Problem

The BM25 codebase indexing feature (Issue #265, PR #266) was implemented as a Beta feature but has significant issues:

Current Behavior

1. Users manually build a local BM25 index of their codebase
2. Index is stored in VSCode's globalStorageUri
3. AI refinement queries trigger keyword searches against the index
4. Search results are displayed alongside AI responses

Issues

❌ Manual index maintenance required (users must rebuild when code changes)
❌ BM25 keyword search has limited semantic understanding
❌ Index data persists in globalStorageUri even after feature removal
❌ Claude Code CLI already has built-in codebase understanding (makes this redundant)

Expected Behavior

✅ Claude Code CLI's native codebase understanding should be used instead
✅ No manual index building required
✅ No orphaned data left on user devices

Solution

Completely removed the BM25 codebase indexing feature and added automatic cleanup for existing users.

Key Changes

Removed Files (8):

• src/extension/commands/codebase-index-handlers.ts (489 lines)
• src/extension/services/codebase-index-service.ts (633 lines)
• src/extension/services/vector-search-service.ts (310 lines)
• src/shared/types/codebase-index.ts (263 lines)
• src/webview/src/services/codebase-search-service.ts (585 lines)
• src/webview/src/components/chat/CodebaseSearchResults.tsx (212 lines)
• src/webview/src/components/chat/CodebaseStatusBadge.tsx (147 lines)
• src/webview/src/components/dialogs/CodebaseSettingsDialog.tsx (462 lines)

Modified Files (16):

• Removed @orama/orama and @orama/plugin-data-persistence dependencies
• Removed all BM25 message types and handlers
• Removed BM25 state from refinement store
• Removed BM25 UI components from chat interface
• Removed all BM25 translations (5 languages)
• Removed cc-wf-studio.codebaseReference.enabled setting

Added Cleanup (1):

• src/extension/extension.ts - Added cleanupLegacyBM25Index() function
  • Automatically removes legacy index data on extension activation
  • Runs once per user after upgrade
  • Logged to Output Channel for visibility
  • Tracked as technical debt in Issue Remove temporary BM25 cleanup code after sufficient user migration period #302

Changes

Dependencies

File: package.json

Removed:

"@orama/orama": "^2.2.0",
"@orama/plugin-data-persistence": "^2.2.0"

Extension Activation

File: src/extension/extension.ts

Added automatic cleanup:

async function cleanupLegacyBM25Index(context: vscode.ExtensionContext): Promise<void> {
  // Removes {globalStorageUri}/indexes/ directory
  // Logged to "Claude Code Workflow Studio" Output Channel
}

Message Types

File: src/shared/types/messages.ts

Removed all BM25 message types:

• BUILD_INDEX, CANCEL_INDEX_BUILD, GET_INDEX_STATUS, CLEAR_INDEX, SEARCH_CODEBASE
• INDEX_BUILD_PROGRESS, INDEX_BUILD_SUCCESS, INDEX_BUILD_FAILED
• SEARCH_CODEBASE_RESULT, SEARCH_CODEBASE_FAILED
• GET_SETTING, SET_SETTING (Settings payloads)
• OPEN_FILE_AT_LINE (Search result navigation)

UI Components

Files: MessageBubble.tsx, RefinementChatPanel.tsx, SettingsDropdown.tsx

Removed:

• BM25 search results display
• Codebase search integration
• Settings dialog integration
• @codebase command parsing

Translations

Files: en.ts, ja.ts, ko.ts, zh-CN.ts, zh-TW.ts

Removed all codebaseIndex.* and codebaseReference.* keys

Impact

• Breaking Change: Users who enabled the Beta feature will lose access
• Data Cleanup: Legacy index data automatically removed on next extension activation
• Bundle Size Reduction: ~3,600 lines of code removed
• Dependency Reduction: 2 fewer npm packages (@orama)
• No Migration Required: Cleanup is automatic and transparent

Testing

• Code quality checks passed (format, lint, check)
• Build verification passed (npm run build)
• Manual verification: All BM25 code removed (verified via PR feat: add local codebase indexing with BM25 full-text search (#265) #266 diff)
• Cleanup code tested: Logs to Output Channel, handles missing directory gracefully
• Technical debt tracked: Issue Remove temporary BM25 cleanup code after sufficient user migration period #302 created for cleanup code removal (3-6 months)

Notes

• Why fix: instead of chore:: This is a user-facing change (Beta feature removal) that should appear in CHANGELOG
• Technical Debt: Cleanup code in extension.ts should be removed after 3-6 months (Issue Remove temporary BM25 cleanup code after sufficient user migration period #302)
• CHANGELOG: This will appear as a bug fix (patch version bump)
• Future: Rely on Claude Code CLI's built-in codebase understanding for AI refinement

🤖 Generated with Claude Code