git-semantic-1.0.4 is not a library.

Visit the last successful build: git-semantic-1.3.2

git-semantic

Search your git history using natural language - find commits by what they mean, not just what they say.

$ git-semantic search "fixed race condition in authentication"

🎯 Most Relevant Commits:

1. abc1234 - Resolve concurrent login session handling (0.89 similarity)
   Author: Alice Chen, 6 months ago
   
2. def5678 - Synchronize user token refresh logic (0.84 similarity)
   Author: Bob Martinez, 4 months ago

Stop scrolling through hundreds of commits with git log --grep. Just describe what you're looking for in plain English.

Why?

Traditional git search is keyword-based. You need to guess the exact words the author used:

git log --grep="race"     # 847 results 😵
git log -S "mutex"        # Maybe? 🤷

git-semantic understands meaning. Search for "race condition" and find commits about "concurrent access" or "synchronization bugs" - even if those exact words aren't in the message.

Features

🔍 Natural language search - "fix memory leak" finds more than just those exact words
🚀 Fast - Results in < 100ms
🔒 Private - Everything runs locally with ONNX, no API keys or cloud services
📦 Zero config - Works out of the box
🎯 Smart filtering - By author, date, file, and more

Installation

Using Cargo (Recommended)

cargo install git-semantic

Alternatively, you can also install from the latest release compatible with your OS on the releases page.

Quick Start

# 1. One-time setup (downloads AI model, ~130MB)
git-semantic init

# 2. Index your repository
cd /path/to/your/repo
git-semantic index

# 3. Search!
git-semantic search "your query here"

Usage

Basic Search

git-semantic search "fix memory leak"
git-semantic search "add authentication feature"
git-semantic search "refactor payment logic"

Filters

# By author
git-semantic search "refactor" --author=alice

# By date
git-semantic search "bug fix" --after=2024-01-01

# By file
git-semantic search "optimization" --file=src/auth.rs

# Limit results
git-semantic search "feature" -n 5

Index Management

# Update index with new commits
git-semantic update

# Show index statistics
git-semantic stats

# Quick index (messages only, faster)
git-semantic index --quick

# Full index (messages + diffs, more context)
git-semantic index --full

How It Works

Downloads BGE-small-en-v1.5 - A compact AI model (130MB) for semantic embeddings
Indexes your repo - Converts each commit into a 384-dimensional vector
Stores locally - Binary index saved in .git/semantic-index (ignored by git)
Searches by meaning - Your query becomes a vector, finds similar commit vectors using cosine similarity
ONNX Runtime - Fast local inference, no cloud services needed

Stored locations:

Model: ~/Library/Application Support/com.git-semantic.git-semantic/models/ (macOS)
Index: .git/semantic-index (per repository)

Technical Details

Model: BGE-small-en-v1.5 (BAAI)
Runtime: ONNX Runtime for fast local inference
Storage: Bincode serialization (~0.04MB per 7 commits)
Search: Cosine similarity with L2 normalization
Inference: < 100ms per query

Real Example

$ git-semantic search "ONNX integration"

🎯 Most Relevant Commits for: "ONNX integration"

1. 4d8acb9 - docs: Update README with complete ONNX integration details (0.73 similarity)
   Author: yan, 2025-10-13 08:17:23 UTC
   -# git-semantic (IN DEVELOPMENT)
   +# git-semantic

2. 776ff32 - feat: Complete ONNX integration with real BGE embeddings (0.73 similarity)
   Author: yan, 2025-10-13 07:24:37 UTC
   -    let engine = SearchEngine::new(model_manager)?;
   +    let mut engine = SearchEngine::new(model_manager)?;

3. 28e9c31 - Implement ONNX model inference and HuggingFace download (0.69 similarity)
   Author: yan, 2025-10-13 06:50:59 UTC
   +use indicatif::{ProgressBar, ProgressStyle};
   +use ndarray::Array1;

Contributing

Contributions welcome! Please use Conventional Commits format:

feat: add new search feature
fix: resolve memory leak in indexing
docs: update installation instructions

Requirements

Git repository (obviously!)
~130MB disk space for the AI model
Rust 1.70+ (if building from source)

License

MIT

Built with: Rust 🦀 and ❤️

git-semantic 1.0.4