Food information for everyday shoppers in Poland and Germany.
Find → inspect recorded facts and limitations → compare compatible quantities → save and retrieve products.
TryVit is invitation-only. It is not a medical-advice service, an allergen-safety certification, a complete catalog of everything sold locally, or a personal dietary recommendation system.
The current implementation work replaces the unsupported weighted health score with source-linked facts and explicit uncertainty. Historical scores are retained for audit, not republished as current winners or health outcomes. Implementation status is not deployment status:
- Release and verification ledger
- Product, mathematical and source-data policy
- Product positioning and open assumptions
- Current project status
Nutrition values retain their unit, basis, preparation state and qualifiers. Missing is not zero; recently retrieved is not package-verified. Recorded fields must resolve to source observations. Invalid or incompatible evidence withholds comparison instead of silently producing a reassuring result.
- Next.js / React frontend with shared neutral UI, TypeScript and TanStack Query.
- Supabase/Postgres with server-owned authorization and runtime-validated RPCs.
- Python acquisition and normalization, stable source identity and immutable observations, transactional source-owned projections.
- GitHub Actions and the established Vercel deployment path; required checks, exact migration manifests, staging validation and scoped recovery evidence.
No additional framework or search service is required for this milestone.
Read repository instructions and environment setup before running commands. Use local/staging fixtures for mutation tests; never point a test at real users. Keep credentials in the approved local environment, never source or reports.
cd frontend
npm ci
npx tsc --noEmit
npm run lint
npm run testGuarded browser commands and their runtime requirements live in the frontend documentation. Production builds and browser evidence must identify the actual backend/fixture environment. A build alone is not proof of a deployed working journey.
The frozen source-reconciliation cohort is in data-quality/cohorts/evidence-first-v1.json. Retain every observation and discrepancy. It checks extraction and reconciliation, not universal food accuracy or validation of a nutrition scoring model.
The repository's software license does not replace source licenses. Open Food Facts database data is ODbL; individual contents and images have their own applicable terms. Preserve attribution and source-specific notices: OFF licensing guidance. See the evidence policy.
Preserve user-owned collections, product IDs, historical evidence and recovery options. Do not enable public signup, disclose credentials or automate a human verification challenge to make a check pass. The production Turnstile first-use/replay proof remains unresolved until genuine evidence establishes it. No static badge or historical test count should be interpreted as current CI, scientific validity, deployment or launch certification.