Skip to content

Installation

Platforms: macOS, Linux

Prerequisites

  • Python 3.11+
  • Go 1.21+ (or use pre-built binary)

Build from Source

git clone --depth 1 --branch v1.5.0 https://github.com/upship-ai/giddyanne.git
cd giddyanne
make install

This builds the CLI, creates a Python virtualenv, and symlinks giddy to ~/bin.

To install elsewhere:

make install BIN_DIR=/usr/local/bin  # or any directory in your PATH

Pre-built Binary

If you don't have Go installed:

# Clone the repo (needed for Python server code)
git clone --depth 1 --branch v1.5.0 https://github.com/upship-ai/giddyanne.git
cd giddyanne

# Download binary for your platform from Releases:
# https://github.com/upship-ai/giddyanne/releases
# Move it into the repo root (same directory as http_main.py)

# On macOS, clear the quarantine flag
xattr -d com.apple.quarantine giddy

# Set up Python environment
make python

# Symlink to PATH (preserves path to Python code)
ln -sf "$(pwd)/giddy" ~/bin/giddy

First Run

First install downloads dependencies (~2GB for PyTorch). This is a one-time download.

Then download the embedding model:

giddy install      # Download embedding model (~90MB, one-time)

If you skip this step, the model downloads automatically on first giddy up — but the explicit install step makes the wait visible instead of silently blocking startup.

Initial indexing takes time depending on codebase size: - ~750 files: ~45 seconds on M1 Pro - Runs once at startup, then watches for changes - Subsequent starts are faster (unchanged files use cached embeddings)

Verify Installation

giddy help
# Shows available commands

Uninstall

# Remove symlink
rm ~/bin/giddy  # or wherever you installed it

# Remove the repo
rm -rf /path/to/giddyanne

# Optional: remove cached models
rm -rf ~/.cache/huggingface/hub/models--sentence-transformers--all-MiniLM-L6-v2