Recipe Doctor

📖 About the Project

## Inspiration

The "fridge full of nothing" dilemma. I built Recipe Doctor to turn random leftovers into cohesive meals, helping home cooks reduce food waste through AI-driven creativity.

## What it does

It transforms a list of ingredients into fully detailed recipes. Users filter by diet, cuisine, and skill level to get instant, step-by-step cooking guidance and a dynamic shopping list.

## How we built it

• UI: Streamlit for a responsive, Python-native frontend.
• AI: Google Gemini API for recipe generation.
• Logic: Custom Python parsers to handle dietary constraints and ingredient scaling: $$Q_{new} = Q_{original} \times \left( \frac{S_{target}}{S_{original}} \right)$$

## Challenges we ran into

• Hallucinations: Preventing the AI from "inventing" ingredients.
• API Security: Managing leaked keys and migrating to secure st.secrets.
• Prompt Tuning: Finding the perfect temperature ($T \approx 0.7$) for creative yet edible results.

## Accomplishments that we're proud of

• Implementing a structured parsing system that turns raw AI text into clean UI components.
• Building a "zero-waste" logic that prioritizes what you already have.

## What we learned

I mastered Prompt Engineering and the importance of schema validation. I also learned that environment management is critical when working with evolving AI SDKs.

## What's next for Recipe Doctor

• [ ] Nutritional Macros: Automatic calorie and protein tracking.
• [ ] Vision: Snapping a photo of the fridge to "scan" ingredients.
• [ ] Scaling: One-click adjustment for different serving sizes.

Built With

• genai
• llm
• natural-language-processing
• python
• streamlit