graphify

Run /graphify on the project directory. It parses code files with AST extraction for imports and structure, then uses parallel subagents for semantic relationships, and outputs an interactive HTML graph, JSON, and an Obsidian vault.

The pipeline detects code files, Markdown and text documents, PDF papers, and images. Images are interpreted with vision, so screenshots, charts, diagrams, and whiteboard photos become nodes with typed relationships.

Use the --update flag for incremental re-extraction. A saved manifest tracks file state, so only new or modified files are re-processed and merged into the existing graph, saving tokens and time.

Yes. The --neo4j flag generates a Cypher file, --neo4j-push writes directly to a running Neo4j instance via bolt, and --graphml exports a GraphML file compatible with Gephi and yEd.

An empty graph means extraction produced no nodes, usually because all files were skipped, the corpus contains only unsupported binary files, or extraction failed. Check the detection summary for skipped or unsupported files.

Every edge carries a confidence tag: EXTRACTED for relationships explicit in the source, INFERRED for reasonable deductions, and AMBIGUOUS for uncertain cases flagged for review. The audit report separates these categories.