Kappa Graph

About

Kappa Graph (κ(G)) is a semantic knowledge graph system designed to not only extract concepts and relationships from various document types but also to critically evaluate and track their epistemic status. Unlike traditional systems, it assigns a 'grounding score' to each piece of knowledge, indicating its evidential support, and preserves disagreements between sources, rather than reconciling them. This allows users to understand the confidence level of information, trace every concept back to its origin, and leverage semantic diversity as a signal for authenticity, enabling robust knowledge management for complex information environments.

Key Features

• Trace every concept and claim directly back to its original source evidence.
• 21 GitHub stars
• Perform semantic search to find content based on meaning and explore connections between ideas.
• Assess knowledge confidence through continuous grounding scores and identify where sources disagree.
• Ingest diverse documents (PDFs, markdown, images, text) for automatic concept and relationship extraction.
• Query the knowledge graph via AI assistants (MCP server integration) or navigate it as a FUSE filesystem.

Use Cases

• Provide AI agents with persistent, grounded memory, enabling them to validate claims and explore relationships.
• Extract and query architecture concepts from technical documentation like diagrams, notes, and design docs.
• Synthesize research by ingesting papers to find connections, identify disagreements, and assess claims' support.