CodexA

Appearance

Installation ​

Requirements ​

  • Python 3.11+ (3.12 recommended)
  • pip or uv package manager
  • Git (for version-controlled projects)

Install from PyPI ​

bash
pip install codexa

This gives you the full CLI (39 commands), tree-sitter parsing (11 languages), quality analysis, grep, explain, and all non-ML features.

ML Extras (Semantic Search & Vector Indexing) ​

For embedding-based semantic search, install with the [ml] extra:

bash
pip install "codexa[ml]"

This adds sentence-transformers, torch, and faiss-cpu. Requires ~2 GB RAM for model loading.

TIP

If you only need structural commands (explain, grep, quality, languages), the lightweight pip install codexa is sufficient. Install [ml] when you want codexa search with vector similarity.

Install from Source ​

bash
git clone https://github.com/M9nx/CodexA.git
cd CodexA
pip install -e "."

Install via Docker ​

bash
docker build -t codexa .
docker run --rm -v /path/to/project:/workspace codexa search "auth"

The Docker image includes ripgrep, git, and a pre-loaded embedding model.

Install via Homebrew (macOS) ​

bash
brew install --formula Formula/codexa.rb

Optional Extras ​

Install with optional dependencies for additional features:

bash
pip install "codexa[ml]"
# Includes: sentence-transformers, torch, faiss-cpu
bash
pip install -e ".[dev]"
# Includes: pytest, pytest-cov
bash
pip install -e ".[tui]"
# Includes: textual (interactive terminal UI)
bash
pip install -e ".[dev,tui,ml]"

Verify Installation ​

bash
codexa --version
# codexa 0.4.4

codexa doctor
# Checks environment health

First-Time Setup ​

Initialize CodexA in your project:

bash
cd /path/to/your-project
codexa init

This creates a .codexa/ directory with:

.codexa/
├── config.json     # Configuration (embedding, search, LLM settings)
├── index/          # FAISS vector index (after indexing)
├── cache/          # Query and embedding caches
└── plugins/        # Custom plugins directory

LLM Configuration (Optional) ​

For AI-powered commands (ask, review, refactor, chat, etc.), configure an LLM provider in .codexa/config.json:

json
{
  "llm": {
    "provider": "openai",
    "model": "gpt-4",
    "api_key": "sk-...",
    "temperature": 0.2,
    "max_tokens": 2048
  }
}

Supported providers: openai, ollama (local), mock (testing).

Next Steps ​