RMA-MUN/RAGNotebook

What it solves

RAG Notebook is an AI-powered personal knowledge management tool designed to prevent notes from becoming "isolated islands" of information. It solves the problem of users writing notes but never revisiting them by integrating note management with a knowledge graph and AI-driven retrieval.

How it works

The system combines a Markdown editor with a GraphRAG architecture. It uses LLMs to automatically extract entities and relationships from uploaded documents (PDF, DOCX, PPTX, TXT, MD) and user notes, storing them in a Neo4j graph database. When a user asks a question, the system employs an Agentic RAG approach—using the knowledge graph to guide retrieval and fusing evidence from both the graph and vector search (RRF fusion) to provide cited answers. It also includes an Ebbinghaus forgetting curve algorithm for spaced repetition review.

Who it’s for

• Individuals needing a sophisticated AI-driven system to manage their personal knowledge base.
• Developers or job seekers looking for a practical implementation of Agentic RAG and Knowledge Graph integration.

Highlights

• Graph-Guided RAG: Uses Neo4j to store knowledge as a graph, allowing the AI to traverse relationships for more accurate retrieval.
• AI Writing Suite: Features real-time inline completion (Tab-to-accept), expansion, summarization, and cross-source related note recommendations.
• Automated Organization: LLMs automatically generate tags and categories for notes upon saving, removing the need for manual sorting.
• Spaced Repetition: Built-in support for the Ebbinghaus forgetting curve to help users retain information.
• Hybrid Search: Combines Neo4j vector search and full-text search using Reciprocal Rank Fusion (RRF).