Methodology & changelog
How the library is built, what is verified, and what is still unknown. This page is the contract behind the dataset.
Build
- Sources: official product pages, Shopify product catalogs, PDF manuals, certification pages (public pages only; robots respected).
- Record mix: 87 curated full-spec records + 278 catalog-derived records; catalog rows carry price plus whatever specs the storefront published.
- Extraction: deterministic parsing with a verbatim quote stored for every populated field.
- QA gate: PASS — 2277 populated values, 100.0% with provenance.
- Bundle manifest: inputs 3, files 17, generated 2026-11-15T06:47:11Z, credits used 1324.0.
Changelog
| Version | Date | Change |
|---|---|---|
| v2_20261115 | 2026-11-15 | Initial seed build: 87 models / 8 brands from existing NRS assets; supporting tables included. |
Known limitations
- Seed coverage: 32 brands collected so far; expansion to the full market list is in progress.
- Some fields are legitimately absent on vendor pages; missing values are null, never guessed.
- Prices are snapshots with fetch dates; they are not live quotes.
- Source licensing: public data compiled with citation; resale terms to be finalized per source before first sale.
Labels
OBSERVED = counts, coverage, provenance in this build. INFERRED = pricing tiers and buyer demand. UNKNOWN = source licensing constraints and live price accuracy after the build date.