We’re building Largemem, (https://largemem.com) a shared knowledge base where groups upload and maintain a common set of documents (PDFs, scans, audio) and query them conversationally.

Each group has its own persistent knowledge base. We parse content into chunks, extract entities, and combine vector search with a lightweight knowledge graph, so answers can synthesize information across multiple documents and across the group’s shared context - not just retrieve isolated snippets.

We would love to hear your feedback!

This is a cool feature, given that ChatGPT introduced the group chats fairly recently (I am not sure about usage statistics), they could introduce this any day now. Perhaps, this will be very useful for Enterprise, but Enterprise could just use the ChatGPT Enterprise and fine tune a model on their data for their company.

Do you have any evals comparing your approach to alternatives?

A demo video would've been nice