scarletkc/vexor

What it does

Vexor is a semantic search engine designed for developers to discover files and code repositories by describing what they do functionally, rather than searching by filename or keywords. It creates searchable indexes of your codebase using semantic embeddings, allowing natural language queries to locate relevant code.

Key features

• Semantic search across files and code using natural language descriptions
• Local reusable indexes for persistent, offline searching
• Configurable embedding providers including OpenAI, Gemini, Voyage, and OpenAI-compatible APIs
• Support for fully offline local embedding models
• AST-aware code chunking for intelligent code parsing
• Optional reranking to improve search result relevance
• Cross-platform support (macOS, Windows, Linux, cloud)

How to set up

Install via pip or npm depending on your environment. Configure your preferred embedding provider (cloud-based or local) in the settings. Index your codebase by pointing Vexor to your project directories. Query your indexed code using natural language descriptions of functionality you're looking for.