jawntrack
Reconstructs a normalized possessions graph — what you own, when you bought it, what it cost, which email proves it — by mining receipt emails. The intake funnel for the Jawnverse. The Gmail source adapter has just landed.
jawntrack (GSV-D5, hexagon pace-layer designation) is a General Systems Ventures project in the media-ecology domain that answers a single question: what do you own, and where did it come from. It mines receipt and order-confirmation emails and normalizes them into a possessions graph, where every item carries full provenance — the item itself, its date, price, merchant, and the specific source email that proves the claim. The design is deliberately narrow in scope: jawntrack performs ingest and provenance only, and explicitly excludes appraisal (valuation and sell/keep decisions, which belong to agentique-roadshow) and resale/storefront functions (which belong to jawncloud). This makes jawntrack the shared ingest funnel that downstream GSV apps read from, so no other project has to re-derive ownership facts from raw email data itself. The premise was validated prior to this repo's existence: an earlier, narrower workflow scanned 385 order emails spanning two years and distilled them into 11 named wearable items with complete provenance, demonstrating that receipt emails are a viable ground-truth source for ownership claims. jawntrack generalizes that proof of concept from wardrobe items to possessions of any kind. Architecturally, the pipeline runs source-specific email adapters (covering Amazon, Shopify, PayPal, Grailed, plus a generic fallback) through a harvest, classify, and extract stage, then normalizes and deduplicates results into the possessions graph, and finally exposes an outbound emit contract for consuming apps like roadshow and jawncloud. Current status: the Gmail source adapter has just landed with an accompanying test file, marking the first live, wired transport in an architecture that until this commit existed only as stubs against documented interfaces rather than connected to a real data source.