Fractal Search for VS Code
Find what you mean, not just what you type.
Built-in VS Code search matches exact text. Fractal Search indexes your workspace at 3 levels — file, section, and line — so related content surfaces even without exact keyword matches.
Why?
Ever searched for "auth middleware" but the code calls it "session validator"? Built-in search returns nothing. Fractal Search finds it because both files share structural patterns with authentication concepts.
Features
- 3-level indexing — file → section → paragraph, each weighted by relevance
- Related term detection — search "database" also finds files about "query", "schema", "migration"
- Smart ranking — direct matches rank highest, related content surfaces naturally
- Filename search — quick search by file/directory names
- Auto-indexing — workspace indexed on open, re-indexes on save
- Zero config — works out of the box with all major languages
Usage
Cmd+Shift+P → "Fractal Search: Search Workspace"
- Type your query
- Results ranked by multi-level relevance
- Click to open
Settings
| Setting |
Default |
Description |
fractalSearch.maxResults |
20 |
Maximum results |
fractalSearch.maxDepth |
4 |
Search depth (1=shallow, 4=deep) |
fractalSearch.fileExtensions |
.md .txt .py .js .ts .go .rs + more |
File types to index |
How it works
Documents are indexed at three self-similar scales. Results at each level are weighted with natural decay:
File level → weight: 1.0 (direct match)
Section level → weight: 0.618 (contextual match)
Paragraph level → weight: 0.382 (deep match)
Related terms → weight: 0.236 (semantic match)
Terms that frequently appear together across your codebase are automatically linked — no configuration needed.
Supported Languages
Works with any text file. Optimized for: Python, JavaScript, TypeScript, Go, Rust, Java, Markdown, YAML, JSON, HTML, CSS, Shell scripts.
Author
Abhishek Srivastava — github.com/0x-auth
Report issues | Source code