KitchenFlow - Gemini Integration Description
KitchenFlow transforms cooking from inspiration to action using Google Gemini 3's multimodal capabilities to bridge fridge inventory, food cravings, and optimized shopping—embodying "Crave it, Snap it, Shop it."
Gemini 3 Features Used:
Vision API: Analyzes fridge photos identifying ingredients, quantities, and freshness with real-time recognition. Handles cluttered shelves and partial visibility. Calculates coverage score:
$$\text{Coverage} = \frac{\text{Available Ingredients}}{\text{Required Ingredients}} \times 100\%$$
Multimodal Understanding: Simultaneously processes visual recognition, OCR text extraction (labels, expiry dates), and contextual reasoning (understanding "half-used milk expiring tomorrow"). Enables intelligent recipe matching based on current inventory.
Advanced Reasoning: Matches user cravings against available ingredients to suggest optimal recipes. Calculates shopping efficiency:
$$\text{Efficiency} = \frac{V_r}{C_s}$$
where \(V_r\) is recipe value, \(C_s\) is shopping cost. Generates: "Make carbonara with 80% current ingredients; add £3.50 for pancetta and parmesan."
Structured Output: Custom prompts return consistent JSON for recipe recommendations, ingredient gaps, and prioritized shopping lists.
Reduced Latency: \(T_{\text{scan}} < 2.5s\) enables instant suggestions while browsing fridge.
Why Central: Without Gemini 3, this requires separate computer vision, NLP, and recommendation systems. Gemini 3 unifies these into one API, enabling seamless fridge-to-recipe-to-shop workflow—demonstrating practical AI value in home cooking beyond chat interfaces.
Built With
- google-gemini-3-vision-api
- google/generative-ai-sdk
- next.js
- paw
- react
- remotion-(for-video-demo)
- supabase
- tailwind-css
- typescript
- vercel
Log in or sign up for Devpost to join the conversation.