Metadata Engine turns a raw audio file into professional, distribution-ready metadata: DSP measurements (BPM, key, loudness, spectral profile), a multi-model AI classification (genre, mood, instrumentation, vocal style, a written description), identification against public databases, and exports in the formats DSPs and PROs accept (DDEX ERN 4.3, CWR 2.1, MP3Tag CSV, JSON). It also issues a signed Certificate of Authenticity pinned to IPFS.
Operated by HardbanRecords Lab (Poland, EU). App:
https://app-metadata.hardbanrecordslab.online.
| Document | Audience | Covers |
|---|---|---|
| BUSINESS.md | Founders, partners, investors | The metadata problem, target users, competitive position, pricing, revenue model, roadmap |
| ARCHITECTURE.md | Engineers | The real system — services, the LLM consensus ensemble, exports, certificates, deploy topology |
| USER-GUIDE.md | End users | Signing up, credits, analysing a track, reading the result cards, exports, certificates, batch, settings |
| API.md | Integrators | Submitting audio programmatically with an X-API-Key, polling for results, exports |
| LEGAL-OVERVIEW.md | Everyone | Plain-language map of the binding legal documents shown in the app |
| OPERATIONS.md | Whoever runs the deploy | Env vars, GitHub Secrets, the deploy pipeline, the VPS, admin, backups |
| SEO.md | Whoever maintains the site | Canonical domain, meta/JSON-LD strategy, sitemap upkeep |
archive/ holds one-off scripts kept only for history.
- Legal: the text rendered by
frontend/components/LegalModal.tsxis binding.LEGAL-OVERVIEW.mdonly summarises it. - Env / secrets:
backend/.env.examplelists every variable; GitHub repo Secrets is what the deploy actually reads. - Code: where a doc and the code disagree, the code wins — open an issue to fix the doc.