Products

Catalog Attribute Normalizer

Clean, normalized product attributes and taxonomy mappings from messy multi-source catalogs. You supply the product data; we canonicalize size/color/attributes to a fixed vocabulary and classify each product into the Google and Shopify taxonomies.

What makes it different: category classification is retrieval-grounded against the real, current taxonomy files — not recalled from the model’s memory — so it returns a genuine leaf_id or an honest null, never a plausible-but-fabricated ID. Measured against a live model API at 22 of 24 checks (91.7%) exact path + leaf-ID matches on our 12-product evaluation set — each product checked against both Google and Shopify. A later live check through the deployed service itself scored 11 of 12 on the Google side of that same 12-product evaluation set — one check per product. Accuracy tracks input quality: give it a bare title and it says so with lower confidence; give it the full record and it’s confident and correct.

Receipt Extraction (AU GST-aware)

Turn a receipt or invoice image/PDF into structured data — vendor, ABN, GST-aware line items and totals.

What makes it different: it’s built against the real ATO GST rules, not just “find the total.” Each line item carries a tax_code (taxable / gst_free / entertainment_denied), a confidence, and a one-line rationale naming the actual rule — e.g. “entertainment ITC denied, no exception context” or “entertainment incidental to travel exception applies.” It distinguishes a client dinner (ITC denied) from an in-flight meal on a business trip (claimable) even though the receipts look alike. ABN checksum validation is built in.

GST handling reflects publicly documented Australian tax law and is provided as software output, not tax advice.

x402 Bazaar Listing Monitor

Sellers who settle real x402 payments have no reliable way to confirm their resources stay discoverable in Coinbase CDP’s Bazaar catalog. The discovery API has a recurring, publicly reported class of bug where a resource silently stops appearing despite confirmed settlements — and sellers only catch it by re-checking by hand, or by noticing a revenue dip days later.

What makes it different: it is the checking you would otherwise do yourself, kept up. Register a payTo address and the resource URLs you serve; every six hours the poll diffs them against the live discovery catalog and records what changed, so you have a timestamped history to cite if you need to escalate. The value is continuous vigilance rather than a one-off answer — the same check run on a schedule you do not have to maintain.

Outbound alert delivery is not enabled yet: changes are detected and recorded with a timestamped history, and delivery is being wired next.

Safe Sun Window

A free web app that answers the two halves of the same daily question: roughly how long your skin type can take today’s sun before burn risk rises, and the shorter window relevant to vitamin-D synthesis. Pick your Fitzpatrick skin type, allow location or type today’s UV index, and it computes both.

What makes it different: it computes published models and shows the arithmetic — Fitzpatrick phototype scales, the WHO UV index, peer-reviewed sun-exposure models — with the sources cited on the page, so you can check where each number came from. The estimates are labelled as estimates: skin varies by roughly ±30–50% within a type, UV varies with cloud, altitude and season, and the app says so rather than presenting a window as a certificate. There is also a reference set covering every UV index from 3 to 11, each with the full type I–VI table computed at that level and the calculation laid out.

Estimates computed from published models and the UV reading you enter or allow the app to fetch. It does not diagnose skin type, does not promise safety, and is not medical advice.

Mediterranean Diet Score

A free web app built around MEDAS-14 — the fourteen-item screener developed and validated for the PREDIMED trial. Answer the fourteen questions, get the standard 0–14 score, and see exactly which criteria you met and which you missed.

What makes it different: it applies the published MEDAS-14 criteria rather than a food diary. No barcode scanning, no meal photos, no daily logging — about two minutes, once a week. The per-item breakdown is the useful output: not a grade, a list of the specific thresholds you didn’t clear. Re-check weekly and the trend renders locally from your browser’s own storage, so there is no account to make and no server copy of your answers.

The app applies the published MEDAS-14 scoring criteria and cites its sources. It does not diagnose anything, does not prescribe anything, and a score is not medical advice.