What problem does it solve?
It helps you translate and safely evolve the KPKN nutrition AI workflow so Android free-form food logging produces trustworthy, review-aware results with correct save semantics and deterministic fallback.
Core Features & Use Cases
- Android-first translation and validation: Port or validate behavior from the PWA nutrition pipeline into Kotlin/Compose, especially around free-form logging and drawer UX.
- Robust AI + deterministic fallback: Ensure parsing never fails hard when local AI is missing or slow, while preserving the analyze-when-requested flow.
- Correct resolution and persistence: Guarantee proper transformation from parsed items to
LoggedFood, then write-through state updates into NutritionRepository (including Room-backed persistence) so the screen and stored logs stay consistent.
- Locale-aware food parsing: Preserve Spanish/Chilean parsing intent, aliases/tokens, and protected phrases so behavior does not regress during refactors.
- Guardrails against false certainty: Keep
analysisEngine, modelVersion, containsEstimatedItems, and requiresReview semantics coherent and reflected honestly in the UI.
Quick Start
Use the kpkn-nutrition-ai skill to validate a proposed change to Local AI nutrition parsing in android-native and confirm the unresolved/estimated/review-required behavior and NutritionRepository save semantics still match the PWA oracle for the impacted flow.