Graphiti
Graphiti is an open-source platform designed to improve Retrieval Augmented Generation (RAG) systems by incorporating knowledge graph structures. While traditional RAG systems retrieve relevant documents or text passages to provide context for language models, Graphiti instead organizes information as interconnected entities and their relationships, creating a structured semantic representation of knowledge rather than relying on flat document collections.
How It Works
The platform constructs knowledge graphs that map entities and their connections, allowing the system to understand context and relationships between concepts more explicitly. This structured approach enables more precise retrieval of relevant information by traversing relationships in the graph rather than performing keyword-based document matching. When a query is processed, Graphiti can return semantically related entities and their connections alongside retrieved content, providing richer contextual information to language models for answer generation.
Key Advantages
By representing knowledge as graphs, Graphiti addresses several limitations of traditional RAG systems. The explicit relationship structure reduces reliance on document-level chunking strategies and can better handle complex queries that require understanding connections between multiple concepts. This approach also facilitates updates and maintenance of knowledge bases, since new information can be integrated into the graph structure more systematically than adding new documents to unstructured collections.