Project Info

Winner

Google: Most useful app

Refrigerator Ramsay

Inspiration

The idea for Refrigerator Ramsay came from a common problem we all face: food waste. So many times, we forget about ingredients in the fridge, or we don't know how to use them, and they go bad. We wanted to create something that could help us not only use those ingredients but also inspire us to try new recipes and reduce food waste in the process.

What it does

Refrigerator Ramsay is an AI-powered kitchen assistant that helps users make the most of their groceries. You can take a picture of your fridge or pantry, and the AI will identify the ingredients. Then, it suggests recipes based on what you have. It also has an interactive voice feature, so you can ask for cooking tips or get help with recipes in real-time.

How we built it

We used Google Gemini AI for image recognition to identify ingredients from photos. and then, to come up with creative meal ideas. We also added Hume EVI, which is an emotionally intelligent voice AI, to make the experience more interactive. This lets users chat with the AI to get tips, learn new cooking techniques, or personalize recipes.

Challenges we ran into

One challenge we faced was working with Hume EVI for the first time. It was really fun to explore, but there was definitely a learning curve. We had to figure out how to make the voice assistant not just smart but also engaging, making sure it could deliver cooking tips naturally and understand what users were asking. It took some trial and error to get the right balance between information and interaction. Another challenge was training the Google Gemini AI to be as accurate as possible when recognizing ingredients. It wasn't easy to ensure the AI could reliably detect different foods, especially when items were stacked or grouped together in photos. We had to spend extra time fine-tuning the model so that it could handle a variety of lighting conditions and product packaging.

Accomplishments we're proud of

We're proud of creating a solution that not only helps reduce food waste but also makes cooking more fun. The AI's ability to suggest creative recipes from random ingredients is really exciting. We're also proud of how user-friendly (and funny) the voice assistant turned out to be, making it feel like you're cooking with a friend.

What we learned

We learned a lot about how AI can be used to solve everyday problems, like reducing food waste. Working with Hume EVI taught us about building conversational AI, which was new for us. It was really fun, but there was definitely a challenge in figuring out how to make the voice interaction feel natural and helpful at the same time. We also learned about the importance of training AI models, especially with the Gemini AI. Getting the image recognition to accurately identify different ingredients required us to experiment a lot with the data and train the model to work in a variety of environments. This taught us that accuracy is key when it comes to creating a seamless user experience.

What's next

Next, we want to improve the AI’s ability to recognize even more ingredients and maybe even offer nutritional advice. We’re also thinking about adding features where the AI can help you plan meals for the whole week based on what you have. Eventually, we’d love to partner with grocery stores to suggest recipes based on store deals, helping users save money too!

Analysis

Compare with all teams

View

Metric

Sherwin Vishesh Jathanna
15

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only

CSSIn code
FastAPIIn code
HTMLIn code
JavaScriptIn code
Next.jsIn code
PythonIn code
ReactIn code
Tailwind CSSIn code
TypeScriptIn code

9 of 9 appear in the indexed code.

AI coding agents

No AI coding agent signals were found in this repository.

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

194 KB

Source files

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.

Feature verification

4 Verified1 Code-supported2 Claimed only

AI recipe generation from identified ingredientsVerified
It suggests recipes based on what you have
Claimed on Devposthigh confidence
- backend/main.py:72 — Second Gemini call (gemini-1.5-pro) generates a recipe from the identified groceries and user description, returned to the frontend
- frontend/src/components/ChefChat.js:13 — ChefChat page renders the returned recipe and groceries as Markdown
FastAPI backendVerified
Built with fastapi
Claimed on Devposthigh confidence
- backend/main.py:2 — FastAPI app defined with CORS middleware and a POST endpoint
Ingredient recognition from fridge/pantry photo via Google GeminiVerified
Take a picture of your fridge or pantry, and the AI will identify the ingredients
Claimed on Devposthigh confidence
- backend/main.py:39 — /analyze_image endpoint accepts an uploaded image and calls genai.GenerativeModel('gemini-1.5-flash').generate_content to identify groceries
- frontend/src/components/EntryDescription.js:51 — Frontend uploads the image file via FormData to the backend /analyze_image endpoint
React frontendVerified
Built with react
Claimed on Devposthigh confidence
- frontend/src/App.js:1 — React app with react-router-dom routes for HomePage, EntryDescription, ChefChat, CallPage
Interactive voice feature via Hume EVI for cooking tips/help in real-timeCode-supported
It also has an interactive voice feature, so you can ask for cooking tips or get help with recipes in real-time
Claimed on Devpostmedium confidence
- evi-files/utils/getHumeAccessToken.ts:5 — Fetches a Hume access token via the hume SDK
- evi-files/components/Chat.tsx:26 — Wraps chat UI in Hume's VoiceProvider from @humeai/voice-react, wiring up the voice session
- frontend/src/components/CallPage.js:15 — Main React app only embeds the EVI voice UI via an iframe pointing to an external deployed URL (refrigerator-ramsay-evi.vercel.app) rather than including the evi-files code directly, so end-to-end integration within this repo can't be fully confirmed
Nutritional advice / weekly meal planning / grocery store deal integration (future work)Claimed only
Next, we want to improve the AI's ability to recognize even more ingredients and maybe even offer nutritional advice... plan meals for the whole week... partner with grocery stores
Claimed on Devposthigh confidence
Personalized, data-driven meal suggestions learning from user history/preferencesClaimed only
The AI learns from your ingredient usage patterns, preferences, and cooking history, offering personalized meal suggestions
Claimed on readmehigh confidence

An AI agent derived these features from the project’s Devpost page and readme, then searched the code for each one. Verified features are backed by cited code; claimed-only features had no supporting code, which is not by itself proof a feature is missing.