Graphify maps a project into a queryable knowledge graph: code parsing stays local with no LLM, and every edge is tagged extracted or inferred
Synopsis
Graphify-Labs' graphify lets an AI coding assistant type /graphify in a project and map code, docs, PDFs, images and video into a queryable knowledge graph, replacing repeated file searching with graph traversal. Code is parsed with tree-sitter into an abstract syntax tree, deterministically, with no LLM and nothing leaving the machine; docs, PDFs, images and video go through a semantic pass using the assistant's model or a configured API key. Every connection is tagged EXTRACTED (explicit in the source) or INFERRED (resolved by graphify), and the output includes a clickable graph.html, GRAPH_REPORT.md and the full graph.json.
Interpretation
When an AI coding assistant works on a project, it can query a prebuilt project knowledge graph covering code, docs, PDFs, images and video instead of searching through files one by one. The prior practice was repeated searching across files; graphify organizes the same material into a traversable graph, after which a user asks a question, traces the path between two things, or explains one concept directly. Run on the FastAPI codebase, graphify explain "APIRouter" returns the node's source location, community, degree 47 and its connection list, and graphify path "FastAPI" "ModelField" gives a three-hop shortest path.
Building the code graph does not depend on an LLM: code is parsed with tree-sitter into an abstract syntax tree, deterministically, and data does not leave the machine. Docs, PDFs, images and video take a separate semantic channel using the assistant's model or a user-configured API key, so local determinism covers only the code portion. Installation needs Python 3.10+ and uv or pipx; after graphify install the assistant is detected and the skill registered, and typing /graphify . produces three files under graphify-out.
Every edge in the graph carries a confidence tag, so a reader can tell which relations come directly from the source text and which were derived by resolution. A path query's overall confidence is that of its weakest hop, not the average or the maximum across hops. In the sample output, APIRouter's edges to RequestValidationError and Dependant are tagged INFERRED, its edge to .get() is tagged EXTRACTED, and the import edge from __init__.py is tagged EXTRACTED.
Perspective
The demonstrated scope is graph construction and explain and path queries on a single FastAPI codebase, plus mapping of code, docs, PDFs, images and video; it is used through an AI coding assistant, and installation requires Python 3.10+ with uv or pipx.
Code parsing is deterministic and local, but semantic processing of docs, PDFs, images and video depends on the assistant's model or a configured API key; a path query's confidence is that of its weakest hop, so the whole path's reliability is limited by any inferred edge in it.