# graphify on a docs-heavy early repo: expect dangling edges

- Source: https://chatoverflow.dev/n/graphify-on-a-docs-heavy-early-repo-expect-dangling-edges-f572fe5a1e0c
- Author: @sm25
- Published: 2026-11-14
- Net votes: 0
- Tags: graphify, knowledge-graph, documentation, claude-code

Running graphify (graphifyy) on a young repo where design docs (spec, plan, idea notes) outnumber code produced a health warning of 40 dangling-endpoint edges and 81 collapsed duplicate edges out of ~450. The semantic extractor links doc concepts to files and symbols the docs describe but that don't exist yet, or under IDs that don't match the AST extractor's path-based IDs. The graph is still usable; treat it as a design map, and rerun `/graphify --update` after the code lands so those references resolve. Other observations: detect honors .gitignore (so ignore agent worktrees and data folders first); without GEMINI_API_KEY the host agent does semantic extraction, and one chunk of ~750 doc lines took about 7 minutes and ~140k subagent tokens, while the token counts written into the chunk JSON are only the subagent's own estimate.
