Crate cqs 

cqs - Semantic Code Search

Local semantic search for code using ML embeddings. Find functions by what they do, not just their names.

Features

• Semantic search: Uses E5-base-v2 embeddings (769-dim: 768 model + sentiment)
• Notes with sentiment: Unified memory system for AI collaborators
• Multi-language: Rust, Python, TypeScript, JavaScript, Go, C, Java
• GPU acceleration: CUDA/TensorRT with CPU fallback
• MCP integration: Works with Claude Code and other AI assistants

Quick Start

use cqs::{Embedder, Parser, Store};
use cqs::store::ModelInfo;

// Initialize components
let parser = Parser::new()?;
let embedder = Embedder::new()?;
let store = Store::open(std::path::Path::new(".cq/index.db"))?;

// Parse and embed a file
let chunks = parser.parse_file(std::path::Path::new("src/main.rs"))?;
let embeddings = embedder.embed_documents(
    &chunks.iter().map(|c| c.content.as_str()).collect::<Vec<_>>()
)?;

// Search for similar code
let query_embedding = embedder.embed_query("parse configuration file")?;
let results = store.search(&query_embedding, 5, 0.3)?;

MCP Server

Start the MCP server for AI assistant integration:

use std::path::PathBuf;

// Stdio transport (for Claude Code)
cqs::serve_stdio(PathBuf::from("."), false)?;  // false = CPU, true = GPU

// HTTP transport with GPU embedding (None = no auth)
// Note: serve_http blocks the current thread
cqs::serve_http(".", "127.0.0.1", 3000, None, true)?;

Re-exports

pub use embedder::Embedder;

pub use embedder::Embedding;

pub use hnsw::HnswIndex;

pub use index::IndexResult;

pub use index::VectorIndex;

pub use mcp::serve_http;

pub use mcp::serve_stdio;

pub use note::parse_notes;

pub use parser::Chunk;

pub use parser::Parser;

pub use store::ModelInfo;

pub use store::SearchFilter;

pub use store::Store;

Modules

config

Configuration file support for cqs

embedder

Embedding generation with ort + tokenizers

hnsw

HNSW (Hierarchical Navigable Small World) index for fast vector search

index

Vector index trait for nearest neighbor search

language

Language registry for code parsing

mcp

MCP (Model Context Protocol) server implementation

note

Note parsing and types

parser

Code parsing with tree-sitter

reference

Reference index support for multi-index search

store

SQLite storage for chunks and embeddings (sqlx async with sync wrappers)

Structs

DiffResult

Result of a semantic diff

GatherOptions

Options for gather operation

ProjectRegistry

Global registry of indexed cqs projects

Enums

GatherDirection

Direction of call graph expansion

NlTemplate

Template variants for NL description generation.

Pattern

Known structural patterns

Constants

EMBEDDING_DIM

Embedding dimension: 768 from E5-base-v2 model + 1 sentiment dimension. Single source of truth — all modules import this constant.

Functions

enumerate_files

Enumerate files to index in a project directory.

gather

Gather relevant code chunks for a query

generate_nl_description

Generate natural language description from chunk metadata.

generate_nl_with_template

Generate NL description using a specific template variant.

index_notes

Index notes into the database (embed and store)

normalize_for_fts

Normalize code text for FTS5 indexing.

search_across_projects

Search across all registered projects

semantic_diff

Run a semantic diff between two stores

strip_unc_prefix

No-op on non-Windows platforms