graphify

Run /graphify on a local path or GitHub URL to extract entities and relationships into a knowledge graph. Code is parsed structurally via AST with no LLM required, and the output includes graph.json, an interactive HTML visualization, and a GRAPH_REPORT.md summary.

Use graphify query followed by your question once graphify-out/graph.json exists. It expands your query against the graph's actual vocabulary, then runs BFS traversal for broad context or DFS with --dfs to trace a specific dependency path.

No API key is required. Code extraction is fully structural via AST, and semantic extraction for docs and images falls back to the host agent itself. Setting GEMINI_API_KEY or GOOGLE_API_KEY optionally enables Gemini for semantic extraction.

Yes, the --neo4j and --falkordb flags generate a Cypher file, while --neo4j-push and --falkordb-push load the graph directly into a running instance. Both use MERGE statements, so re-running is safe without creating duplicates.

Run /graphify with --update to re-extract only new or changed files and prune deleted ones. Code-only changes skip LLM extraction entirely, making incremental updates fast and free of token costs.

Supported inputs include source code in many languages, Markdown and text docs, PDF papers, images analyzed with vision, and video or audio transcribed via Whisper. Multiple GitHub repos can also be cloned and merged into one cross-repo graph.