About the project

What inspired us

RAG7 is a Retrieval-Augmented Generation (RAG) platform designed to transform static information into intelligent, searchable knowledge. Inspired by the frustration of constantly digging through PDFs, documentation, and GitHub repositories, we built RAG7 to let users upload, organize, and query documents through an MCP server that can integrate well with modern AI IDEs like Cursor, claude, and we also made an AI chat interface.

The system supports PDF, DOCX, and TXT uploads, automatic summarization, web and GitHub scraping and a local chromaDB to scale up to thousands of documents.

Tech stack

Built with FastAPI, PostgreSQL, and ChromaDB, and powered by GPT-4o and sentence-transformer embeddings, RAG7 manages multiple knowledge bases, provides hierarchical summarization, and enables real-time retrieval across large document sets.

Challenges

We faced lots of challenges developing the MCP server as testing it takes really long, cursor had to restart every time to pick up the new changes. Another design choice we made was using Docker because we faced problems using web scraping tool which doesn't support natively on windows, which had led us to change to using Docker.

What we learnt

From this project, we learnt a lot about how to use git efficiently, docker desktop, and knew how to develop MCP servers from scratch (which we never had any idea how it worked).

Built With

• chromadb
• fastmcp
• openai
• postgresql
• python
• typescript