya / docs / reference

Open data

Six reference-grade sources, all 1¢ flat: the Internet Archive’s Wayback Machine, Wikipedia + Wikidata, USPTO patents, the GDELT global news firehose, the OpenAlex scholarly graph, and the US Federal Register. These are primary or canonical sources — prefer them over web search when the question is historical, encyclopedic, patent-, regulation-, or registry-shaped, and cite them over aggregators when they conflict.

When to use which

wayback — the internet’s history

Price

1¢ per action

Source

Internet Archive

• wayback/snapshots — list archived captures of a url; optional from/to timestamp prefixes (like 2023) and limit. Returns rows with timestamps to feed to fetch.
• wayback/available — the closest single snapshot to a timestamp (YYYYMMDD). The quick rung.
• wayback/fetch — the archived page itself ({ url, timestamp }), original bytes without archive chrome.

curl -X POST https://vaaya.ai/api/run/wayback/fetch \
  -H "Authorization: Bearer $VAAYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/pricing", "timestamp": "20240101"}'

Use for: what a pricing/docs/landing page said at some date, deleted posts, dead links from old citations, product pivots over time. Compare two snapshots of the same URL to diff messaging. Timestamps are UTC YYYYMMDDhhmmss prefixes.

wikipedia + wikidata — canonical facts & entity resolution

Price

1¢ per action

• wikipedia/search (q, optional lang) → titles; wikipedia/page (title) → the full article as clean plain text — cheaper than scraping.
• wikidata/search (q) → Q-ids. Start here to disambiguate any entity before fanning out to other sources.
• wikidata/entity (id, e.g. Q95) → all structured claims: founding date, HQ, subsidiaries, execs, and cross-registry identifiers (LEI, tickers, Crunchbase, ROR).
• wikidata/sparql (query) → set-shaped answers no search engine can do. Keep queries LIMITed — the upstream timeout is 60s, so small queries only.

uspto — US patents

Price

1¢ per action

Coverage

applications + grant metadata, newest-first

• uspto/patents — q is a plain full-text string ("machine learning") or field-qualified (applicationMetaData.inventionTitle:"neural network"). Optional date_gte/date_lte (effective-filing-date window), limit (≤100), offset.
• uspto/assignees — pass organization; returns that applicant’s filings. The “does X hold patents” lookup before a patents sweep.

Use for patent portfolios in company research, prior-art scans, and who’s patenting in a space. EU trademarks are a different register — /x402/eu-trademark-search (75¢) on Compliance & KYB.

gdelt — global news firehose

Price

1¢ per action

Coverage

65 languages, ~15-min latency

• gdelt/news — search worldwide press. query supports inline operators: sourcecountry:chile, sourcelang:spanish, domain:reuters.com, "exact phrase". Window via timespan (1d, 2w, 3m) or startdatetime/enddatetime (YYYYMMDDhhmmss). maxrecords ≤250.
• gdelt/timeline — coverage-volume or tone series (mode: timelinevol | timelinetone) for narrative tracking and event detection.

curl -X POST https://vaaya.ai/api/run/gdelt/news \
  -H "Authorization: Bearer $VAAYA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "lithium sourcecountry:chile", "timespan": "2w", "maxrecords": 50}'

Use GDELT when the story is non-US/non-English or you need press-coverage shape over time. For mainstream recent headlines, brave/news is still the first rung — see Web search.

openalex — the scholarly graph

Price

1¢ per action

Coverage

250M+ works

• openalex/works — search and/or filter (from_publication_date:2024-01-01,is_oa:true,cited_by_count:>100), per_page ≤50. Metadata + abstracts + citation counts + OA links.
• openalex/work — one work by W… id or DOI: authorships, per-year citations, referenced/related works.
• openalex/authors — disambiguated researcher profiles (h-index, affiliations) — expert-finding and founder/team technical due diligence.

OpenAlex adds the citation graph and filters on top of quick keyword paper lookups — use it when you need “most-cited since X”, author-centric views, or open-access full-text links. For managed multi-step research runs, see Research.

fedreg — US Federal Register

Price

1¢ per action

Coverage

proposed + final rules since 1994

• fedreg/search — term + optional type (RULE | PRORULE | NOTICE | PRESDOCU), agency slug (e.g. securities-and-exchange-commission), date_gte/date_lte, per_page.
• fedreg/document — full metadata by document number: abstract, docket ids, CFR refs, comment deadlines, full-text links.

Use for upstream regulatory signal (rules while still PRORULE), comment-period tracking, and what an agency has done about X. Pairs with the edgar/* actions on Public records for what companies filed in response. EU-side equivalents (eu-regulation-search, eu-ai-act-classify) live on Compliance & KYB.