CLI Reference

Usage

edgevox [mode] [options]

Modes

EdgeVox supports three UI modes (mutually exclusive):

Flag	Mode	Description
(default)	TUI	Interactive terminal UI with waveform, slash commands
--web-ui	Web UI	FastAPI server with browser-based interface
--simple-ui	Simple CLI	Minimal terminal interface

Shared Options

These options work across all modes:

Flag	Default	Description
--language	en	Language code (en, vi, fr, ko, de, th, ...)
--voice	auto	TTS voice name
--stt	auto	STT model size (tiny, base, small, medium, large-v3-turbo, sherpa)
--stt-device	auto	Device for STT (cuda, cpu)
--llm	auto	Path to LLM GGUF file or HuggingFace model spec
--tts	auto	Force TTS backend (kokoro, piper, supertonic, pythaitts)
-v, --verbose	off	Enable debug logging

TUI Options

Flag	Default	Description
--mic	system default	Microphone device index
--spk	system default	Speaker device index
--wakeword	none	Wake word phrase (e.g., "hey jarvis")
--session-timeout	30	Seconds of silence before session ends (with wake word)
--ros2	false	Enable ROS2 bridge

Web UI Options

Flag	Default	Description
--host	127.0.0.1	Bind host
--port	8765	Bind port

Simple CLI Options

Flag	Default	Description
--text-mode	false	Text-only mode, no microphone

Examples

# Default TUI
edgevox

# Web UI on all interfaces
edgevox --web-ui --host 0.0.0.0 --port 9000

# Vietnamese with Sherpa STT
edgevox --language vi

# Korean with Supertonic TTS
edgevox --language ko --voice ko-M2

# German with specific Piper voice
edgevox --language de --voice de-thorsten-high

# Text-only mode for testing LLM
edgevox --simple-ui --text-mode

# Wake word mode with 60s timeout
edgevox --wakeword "hey pilot" --session-timeout 60

# Specific audio devices
edgevox --mic 2 --spk 4

# Custom LLM model from HuggingFace
edgevox --llm hf:bartowski/Phi-4-mini-instruct-GGUF:Phi-4-mini-instruct-Q4_K_M.gguf