Menu Anxiety Solver

Menu Anxiety Solver

Inspiration

The awkward stare of a waiter when you've been staring at the menu for 10 minutes. The regret of ordering something disappointing because you felt rushed. The frustration of asking 'does this have dairy?' for the fifth time.

What it does

Menu Anxiety Solver eliminates decision paralysis at restaurants. Snap a photo of any menu, tell the AI your dietary needs and preferences, and get instant personalized dish recommendations with explanations. It reads menus in any language, understands context beyond keywords (catching hidden allergens and assessing value), and recommends 2-3 perfect dishes in seconds, like having a knowledgeable food expert in your pocket who knows exactly what you'll love.

How we built it

Tech Stack:

Frontend: React + TypeScript (Vite)
AI Engine: gemini-3-flash-preview for vision + language processing
Architecture: Client-side application with intelligent rate limiting and error handling

Challenges we ran into

Rate Limits: Hit Google's API rate limits hard during testing solved with intelligent request throttling and user-friendly wait messages.
AI Response Consistency: Gemini sometimes returned markdown-wrapped JSON or added extra text built robust parsing logic to clean and validate responses.
Real World Menu Chaos: Menus are messy, handwritten text, multiple languages, missing prices, vague descriptions. Extensive prompt engineering helped the AI handle these variations gracefully.

Accomplishments that we're proud of

Built a functional AI product in under 24 hours
Solved a universal problem
Multilingual from day one
Overcame major technical hurdles API rate limits, inconsistent AI responses, model availability
Actually useful - this isn't just a demo, it's something we genuinely want to use ourselves. We all will use it whenever we will visit any restaurant.