Data plane · SourceRegistry
Sources & ingestion
Every source obeys the acquisition-tiering invariant · licensed/API → managed scraper → custom miner · Scout extension only for human-assisted no-API sites · no ToS-hostile scrapingActive sources9across 5 access methods
Tier-2 fail rate4.1%managed scraper · rolling 24h
Batches today214bronze immutable → ADLS
Candidate sources2from discovery agent · await triage
Extraction health by tier
nominalTier 1 · structured API / feed99.2% ok
Tier 2 · managed scraper4.1% fail
Tier 3 · Scout extension (human-assist)manual cadence
💡
Tiering holds strictly: a source only drops to a lower tier when the higher one is unavailable. Custom miners run only on sources with no API and no ToS-hostile posture.
Source discovery agent
PlannedDescribe a niche and the agent probes the open web for viable acquisition targets, then emits a Source Dossier (markdown) plus candidate SourceRegistry rows at status=candidate for human triage. v1 will run per active niche on a cadence.
OutputSource Dossier (markdown)
Emitscandidate SourceRegistry rows
Tieringrespects acquisition invariant
Cadence (v1)per active niche