add-listen-hotkey

/add-listen-hotkey

Install a global hotkey that triggers deus listen from anywhere on the OS. Also installs all required dependencies (sox, whisper-cli) and downloads the whisper model.

Step 1 — Install dependencies and whisper model

Run these checks and installs before asking the user about hotkey preferences.

Dependencies

macOS:

# Check sox
command -v sox || brew install sox

# Check whisper-cli (from whisper-cpp)
command -v whisper-cli || brew install whisper-cpp

Linux:

# Check sox
command -v sox || sudo apt install -y sox libsox-fmt-all

# Check whisper-cli — not in apt, build from source if missing
if ! command -v whisper-cli; then
  echo "whisper-cli not found. Build from: https://github.com/ggerganov/whisper.cpp"
  echo "After building, ensure 'whisper-cli' is on your PATH."
  # Pause and ask user to confirm before continuing
fi

Windows:

# Check sox
if (-not (Get-Command sox -ErrorAction SilentlyContinue)) {
  winget install sharkdp.bat  # or: choco install sox.portable
}

# Check whisper-cli
if (-not (Get-Command whisper-cli -ErrorAction SilentlyContinue)) {
  Write-Host "Download whisper-cli from: https://github.com/ggerganov/whisper.cpp/releases"
  Write-Host "Add it to your PATH, then re-run this skill."
  # Stop and wait for user
}

Whisper model

Resolve the model path from WHISPER_MODEL env var, or default to ~/deus/data/models/ggml-large-v3-turbo.bin (Liam's personal config) or ~/deus/data/models/ggml-base.bin (public default).

If the model file doesn't exist, download it:

MODEL_PATH="${WHISPER_MODEL:-$HOME/deus/data/models/ggml-base.bin}"
MODEL_URL="https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.bin"

if [ ! -f "$MODEL_PATH" ]; then
  mkdir -p "$(dirname "$MODEL_PATH")"
  echo "Downloading whisper model (148 MB)..."
  curl -L --progress-bar -o "$MODEL_PATH" "$MODEL_URL"
  echo "Model saved to: $MODEL_PATH"
fi

For ggml-large-v3-turbo.bin (higher accuracy, 1.5 GB):

https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v3-turbo.bin