vecgrepLocal-first semantic code search

Index codebases, search with natural language, and keep vectors on your machine by default.

Quick Start

Open Studio

Local-first indexing

Store generated indexes under ~/.vecgrep/projects by default and use Ollama for local embeddings.

Hybrid search

Combine semantic vector search with VecLite BM25 keyword matching and metadata filters.

Studio TUI

Browse results, preview code, index projects, and inspect vector status from a Charm v2 terminal UI.

Cloud-ready providers

Use OpenAI, Cohere, or Voyage AI when a managed embedding provider is the right fit.

What vecgrep Does

vecgrep indexes a source tree into code chunks, embeds those chunks, and stores them in VecLite. You can search with natural language, exact keywords, or a hybrid of both.

The default setup is local-first:

vecgrep init registers projects in ~/.vecgrep/config.yaml.
Vector data is stored under ~/.vecgrep/projects/<project>/.
Ollama provides local embeddings with nomic-embed-text.
Repository-local .vecgrep/ directories are only created when you use vecgrep init --local.

Common Workflows

Goal	Start Here
Install and run your first search	Quick Start
Learn the CLI commands	CLI Usage
Configure storage and providers	Configuration
Use the terminal workspace	Studio
Connect an AI assistant	MCP Integration
Understand VecLite storage	VecLite Integration