Problem
contentAtDepth() in src/graph/packer.ts generates descriptive stubs instead of calling depthFilter.ts for real filtered content. The bridge is not wired. The depth system's core value (3-5x more context in same token budget) doesn't actually work in the graph pipeline.
Additionally, calls detection in extractors/code.ts matches too aggressively: any identifier + ( matches against every export.
Success criteria
contentAtDepth() calls depthFilter.ts and returns real depth-filtered content
- COMMON_IDENTIFIERS blocklist extended
- Scope filter reduces false positive calls edges by 5x
Implementation
src/graph/packer.ts → import depthFilter, replace contentAtDepth() stub
src/graph/extractors/code.ts → add scope filter (same-file = ignore)
- Extend COMMON_IDENTIFIERS
- Test: scan modular-patchbay, verify
calls edges < 200
Problem
contentAtDepth()insrc/graph/packer.tsgenerates descriptive stubs instead of callingdepthFilter.tsfor real filtered content. The bridge is not wired. The depth system's core value (3-5x more context in same token budget) doesn't actually work in the graph pipeline.Additionally,
callsdetection inextractors/code.tsmatches too aggressively: any identifier +(matches against every export.Success criteria
contentAtDepth()callsdepthFilter.tsand returns real depth-filtered contentImplementation
src/graph/packer.ts→ importdepthFilter, replacecontentAtDepth()stubsrc/graph/extractors/code.ts→ add scope filter (same-file = ignore)callsedges < 200