Sem 47 sha based drift detection

[Kadajett]

• Implements SHA-based drift detection for the layered index system, replacing time-based staleness checks

  • Adds DriftStatus struct to track indexed vs current SHA, changed files, and drift percentage
  • Adds UpdateStrategy enum with threshold-based strategy selection (Fresh, Incremental, Rebase, FullRebuild)
  • Detects merge-base changes for branch layer rebase scenarios
  • Working layer checks uncommitted changes via git status

  Test plan

  • 13 tests passing (7 TDD required + 6 additional)
  • test_same_sha_reports_fresh - Old project with same SHA reports fresh
  • test_different_sha_reports_stale - Different SHA correctly detected
  • test_strategy_incremental_under_10_files - <10 files → Incremental
  • test_strategy_rebase_under_30_percent - <30% → Rebase
  • test_strategy_full_rebuild_over_30_percent - ≥30% → FullRebuild
  • test_merge_base_change_detected - Merge-base movement detected
  • test_working_layer_checks_uncommitted - Working layer detects uncommitted changes

  Closes SEM-47

[linear]

SEM-47 SHA-Based Drift Detection

Overview

Implement drift detection using git SHA comparison instead of timestamps.

Key Insight

Time-based staleness is meaningless:

• Old project with same SHA = FRESH (nothing changed)
• 5-minute break with half app rewritten = STALE (everything changed)

Drift Magnitude → Strategy

Drift	Strategy
0 files	No action
< 10 files	Incremental update
< 30% of repo	Rebase overlay
30% of repo

| Full rebuild |

TDD Requirements

#[test] fn test_same_sha_reports_fresh() { }
#[test] fn test_different_sha_reports_stale() { }
#[test] fn test_strategy_incremental_under_10_files() { }
#[test] fn test_strategy_rebase_under_30_percent() { }
#[test] fn test_strategy_full_rebuild_over_30_percent() { }
#[test] fn test_merge_base_change_detected() { }
#[test] fn test_working_layer_checks_uncommitted() { }

Deliverables

• DriftStatus struct:
  • is_stale: bool
  • indexed_sha: String
  • current_sha: String
  • changed_files: Vec<PathBuf>
  • drift_percentage: f64
• UpdateStrategy enum: Fresh, Incremental(Vec<PathBuf>), Rebase, FullRebuild
• check_drift(layer: LayerKind) -> DriftStatus
• Strategy selection based on drift magnitude

Acceptance Criteria

• Old project with same SHA reports as fresh
• Detects changed files via git diff
• Correctly identifies when merge-base moved (branch layer)
• Working layer checks uncommitted changes via git status
• Strategy selection matches drift magnitude thresholds

[Copilot]

Pull request overview

This PR implements SHA-based drift detection for the layered index system, replacing time-based staleness checks. The implementation correctly recognizes that drift should be measured by content changes (SHA differences) rather than time elapsed.

Key changes:

• Adds DriftStatus struct to track indexed vs current SHA, changed files, drift percentage, and merge-base information for branch layers
• Introduces UpdateStrategy enum with threshold-based strategy selection (Fresh, Incremental, Rebase, FullRebuild) based on the magnitude of changes
• Implements layer-specific drift detection for Base, Branch, Working, and AI layers
• Includes comprehensive test coverage (13 tests) covering all TDD requirements and edge cases

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
src/drift.rs	New module implementing SHA-based drift detection with DriftStatus, UpdateStrategy, DriftDetector, and helper functions. Includes comprehensive tests covering all layer types and update strategies.
src/lib.rs	Adds re-exports for drift detection types (DriftDetector, DriftStatus, UpdateStrategy, count_tracked_files) and removes unused SemanticSummary import from benchmark module.
src/benchmark.rs	Auto-formatter changes breaking long format! strings across multiple lines and removing unused SemanticSummary import. No functional changes.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Edge case handling could be clearer: When is_stale is true but changed_count == 0, returning Fresh seems contradictory. This can theoretically occur if git reports a SHA change but no file changes (e.g., commits with only metadata changes). Consider adding a comment explaining this edge case:

// Edge case: SHA changed but no files changed (e.g., empty commit, metadata-only change)
// In this case, no actual update is needed despite the SHA difference
if changed_count == 0 {
    UpdateStrategy::Fresh

Suggested change

	// Edge case: SHA changed but no files changed (e.g., empty commit, metadata-only change)
	// In this case, no actual update is needed despite the SHA difference