scarletkc/vexor
Semantic search engine for locating files and code by meaning.
View on GitHub Updated 2026-06-16
What is scarletkc/vexor?
Vexor is a semantic search system that indexes files and source code so users can retrieve them based on conceptual similarity rather than keyword matches. It supports multiple embedding providers and can run entirely locally or with cloud models, exposing consistent functionality across its API, CLI, and experimental GUI.
The system automatically creates and caches indexes on first use, allowing fast subsequent searches. Configuration covers embedding dimensions, concurrency settings, extraction backends, and provider choices such as OpenAI, Gemini, or VoyageAI. An included agent skill lets AI coding assistants invoke the same search capabilities directly.
It is intended for individual developers, teams maintaining large codebases, and autonomous AI agents that must locate configuration, implementation, or test files during workflows.
What you can build with scarletkc/vexor
Locating forgotten files
Retrieve source files when you recall their purpose but not their exact name or folder location.
AI agent file discovery
Enable coding assistants to find relevant modules and tests through semantic queries instead of brittle path traversal.
Codebase exploration
Quickly surface related implementation and test files during onboarding or refactoring tasks.
Install scarletkc/vexor
Install
pip install vexorQuick start
pip install vexor # also works with pipx, uv- 1Install via pip with the command pip install vexor.
- 2Run vexor init to create a configuration file and choose an embedding provider.
- 3Execute a search such as vexor "config loader" to index the current directory and view results.
- 4Use vexor index --path /your/project for explicit indexing before repeated searches.
- 5Import the library in Python with from vexor import index, search for programmatic use.
scarletkc/vexor: pros & cons
Pros
- +Works across CLI, Python API, and desktop without changing core behavior.
- +Supports multiple embedding providers and local or cloud execution.
- +Includes ready-made skills for popular AI coding agents.
- +Caches indexes for fast repeated searches after initial build.
Cons
- –Desktop application is experimental and not actively maintained.
- –Requires initial indexing time on large directories before first search.
- –Configuration and provider setup needed for optimal results.
Did you find this helpful?
Frequently asked questions
No, it can run with local embedding models, though cloud providers are also supported.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other agent skills options worth comparing.
