Neo4J

The basic idea I’m working on:

This feature adds local, privacy-preserving GraphRAG capabilities to the IS547 committee documents project using the official neo4j-graphrag-python library with Ollama for both embeddings and language model operations. The implementation will enable:

• Semantic document search via vector embeddings stored in Neo4j
• Hybrid retrieval combining vector similarity + graph traversal
• Question-answering about committees, people, and topics using local LLMs
• Full privacy – no external API calls, all processing runs locally via Ollama

The feature leverages the existing Neo4j graph (~1000 documents, 142 people, 70 committees) and text extraction pipeline, adding a semantic layer on top of the structural knowledge graph.