Legit Lift

Inspiration

There is a lot of misinformation and conflicting advice about health/fitness throughout the internet, and we wanted to make an easy way to find out what advice is credible and worth following.

What it does

This project is an AI-powered tool that is designed to give users an analysis of the credibility of a fitness advice YouTube video that they input.

How we built it

We first decided that the user would input a link to a fitness advice video they would like to check the credibility of. Then, we used yt-dlp, a Python library with a Node.js wrapper, to get transcript data from the given link. Then we cleaned the VTT (video text track) files to remove timestamps and make the transcript human/agent readable. Using information from the Gemini API, we generated a credibility score from 1-100, and engineered a prompt that outlines the criteria on which the video transcript is to be scored. The API also generates a risk level (low, medium, high), and an explanation of why the video is/isn't credible. We stored all of these data using a MongoDB Atlas Cluster which was then returned to the user. All the information returned by the API is in JSON, which we parse before displaying to the user. We then used Vercel to connect the project to our domain, which is legitlift.tech.

Challenges we ran into

A challenge we ran into was deploying the backend to the whole project, in which the problem was how the yt-dlp wasn't able to work with Vercel.

Accomplishments that we're proud of

We're proud of all finishing our first hackathon and having an amazing experience with completing our first project!

What we learned

We learned how to scrape YouTube transcript data through a python library with a node.js wrapper. We also learned how to use the Gemini API alongside learning storing data using MongoDB with a full-stack perspective.

What's next for Legit Lift

The next step for Legit Lift is to refine and work for larger user deployment and add more features other than YouTube.

Built With

• css
• express.js
• geminiapi
• json
• mongodb
• node.js
• react
• render
• tailwind
• typescript
• vercel