Chomp - Recipes for Everyone

Demo of the form pre-recipe generation
We added an input for price

Inspiration

Recipe ideation is dominated by out-of-touch food influencers who prioritize flashiness and view counts over practical quality of recipes and accessibility for the average viewer. Instead of pretentious slop, people need tailor-made recipes that fit their needs.

What it does & How it Works:

Currently, there are numerous technologies aimed at cataloging recipes or recommending popular dishes based on user-inputted constraints (like budget, cuisine, etc). There are also emerging tools that use AI to generate recipe ideas in isolation. Chomp is meant to bridge these two approaches by grounding AI-generated recipes in realistic, user-provided context. The system is structured as follows:

Our backend is built in Python using FastAPI and Uvicorn. If necessary, we could make it serverless, but since this project is hosted locally, that is unnecessary. Users are prompted to fill in their personal information (this is not necessarily required, but ensures that recipes are optimized to their current situation), which is then passed into Gemini for recipe generation. Our prompt is designed to maximize factors such as affordability and pure taste, which means that recipes will occasionally be creative variants or combinations of well-known classics. This was a side effect, but from preliminary testing (asking other participants to demo our product), we found that this is not necessarily a negative corollary.

We employ Selenium to scrape various websites containing both local and chain supermarkets (e.g. Wegmans, Walmart, Trader Joe's, etc.), and we pass these results to the frontend, which is built in a compact React+Vite system.

Challenges we ran into

We encountered a small number of miscellaneous issues that mostly derived from the difficulty in executing the higher strategic vision of the project. We were also relatively rusty and spent more time than anticipated in the setup phase, although we were able to rebound into a well-structured flow state in time to finish the project.

Accomplishments that we're proud of

Despite the aforementioned challenges, we built a fully functional end-to-end system within the hackathon timeframe: - We integrated live grocery price scraping across multiple supermarkets, allowing Chomp to ground its recipes in current market reality rather than the whims of TikTok influencers. - We accelerated development velocity at an unexpected rate, which shows how we have grown as developers over the course of the hackathon. - We resisted the urge to add “blockchain” or “Web3” to our Devpost description for engagement farming, which took restraint.

What's next for Chomp

We recognize that within the overall genre of social media food content, there exists a subniche focused on creating recipes adjacent with Chomp's mission. We hope to develop a formula that allows us to leverage the quirks typically found with these recipes and integrate that into our Gemini prompt.