Voice AI
in your terminal.

Talk to AI with your voice. Cross-platform CLI that turns speech into conversation — powered by streaming LLM and real-time TTS.

GitHub Get Started

$ curl -fsSL https://chatbot.longye.dev/install.sh | bash

binary is cb on every channel · pulls latest from github.com/erchoc/chatbot
set GITHUB_TOKEN if rate-limited

Features

Everything you need for hands-free AI conversation, right in your terminal.

🎤

Voice Activity Detection

Smart noise calibration and speech detection. Automatically adjusts to your environment — just speak naturally.

⚡

Streaming Pipeline

LLM responses stream token-by-token. TTS synthesizes each sentence in parallel — no waiting for the full reply.

🔌

Pluggable Providers

Swap ASR, TTS, or LLM providers via trait abstraction. DeepSeek, Claude, GPT — any OpenAI-compatible API works.

🗣

Wake Word

Set a wake word like "Hi Pai" to run as always-on assistant. Only responds when addressed — saves API calls.

🖥

Cross-platform

Native audio via CoreAudio (macOS) and ALSA (Linux). One binary, multiple architectures — Apple Silicon, Intel, ARM64.

🛡

Daemon Mode

Install as a system service with cb install. Auto-start on boot, crash recovery, clean uninstall.

Pipeline

🎤

Microphone Capture

cpal · native rate · VAD

↓

📝

Speech-to-Text

Doubao ASR · WebSocket v3

↓

🧠

LLM Streaming

OpenAI-compatible · SSE

↓

🔊

Text-to-Speech

Doubao TTS · concurrent per sentence

↓

🔈

Speaker Playback

rodio · sequential queue

Commands

cb	Start voice chat (default)
cb config	Interactive setup wizard
cb config show	Display current configuration
cb config set <k> <v>	Set a config value directly
cb install	Install as system daemon service
cb uninstall	Remove daemon and all service files
cb status	Check daemon service status
cb --debug	Enable debug logging

Voice AI in your terminal.