Sensegrep for VS Code
Semantic + structural code search with AI-powered duplicate detection.
Search your codebase by meaning, not just text. Sensegrep uses AI embeddings and tree-sitter AST parsing to find conceptually relevant code.
Features
- Semantic Search - Natural language queries to find relevant code (
Ctrl+Shift+S)
- Find Similar Code - Select code and find similar patterns across your project (
Ctrl+Shift+F)
- Duplicate Detection - AI-powered detection of logical duplicates with impact scoring
- Structural Filters - Filter by symbol type, exports, async, complexity, language, decorators, and more
- Tree-Shaking - Results collapse irrelevant code, showing only what matters
- Code Lens - Inline similarity and complexity info
- Auto-Indexing - Automatic incremental index updates on file changes
- Semantic Folding - Collapse unrelated code when navigating to results
Supported Languages
- TypeScript / JavaScript (TSX/JSX)
- Python (dataclasses, protocols, decorators, async generators, TypedDict)
Commands
| Command |
Shortcut |
Description |
| Sensegrep: Semantic Search |
Ctrl+Shift+S |
Open semantic search |
| Sensegrep: Find Similar Code |
Ctrl+Shift+F |
Find code similar to selection |
| Sensegrep: Semantic Search (Advanced) |
- |
Search with all filter options |
| Sensegrep: Detect Duplicates |
- |
Analyze project for logical duplicates |
| Sensegrep: Index Project |
- |
Incremental index update |
| Sensegrep: Reindex Project (Full) |
- |
Full reindex from scratch |
| Sensegrep: Toggle Watch Mode |
- |
Enable/disable auto-reindexing |
| Sensegrep: Verify Index Integrity |
- |
Check index is up to date |
| Sensegrep: Show Index Stats |
- |
Display index metadata |
| Sensegrep: Fold Unrelated Code |
- |
Collapse code not related to current result |
| Sensegrep: Export Search Results |
- |
Export results to file |
Configuration
| Setting |
Default |
Description |
sensegrep.autoIndex |
true |
Index project on startup |
sensegrep.watchMode |
true |
Auto-reindex on file changes |
sensegrep.watchIntervalMs |
60000 |
Min interval between reindexing (ms) |
sensegrep.showCodeLens |
true |
Show inline similarity/complexity info |
sensegrep.showDiagnostics |
true |
Show duplicate warnings in Problems panel |
sensegrep.semanticFolding |
true |
Collapse unrelated code on navigation |
sensegrep.duplicateThreshold |
0.85 |
Similarity threshold for duplicates |
sensegrep.embeddings.provider |
local |
local (transformers.js) or gemini (recommended) |
sensegrep.embeddings.model |
BAAI/bge-small-en-v1.5 |
Embedding model |
sensegrep.embeddings.device |
cpu |
cpu, cuda, webgpu, wasm |
Using Gemini Embeddings
For cloud-based embeddings (recommended):
- Get an API key from Google AI Studio
- Run Sensegrep: Set Gemini API Key or set the
GEMINI_API_KEY env variable
- Set
sensegrep.embeddings.provider to "gemini"
Why recommended: Gemini generally gives better semantic search quality and handles much larger token contexts than local embedding models.
Local embeddings remain supported, and support for additional providers/APIs is coming soon.
Requirements
- VS Code 1.85+
- Node.js 18+ (for the embedded search engine)
Links
License
Apache-2.0
| |
