Hyperlinks lead to LinkedIn profiles*
- Marcus Lee DEC, Computer Science & Math @ Vanier College 💻
- Carson Spriggs-Audet DEC, Computer Science @ John Abbott College 💻
Note
Linky was developed in <24hrs for MariHacks' 2024 Hackathon!
Researching can be complicated. Reading through possible sources just to see there isn't actually important information, having to read through heaps of information on a research paper to extract one bit of information, etc. Linky is here to change that for you.
Give Linky a URL (an article, Wikipedia page, etc.) and watch it parse all the data into different chunks! Then, ask Linky a question. Linky will leverage the information from that article and answer your question accurately based on what it read! 🤖💡
When diving into new projects, navigating extensive documentation across various technologies can be overwhelming, especially when seeking specific information. We recognized this challenge and aimed to simplify the process by creating a tool that streamlines access to relevant content.
Linky harnesses the power of an AI model, integrated with a Pinecone Vector database. Users input a URL, which is then processed through the vector database. With the selected URL, users can pose questions to Linky. Leveraging the queried data, Linky generates accurate responses, all while crediting the sourced information.
- Implemented using TypeScript
- Utilized Mantine for UI and hooks
- Deployed on Vercel
- Leveraged Pinecone Vector Database for data retrieval
- RAG AI model using Vercel's AI SDK for retrieval and generation
- Addressing text wrapping issues with retrieved data
- Dealing with continuous development challenges, as the model required deployment for updates related to data generation
- Achieving mobile functionality 📱
- Delivering a polished end result ✨
- Deepening our understanding of the technologies involved 🧠
- Successfully staying awake for over 24 hours to meet project goals 😴
- Gained comprehensive knowledge about each technology within our tech stack
- Enhancing mobile responsiveness
- Improving the display of retrieved data 📊