Inspiration Every day, one person in a home answers "what should we cook today?" Mostly a mother, on a tight monthly budget, often away from home and unsure what is left. Students in rented flats abroad face the same thing.

What it does Photograph the month's shopping (several items per photo, with brand and amount). Home shows days to the next shopping and what is running low. Cook suggests dishes, best match first, from her own recipe book of 3,003 dishes, instantly, asking Gemini only when nothing fits. Cooked-this-week dishes are hidden. Log a dish and the groceries go down. English and Urdu, right to left.

How I built it Flutter (Android), Firebase Authentication, Cloud Firestore with per-user rules, and Gemini through Firebase AI Logic (no AI key in the app). The AI never sees names, exact ages or email.

Challenges Gemini was slow and busy, so I wrote my own recipe book and made Gemini the fallback. Screens stuck loading on live Firestore feeds. Photo reading missed items and amounts. I redid every screen mid-build, and tests made that safe.

What I learned Put the cheap, certain thing first. Tell the truth in the interface. Design for the real person. 338 automated tests make bold changes safe.