Build and query a codebase knowledge graph from code, docs, and media.
npx skills add https://github.com/lurodrisilva/personal-skills --skill graphify-lurodrisilva
Please help me install this Agent Skill. Skill: graphify Source: https://github.com/lurodrisilva/personal-skills/tree/main/ai/graphify Command: npx skills add https://github.com/lurodrisilva/personal-skills --skill graphify-lurodrisilva
/graphify . to map a repo into a NetworkX graph.json via a three-pass pipeline (local tree-sitter code parsing, faster-whisper media transcription, LLM semantic extraction), then query with /graphify query, path, and explain. - PR impact and CI integration: Triage pull requests by graph blast-radius (graphify prs --triage), install AST-only git hooks, and run headless extraction in CI with graphify extract --backend claude|gemini|ollama. - MCP serving: Expose the built graph as MCP tools (query_graph, shortest_path, get_pr_impact, and more) via python -m graphify.serve. - Use Case: Onboarding to a 200k-line monorepo — build the graph once, commit graphify-out/ to git, and ask "what connects auth to database?" using roughly 71x fewer tokens than re-reading files. ## Quick Start Ask the assistant to graphify this repository and then query what connects two modules in the code graph.High-intent search queries and answers about installing and using this skill.
Install the graphifyy package with uv tool install graphifyy, register your assistant with graphify install, then run /graphify . in the repo root. The graph, HTML visualization, and report are written to graphify-out/.
The PyPI package is named graphifyy with a double y, while the console command it installs is graphify. Running pip install graphify gets the wrong or nonexistent package, so install graphifyy instead.
No. Code structure is parsed locally with tree-sitter at zero API cost. Only documentation, PDFs, images, and media transcripts go to a configured LLM during the semantic extraction pass.
Yes. Run graphify extract ./docs --backend claude, gemini, or ollama with the matching API key environment variable such as ANTHROPIC_API_KEY. Code extraction stays local regardless of backend.
No. Graphify's only store is the NetworkX node-link graph.json under graphify-out/. The Docker MCP SQLite server is a separate companion for persisting query insights, not Graphify storage.
Avoid it for generic graph database needs like Neo4j administration or for RAG vector-embedding pipelines. Graphify uses graph structure as the similarity signal rather than a vector index.
wanyusheng63-altScan directories recursively and generate a grouped Markdown file listing file paths.
AO-HySExtract business domains and flows from a codebase into an interactive domain graph.
burningportraAudits flywheel dependencies, codebase state, and bead graph into a scored health report.
shalevaminDesigns and implements RAG systems with chunking, embedding generation, vector stores, hybrid search, reranking, and evaluation.