Greptile Summary

This release strips down search output to reduce token usage by ~50% and extracts CLI code from the protocol layer, adhering to the "Janitor > Visionary" principle from AGENTS.md.

Major changes:

• Search output streamlined: Removed lastModified, made Stable trend implicit (only Rising/Declining emitted), condensed relationships to importedByCount/hasTests, added compact type field merging componentType:layer
• Preflight flattened: From nested evidenceLock/epistemicStress to simple {ready, reason} for agent consumption
• Search quality gate: Evidence lock now blocks edits when searchQuality is low_confidence, preventing "confidently wrong" edits
• CLI extraction: Memory management moved from src/index.ts to src/cli.ts per code separation rules
• Framework-agnostic core: Angular patterns self-register via AngularAnalyzer constructor, snippet extraction delegated to analyzers, golden file scoring now counts unique pattern categories instead of hardcoded Angular checks
• Bug fix: Angular providedIn metadata extraction path corrected

All changes align with project constraints (zero-infra, framework-agnostic, privacy-first) and the release checklist was followed (CHANGELOG, README, capabilities.md all updated).

Confidence Score: 5/5

• Safe to merge - well-executed refactor with comprehensive tests and documentation
• This is a mature release with thoughtful architectural improvements. The changes follow established project principles (AGENTS.md rules), maintain backward compatibility at the protocol level, include comprehensive test coverage for new functionality, and all documentation is updated. The refactoring makes the codebase more maintainable by enforcing framework-agnostic patterns.
• No files require special attention - all changes are clean and well-tested

Flowchart

flowchart TD
    A[search_codebase called] --> B{Check intent parameter}
    B -->|explore or undefined| C[Perform search]
    B -->|edit/refactor/migrate| C
    
    C --> D[Load intelligence data]
    D --> E[Assess search quality]
    
    E --> F{Intelligence available?}
    F -->|No| G[Return basic results]
    F -->|Yes| H[Build evidence lock]
    
    H --> I{Search quality status}
    I -->|low_confidence| J[Force ready=false]
    I -->|ok| K[Calculate evidence score]
    
    K --> L{Evidence score >= 40?}
    L -->|No| J
    L -->|Yes| M{Epistemic stress?}
    
    M -->|Yes conflicts/stale| J
    M -->|No| N[Set ready=true]
    
    J --> O[Add reason to preflight]
    N --> P[Preflight ready without reason]
    
    O --> Q[Flatten preflight to ready, reason]
    P --> Q
    
    Q --> R[Streamline result output]
    R --> S[Return: searchQuality + preflight + results]

_{Last reviewed commit: 126b6b5}