What problem does it solve? Understanding a new codebase, research corpus, or mixed folder of notes requires re-reading everything each session, and hidden cross-document connections stay invisible. graphify builds a persistent, queryable knowledge graph from any folder so relationships survive across sessions and every edge carries an honest EXTRACTED, INFERRED, or AMBIGUOUS audit tag. ## Core Features & Use Cases - Multi-format extraction: Combines deterministic AST parsing for code with LLM-based semantic extraction for docs, papers, images, and Whisper-transcribed video/audio. - Community detection and analysis: Clusters the graph, identifies god nodes and surprising cross-document connections, and generates a plain-language GRAPH_REPORT.md. - Multiple outputs and query modes: Produces interactive HTML, GraphRAG-ready JSON, Obsidian vaults, SVG/GraphML exports, Neo4j push, an MCP server, and BFS/DFS natural-language queries with token budgets. - Use Case: Point it at an unfamiliar repository to get an architecture graph, labeled communities, and suggested questions before touching any code, then run incremental --update runs as the code changes. ## Quick Start Ask the agent to run /graphify on the current project directory to generate an interactive HTML knowledge graph and audit report in graphify-out/.