cook-e | Devpost

Home Page
Page in action
Jaeger UI Look with Trace Data

Inspiration

Since we're students living at home, we don't spend too much time learning new cooking things. So when we were thinking of projects to build, something related to food enticed both of us. Using AI to help generate recipes was something that we both found to be really useful and can definitely be used in our day-to-day lives.

What it does

Cook-e takes in a food selection and produces a list of ingredients, nutrition facts, and a recipe. You can take this information and start cooking, or if you see that there's something that you don't like about the recipe, you can ask to cater the recipe more to your needs. Cook-e also utilizes software observability and provides the developer with key information throughout the software cycle. Observability and distributed tracing are important concepts in modern software development because they help developers understand and debug complex distributed systems.

How we built it

For the recipe generator, we used Cohere's command model so that it could give us recipes for any food that we asked for. The nutritional facts were found using an API called CalorieNinjas. Our front-end and back-end uses NodeJS to tie everything together. We introduced the leading opensource technology(Opentelemetry) within our project as we took advantage of auto-instrumenting our code and exporting trace result with our local instance of Jaeger UI, an opensource tool which allows develops to search, view trace data and analyze performance metrics, bottlenecks within a system and the overall flow of the system.

Challenges we ran into

Figuring out how to uses Cohere's model for contextual conversations.
Tying our front-end and back-end data together.

Accomplishments that we're proud of

We're proud of being able to use an LLM for the first time as this was our first venture into something like this. Completing a project which tied together a bunch of different technologies in such a short time is definitely no small feat, but to accomplish a working MVP which is on par with our expectations is definitely something for us to be proud of.

What we learned

How to process data from a Generation model, and make it viewable for the user. How to handle API requests when using multiple other API calls. How to deploy our project on a cloud-based platform. Incorporate distributed tracing within our project.