Porichoy — Turning Chaotic Factory Records Into EU Digital Product Passports

title: "Porichoy \u2014 Turning Chaotic Factory Records Into EU Digital Product Passports" published: true description: "A local-first PWA that maps messy Excel, ERP and paper records into buyer-ready EU Digital Product Passports \u2014 with women DPP Data Stewards at each pilot factory." tags: typescript, react, showdev, opensource

The problem

Bangladesh's garment factories run on messy Excel sheets, ERP exports, and paper trim cards — with Bangla/English mixed headers, duplicate rows, and missing cells. From 2027, the EU's ESPR Digital Product Passport rules start biting, and buyers will expect passport-ready data per production order.

The gap isn't data — it's data shape. Porichoy (পরিচয় — "identity") closes it.

Try it: rudra496.github.io/porichoy

What it does (pilot v1.0)

1. Ingest — drop a factory .xlsx/.csv; the first sheet parses on-device
2. Map — a deterministic engine maps chaotic headers to the canonical DPP schema: documented synonym patterns (English + Bangla), bigram-dice fuzzy fallback, confidence scores with human confirmation, and a learned vocabulary per factory. Bangla numerals supported
3. Score — a DPP Readiness Score (0–100) per production order, weighted by the ESPR Art. 8 category, with the exact missing-attribute list
4. Publish — a QR whose link embeds the passport snapshot (no backend needed): scanning opens the buyer view on any phone
5. Local-first — raw factory data never leaves the device in the pilot

Provenance is an append-only SHA-256 hash chain — deliberately a hash chain, not blockchain, because nothing here needs a consensus layer.

Engineering notes

• Vite + React + TypeScript PWA with an offline-tolerant service worker; zero backend
• src/engine/ holds the versioned schema (12 ESPR-anchored attributes), the mapping engine, readiness scoring, and the provenance chain — covered by 17 unit tests (npm test)
• The textile delegated act is still a draft, so schema mappings are config, not rewrites
• samples/ ships 3 sheets of deliberately messy fictional factory data ("Shonali Textiles Ltd.") — Bangla headers, conflicting duplicates, missing cells — loadable in-app

The human layer

Technology alone doesn't fix traceability. Porichoy's model pairs the tool with DPP Data Stewards — women trained at each pilot factory who own the mapping and verification workflow. The grant-funded roadmap goes OCR for paper trim cards → ERP connector → cloud multi-factory sync → steward certification → a 10-factory paid pilot.

Built for Needle Innovation Challenge 3.0 (Oporajita / H&M Foundation initiative), "Better Process: transparency & traceability."

Links

• Live: rudra496.github.io/porichoy
• Source: github.com/rudra496/porichoy