feat: add floop derive-edges CLI subcommand

[nvandessel]

Summary

• Add floop derive-edges command to derive similar-to and overrides edges from behavior analysis
• Extract IsMoreSpecific to public similarity.IsMoreSpecific() for reuse
• Add similarity.CountSharedTags() for tag co-occurrence counting
• Tag-overlap rule: behaviors sharing >= 2 tags get connected regardless of similarity score — essential for spreading activation (e.g., git + worktree should associate related concepts)

Flags

• --scope (local/global/both, default "both")
• --dry-run — preview without creating edges
• --clear — remove existing similar-to/overrides edges before re-deriving

Results on current data

Store	Behaviors	Edges created	Connected	Islands
Local	19	8	11	8
Global	147	172	62	85

Global validation: 0 errors.

Test plan

• go test ./internal/similarity/... — IsMoreSpecific + CountSharedTags tests pass
• go test ./internal/learning/... — existing place tests pass (refactor safe)
• go test ./cmd/floop/... — 7 new test functions: proposals, skip-existing, connectivity, histogram, clear, tag-overlap, command flags
• floop derive-edges --dry-run --scope both — preview looks correct
• floop derive-edges --clear --scope both — edges created, connectivity improved
• floop validate --scope global — 0 errors

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

Adds floop derive-edges CLI command to automatically derive similar-to and overrides edges from behavior similarity analysis. The implementation extracts IsMoreSpecific() to a public function in the similarity package for reuse, adds CountSharedTags() helper, and refactors place.go to use the shared implementation.

Key changes:

• Tag-overlap rule: Behaviors sharing >= 2 tags get connected via similar-to edges regardless of score (line 228-229 in cmd_derive_edges.go) — essential for spreading activation between related concepts like git + worktree
• Score-based edges: similarity in [0.5, 0.9) creates similar-to edges; when-condition supersets create overrides edges
• Flags: --dry-run for preview, --clear to remove existing derived edges, --scope for local/global/both
• Proper resource cleanup: Uses defer graphStore.Close() (lines 67, 91)
• Comprehensive tests: 7 test functions covering proposals, deduplication, connectivity, histogram, clear, and tag-overlap

Results: Local store created 8 edges (11/19 connected), global store created 172 edges (62/147 connected). Validation passed with 0 errors.

Confidence Score: 5/5

• Safe to merge - well-tested feature with comprehensive coverage and proper resource management
• Score reflects solid implementation quality: proper defer usage for resource cleanup, comprehensive test coverage (7 test functions), clean refactoring to extract shared logic, and successful validation on real data. The tag-overlap rule addresses a real spreading-activation limitation. No logical errors, security issues, or untested edge cases found.
• No files require special attention

Important Files Changed

Filename	Overview
cmd/floop/cmd_derive_edges.go	New CLI command that derives similar-to and overrides edges via all-pairs similarity analysis. Implements dry-run, clear, and scope flags with proper defer for store.Close()
cmd/floop/cmd_derive_edges_test.go	Comprehensive test coverage with 7 test functions covering edge proposals, deduplication, connectivity, histogram, clear flag, and tag-overlap rule
internal/similarity/specificity.go	Extracted IsMoreSpecific() as public function to enable reuse across packages for determining when-condition supersets
internal/similarity/similarity.go	Added CountSharedTags() helper for tag co-occurrence counting to support the tag-overlap rule in edge derivation
internal/learning/place.go	Refactored isMoreSpecific() to delegate to public similarity.IsMoreSpecific() for code reuse, removing 24 lines of duplicate logic

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[User runs: floop derive-edges] --> B{Check scope flag}
    B -->|local| C[Open local .floop store]
    B -->|global| D[Open global ~/.floop store]
    B -->|both| E[Open both stores]
    
    C --> F[deriveEdgesForStore]
    D --> F
    E --> F
    
    F --> G[Load all behaviors from store]
    G --> H{--clear flag?}
    H -->|yes| I[clearDerivedEdges: Remove existing similar-to/overrides edges]
    H -->|no| J[Build existing edge set for dedup]
    I --> J
    
    J --> K[All-pairs comparison loop]
    K --> L[For each pair: Compute similarity score]
    L --> M[Record in histogram bucket]
    
    M --> N{Check specificity}
    N -->|a when ⊃ b when| O[Propose overrides edge]
    N -->|b when ⊃ a when| O
    
    M --> P{Check similarity}
    P -->|score in 0.5-0.9| Q[Propose similar-to edge]
    P -->|shared tags >= 2| Q
    
    O --> R{Edge exists?}
    Q --> R
    R -->|yes| S[Skip: increment SkippedExisting]
    R -->|no| T[Add to ProposedEdges]
    
    T --> U{--dry-run?}
    U -->|yes| V[Return result without creating edges]
    U -->|no| W[Create edges in store]
    
    W --> X[graphStore.AddEdge for each proposal]
    X --> Y[graphStore.Sync]
    Y --> Z[Refresh PageRank]
    Z --> AA[Compute connectivity stats]
    AA --> AB[Return result]
    V --> AA

Loading

_{Last reviewed commit: 8cb5e41}

[greptile-apps]

_{11 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}