What it does
AI Notes can be used to take notes. Both text and images can be used.
AI generated caption will be automatically added to images.
In home page users can see wither the list of their notes or a summarized view of each notes.
User can ask any question on the notes and will get an answer and reference to the notes.
How we built it
The application is built with Next.js. Uses Github OAuth for authentication.
For database Azure Cosmos DB for MongoDB is used.
For storing images Azure Storage is used. OpenAI model "gpt-4o" is used to generate a caption for the image.
The notes content along with image caption is saved as vector in MongoDB. OpenAI's "text-embedding-3-small" embedding model is used for vector conversion.
OpenAI model "gpt-4o" is used to generate a summary for the notes.
The same model with RAG is used for search. Function calling capability of OpenAI is used to call a function which does a vector search in the cosmosdb and returns the result back for the gpt model to process and respond back to user.

What we learned
Learned deploying a complete application in azure with OpenAI and cosmosdb
What's next for AI Notes
Add audio notes option and add Azure Speech to Text to transcribe it. Update UI to support PC screens.
Built With
- azure
- cosmosdb
- openai
Log in or sign up for Devpost to join the conversation.