Pantry Chef | Devpost

Logo

Link

Inspiration

We were inspired with all the food products in our pantries that had just been sitting, collecting dust for who knows how long. We wanted our website to be inclusive for those with different dietary needs such as vegans, vegetarians, and those who have Lactose intolerance.

What it does

Our website takes as an input all of the items in your pantry, taking into account any dietary restrictions/preferences, and gives you possible recipes to be made from what you have.

How we built it

_We originally used py4web + vue to build the webpage, and make API calls. We used datasets of ingredients and recipes from epicurious.com to train and fine-tune a Natural Language Processing model with OpenAI and Co:here to generate results. We used APIs to send prompts and receive recipes from the OpenAI and Co:here models. _

Challenges we ran into

Training the AI to give us quality recipes and not dishes of Random Ingredients that don't taste very well together.

Accomplishments that we're proud of

We were able to train a NLP model to produce a recipes. For some of us, this is the first time we were exposed to the intricacies of Natural Language Processing.

What we learned

We gained a better understanding of working with and developing AI to suit a particular need. We also learned about some new frameworks that allowed us to put an app together quickly. We have also learned how to work together with new people we've never met before.

What's next for Pantry Chef

We hope to implement an Expiration tracker that can also be integrated with the AI to recommend recipes that prioritize items expiring soon. We also hope that with better refinement, we can produce better, tastier results.