Not a filter. Not a search bar.
A conversation that actually understands.
Classifies query into postpartum care, feeding, baby care, or flags medical emergencies with confidence scoring.
LLM parseGenerates bilingual (EN + AR) empathetic response before any product is shown. Normalises concern first.
bilingualMulti-signal keyword + tag + intent scoring retrieves top-5 relevant products. No hallucination possible.
in-memoryModel reasons only over retrieved products. Generates structured JSON recommendations with usage guidance.
groundedAppends context-aware safety advice, medical disclaimers, and follow-up support information.
safetyPydantic schema validation, confidence bounds check, and clean structured JSON delivered to client.
structuredMothers use natural language. Mumcare understands it — not just the words, but the worry behind them.
An evaluation harness with 10+ test cases measures what we claim — intent accuracy, hallucination rate, safety contracts. No vibes, only metrics.
vibrant ★ empathetic ★ grounded
Better to under-promise (uncertainty) than over-promise (hallucination). Medical concerns are never guessed — they're flagged.