Linky

QR Code to website
Home page
Chat page

✨ Inspiration

When we are working on new projects, we often struggle a great deal with having to go through large pages of documentation about these diverse technologies when all we are looking for is things related to X or Y. We decided we wanted to make something to help make things of that nature easier, a way to interact with the article.

🚀 What it does

Linky leverages a RAG AI model with a Pinecone Vector database. The user inputs a URL which is then sent into the vector database. From there, the user can select a stored URL and ask Linky a question. Linky retrieves data from the sent in query and then generates an accurate response, all while sourcing the data it formed it's answer around.

🛠️ How we built it

TypeScript
Mantine
Vercel
Pinecone Vector Database
Open AI Ada 2

🤯 Challenges we ran into

Text wrapping issues with retrieved data
Continuous development (the model is not configured on our localhosts, we needed to deploy first when it came time to changes related to the generations).