Inspiration

Imagine a world where learning is as easy as having a conversation with a friend. Picture a tool that unlocks the treasure trove of educational content on YouTube, making it accessible to everyone, regardless of their background or expertise. This is exactly what our hackathon project brings to life.

  • Current massive online courses are great resources to bridge the gap in educational inequality.
  • Frustration and loss of motivation with the lengthy and tedious search for that 60-second content.
  • Provide support to our students to unlock their potential.

What it does

Think of our platform as your very own favorite personal tutor. Whenever a question arises during your video journey, don't hesitate to hit pause and ask away. Our chatbot is here to assist you, offering answers in plain, easy-to-understand language. Moreover, it can point you to external resources and suggest specific parts of the video for a quick review, along with relevant sections of the accompanying text. So, explore your curiosity with confidence – we've got your back!

  • Analyze the entire video content 🤖 Learn with organized structure and high accuracy
  • Generate concise, easy-to-follow conversations⏱️Say goodbye to wasted hours watching long videos
  • Generate interactive quizzes and personalized questions 📚 Engaging and thought-provoking
  • Summarize key takeaways, explanations, and discussions tailored to you 💡 Provides tailored support
  • Accessible to anyone with an internet 🌐 Accessible and Convenient

How we built it

Vite React,js as front-end and Flask as back-end. Using Cohere command-nightly AI and Similarity ranking.

Challenges we ran into

  • Increased application efficiency by 98%: Reduced the number of API calls lowering load time from 8.5 minutes to under 10 seconds. The challenge we ran into was not taking into account the time taken for every API call. Originally, our backend made over 500 calls to Cohere's API to embed text every time a transcript section was initiated and repeated when a new prompt was made -- API call took about one second and added 8.5 minutes in total. By reducing the number of API calls and using efficient practices we reduced time to under 10 seconds.
  • Handling over 5000-word single prompts: Scraping longer YouTube transcripts efficiently was complex. We solved it by integrating YouTube APIs and third-party dependencies, enhancing speed and reliability. Also uploading multi-prompt conversation with large initial prompts to MongoDB were challenging. We optimized data transfer, maintaining a smooth user experience.

Accomplishments that we're proud of

Created a practical full-stack application that I will use on my own time.

What we learned

  • Front end: State management with React, third-party dependencies, UI design.
  • Integration: Scalable and efficient API calls.
  • Back end: MongoDB, Langchain, Flask server, error handling, optimizing time complexity and using Cohere AI.

What's next for ChicSplain

We envision ChicSplain to be more than just an AI-powered YouTube chatbot, we envision it to be a mentor, teacher, and guardian that will be no different in functionality and interaction from real-life educators and guidance but for anyone, anytime and anywhere.

Share this project:

Updates