Skip to content

plan3.md

Latest commit

 

History

History
50 lines (41 loc) · 1.89 KB

plan3.md

File metadata and controls

50 lines (41 loc) · 1.89 KB

Plan 3 — Next Steps (from docs/review2.md Part 3)

Cross-Cutting Observations

Fix existing gaps before building new features:

  1. Missing MCP tools — 8 commands lack MCP equivalents.
  2. Missing --model flags — 4 commands embed queries without model selection.
  3. Missing --branch flags — 4 analysis commands lack branch scoping.
  4. Missing HTTP routes — 6 analysis commands lack API equivalents.
  5. Missing --html on 4 search/analysis commands that would benefit.

Architectural recommendations:

  1. SearchPipeline abstraction — composable features (boolean queries → expansion → explanation → saved queries) without duplicating scoring logic.
  2. concept_snapshot table — cache per-commit concept centroids. Enables semantic bisect, lifecycle analysis, and health timeline without recomputing.
  3. FeatureFlags system — config-based toggles for gradually rollable features.

Recommended Implementation Order

Phase 37: Quick Wins (1-2 days each)
  - Code-to-Code Search (Low complexity, High value)
  - Negative Examples Search (Low complexity, High value)
  - Partial/Selective Indexing (Low complexity, High value)
  - Result Explanation (Low-Med complexity, High value)

Phase 38: Medium Effort, High Impact (3-5 days each)
  - Semantic Git Bisect
  - Garbage Collection
  - Boolean/Composite Queries
  - Documentation Gap Analysis

Phase 39: Analysis Features (1-2 weeks each)
  - Refactoring Suggestions
  - Concept Lifecycle Analysis
  - Contributor Semantic Profiles
  - CI/CD Semantic Diff in PRs

Phase 40: Visualization & Scale (2-4 weeks each)
  - Semantic Codebase Map (UMAP/t-SNE)
  - Temporal Heatmap
  - Remote Index Sharing
  - Semantic Cherry-Pick Suggestions

Phase 41+: Large Investments (4+ weeks each)
  - Multi-Repo Unified Index
  - IDE / LSP Integration
  - Security Pattern Detection
  - Codebase Health Timeline
  - Technical Debt Scoring