I have updated this application if anyone would like to check it out!
#Memetic-Agent-And-Graph-RAG Tool
narrow elbow
narrow elbow
🧠 Memetic Agent - Local-First GraphRAG & Knowledge Extraction Pipeline
Hey everyone! 👋
I've been working on a new open-source tool called Memetic Agent, and I successfully got it running with a local Neo4j instance + local LLMs (Ollama).
It's a Hybrid Graph Extraction Pipeline designed to turn unstructured text documents into a structured Knowledge Graph, which you can then chat with using a RAG agent.
Key Features:
🕷️ Hybrid Extraction: Uses GLiNER for zero-shot Named Entity Recognition (fast & accurate) + LLMs for semantic Relationship Extraction.
🤖 Agentic Workflow: Includes an "Auto-Pilot" mode that ingests files, builds a vector index, discovers schema, extracts the graph, and resolves entities automatically.
🔍 GraphRAG: Chat with your data! The agent generates Cypher queries on the fly to answer questions based on the graph structure.
🔒 100% Local: Works with Ollama (Llama 3, Mistral, Qwen) and local embeddings, keeping your data private.
📊 Dynamic Schema: Can infer node labels and relationship types from your data, or enforce a strict schema.
🖥️ GUI Console: Built with CustomTkinter for easy configuration and monitoring.
Tech Stack:
Python
Neo4j (Knowledge Graph)
LangChain (Orchestration)
Ollama (Local Inference)
I'm looking for feedback on the extraction logic and Cypher query generation patterns. Let me know what you think!
https://github.com/cherubski/Memetic-Agent-And-Graph-RAG/blob/master/README.md
buoyant forgeBOT
cherubski has been warned