semtools

Enables meaning-based search across code and documents using embeddings; supports document parsing and workspace caching.

• Semantic search by meaning, not keywords
• Supports PDF, DOCX, PPTX document parsing with API key
• Manages workspaces for fast repeated searches
• Uses embedding-based similarity matching for code and text

• Finds code by concept, not exact keywords
• Caches embeddings in workspaces for fast repeated searches
• Supports context lines and similarity thresholds
• Converts documents to searchable markdown (requires API key)
• Works with mixed code and text files
• Provides top-K results and configurable similarity distance

1. 1.Run `search` with a concept to find relevant code or text
2. 2.Use `workspace` to index large codebases and cache embeddings
3. 3.Parse documents with `parse` (requires API key)
4. 4.Combine semantic search with exact string search (ripgrep) for follow-up
5. 5.Set environment variables for API keys and workspaces