What problem does it solve? Understanding a large codebase or mixed corpus of documents, papers, and media requires reading everything manually; graphify turns any folder into a persistent, queryable knowledge graph so questions about architecture, file relationships, and concepts are answered by graph traversal instead of re-reading files. ## Core Features & Use Cases - Knowledge graph construction: Extracts entities and relationships from code (AST-based, no LLM needed) and from docs, papers, images, and transcribed video (semantic extraction via subagents or Gemini), with an honest EXTRACTED/INFERRED/AMBIGUOUS audit trail. - Analysis and outputs: Runs community detection, god-node and surprising-connection analysis, and produces an interactive HTML visualization, GraphRAG-ready JSON, and a plain-language GRAPH_REPORT.md. - Query and maintenance: Answers natural-language questions via BFS/DFS traversal, finds shortest paths between concepts, explains nodes, supports incremental --update re-extraction, watch mode, git commit hooks, and exports to Neo4j, FalkorDB, GraphML, SVG, Obsidian, wiki, and MCP server. - Use Case: Point it at a research repository or a GitHub URL, then ask "How does the training pipeline connect to data loading?" and get a graph-grounded answer with source citations. ## Quick Start Run /graphify on the current directory to build the knowledge graph, then ask any question about the codebase to query it.
