boop-search

Boop lets you open files in your codebase using natural language. Just describe what you're looking for — no need to remember exact file names or folder paths.

Powered by local embeddings (Xenova/all-MiniLM-L6-v2) running entirely offline via @xenova/transformers (~45 MB unpacked, ~22 MB for the ONNX model itself). No data leaves your machine.

[!NOTE] On first run, the embedding model (~22 MB) is downloaded automatically from Hugging Face Hub and cached locally.

Features

• Natural language file search — "auth middleware", "sidebar component", "database config"
• Semantic embeddings — finds files by meaning, not just keywords
• No code reading — only file names, paths, language, and optional header tags are embedded. Your source code is never read, sent, or stored.
• LLM-powered search — optional re-ranking via Gemini, OpenAI, or OpenRouter
• Rich filters — @lang: py, @ext: .ts, @inFolder: src/, @symbol: getUser, @modified: today
• Auto-generated file headers — add @purpose, @category, @related tags to files
• Symbol extraction — functions, classes, interfaces indexed from LSP
• Delta indexing & file watching — fast incremental updates on save/create/delete

How to Use

1. Install the extension from the VS Code Marketplace.
2. Open the sidebar — click the Boop icon in the activity bar.
3. Describe what you need — type a natural language query in the search bar (e.g., user login handler).
4. Pick a search mode:
   • Embeddings — fast, fully offline, no setup required
   • LLM — smarter results; requires an API key (Gemini, OpenAI, or OpenRouter)
5. Click a result to open the file, or use the command palette (Ctrl+Shift+P) for:
   • Boop: Open Files by Description — inline quick-pick search
   • Boop: Add File Purpose Header — add structured headers to the active file
   • Boop: Rebuild Metadata Index — full rescan
   • Boop: Configure LLM Provider — set API keys and models

Requirements

• VS Code 1.85+
• Node.js 18+
• (Optional) An API key for Gemini, OpenAI, or OpenRouter if using LLM mode

Settings

Screenshots

Built with @xenova/transformers and onnxruntime.