Overview
KRAGEN (Knowledge Retrieval Augmented Generation ENgine) is a tool that combines knowledge graphs, Retrieval Augmented Generation (RAG), and advanced prompting techniques to solve complex problems with natural language, specifically designed for biomedical applications.
Key Features
- Knowledge Graph Integration: Converts knowledge graphs into vector databases using embeddings for searchable retrieval
- Graph-of-Thoughts (GoT): Dynamically breaks down complex problems into smaller subproblems, solves each using relevant knowledge through RAG, then consolidates solutions
- Interactive Visualization: Provides graph visualization for users to interact with and evaluate solution quality and logic
- Hallucination Reduction: Limits LLM hallucinations by grounding responses in retrieved knowledge graph facts
Technical Implementation
KRAGEN uses Weaviate, an open-source vector database, to store embedded knowledge graph information and support various embedding models and query methods. The system is deployed using custom Docker containers.
Workflow
- Embeds knowledge graph information into vector embeddings
- Creates searchable vector database from knowledge graph
- Uses GoT to decompose complex queries into subproblems
- Retrieves relevant facts for each subproblem via RAG
- Consolidates subproblem solutions into final answer
- Provides visualization of reasoning structure
Use Cases
- Biomedical problem-solving
- Scientific question answering
- Complex reasoning tasks requiring evidence-based answers
- Multi-hop reasoning over knowledge graphs
Publication
Published in Bioinformatics journal (June 2024, Oxford Academic).
Pricing
Free and open-source.
