Active Open source · MIT

Veles

Millisecond-latency hybrid code search over a persistent on-disk index — no GPU, no API calls.

Language Rust

Started 2026

Stack tree-sitter · model2vec · BM25 · MCP · gRPC

Role Creator

View on GitHub→

Overview

Fast hybrid (BM25 + semantic) local code search for AI agents and humans, written in pure Rust.

Veles interactive TUI in action

Tech: Rust, tree-sitter, model2vec, BM25, MCP, gRPC (tonic), Tokio
Purpose: Millisecond-latency code search over a persistent on-disk index — no GPU, no API calls
Features: Hybrid retrieval with Reciprocal Rank Fusion, identifier-aware tokenizer, query-type detection, definition boosting, file saturation
Symbols: Tree-sitter symbols / defs / refs for Rust, Python, JavaScript, TypeScript, Go
Interfaces: CLI, interactive TUI, MCP stdio server, gRPC service, six pipe-friendly output formats
Distribution: Published crates (veles-cli, veles-core), prebuilt binaries for macOS / Linux / Windows, Homebrew tap
Focus: CPU-only static embeddings, persistent + incremental indexing, AI-agent integration (Claude, Cursor)