Pixel vs graphify

A skill that turns a codebase, with its docs, schemas, PDFs and media, into a knowledge graph an agent can query.

Updated

Short answer

graphify builds a knowledge graph of code and of everything around it: docs, SQL schemas, PDFs, images, video. Code is parsed with tree-sitter and no LLM; docs, PDFs and images go through a model, and video is transcribed locally before it does. Pixel stays on code and Git: callers, blast radius, task scope, history search and guarded Git writes, in one binary.

Side by side

Not benchmarked. No figure: how Pixel and graphify work, side by side. Nothing on this page is a measurement.

MeasurePixelgraphify
Reads docs, PDFs, mediaNo, code onlyYes: a model, after local transcription for video
Graph viewNoInteractive HTML
Blast radius before an editYes, pixel impactPaths between two nodes, PR impact
Git historySearch and archaeology, guarded writesRebuilds on commit

Where graphify wins

Non-code inputs (docs, PDFs, media, database schemas), community clustering and an interactive graph view.

How they differ

graphify parses code with tree-sitter into a graph written to local files, clusters it into communities, and adds docs, SQL schemas, PDFs and images through a model; video and audio are first transcribed locally with faster-whisper, and the transcripts then go through the same model. An agent reaches it through a /graphify skill, a CLI, hooks or an optional MCP server. Pixel indexes code only, and answers the questions an edit raises: pixel impact before it, pixel scope-task to find the files, pixel dig-history for why the code is the way it is, pixel commit to land it.

Choosing

Use graphify when the answer lives outside the code too (a design doc, a schema, a PDF) or when you want to see the project as a graph. Use Pixel for the code-and-Git loop of an agent’s edit. The two hook into the same agents, so they can run side by side.