graphify

What's new

Surprising Connections noise reduction

Two new suppression rules stop false positives from dominating the headline output:

• Cross-language INFERRED edges suppressed — in monorepos the call/import resolver matches symbol names across language boundaries (e.g. Python AuthError → TypeScript Member). These calls/uses edges are now zero-scored so they don't crowd out real structural surprises. semantically_similar_to and EXTRACTED edges are unaffected.
• Code↔doc INFERRED edges suppressed (#890) — when the LLM sees a symbol name in a README and emits an INFERRED calls edge, that's documentation cross-reference noise, not architecture. Same suppression applied. Code↔paper edges are preserved (a code file referencing a research paper is a genuine cross-format signal).

god_nodes noise reduction

• JSON key nodes filtered (#890) — the JSON extractor emits one node per key, so generic keys (name, id, type, start, end, key, value, data, items, title, description, version, properties) accumulate positional degree from sibling records rather than architectural meaning. These are now excluded from god_nodes. Domain-specific labels in JSON files still rank normally.

Rationale extraction

• Autogenerated file docstrings suppressed — module-level docstrings in Alembic migrations, Django migrations, and protobuf-generated files are boilerplate, not design intent. They are now skipped. Function docstrings inside migration files (e.g. upgrade()) are still captured.

Other

• Auto-detect symlinks (#887) — follow_symlinks is now enabled automatically when symlinked children are detected in the target directory, no flag needed.
• Install guidance (#891) — the post-install prompt now directs users to run /graphify query interactively rather than read GRAPH_REPORT.md first.