Voice OpenCode - Development on Hold see LegionForge or https://legionforge.org

A cross-platform voice interaction layer for OpenCode AI coding agent.

Overview

Voice OpenCode enables bi-directional voice communication with OpenCode. Speak your coding instructions, hear AI responses - hands-free programming powered by AI.

Features

• Voice Input: Speak naturally to direct OpenCode
• Voice Output: Hear AI responses and status updates
• Cross-Platform: macOS, Windows, Linux support
• Modular Architecture: Pluggable STT/TTS engines
• OpenCode API Integration: Works with any OpenCode-compatible backend

Phase 1 Architecture

┌─────────────────────────────────────────────────────────────────┐
│                     Voice OpenCode CLI                          │
├─────────────────────────────────────────────────────────────────┤
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────────────┐ │
│  │   STT       │    │   Core      │    │   TTS               │ │
│  │  Engine     │───▶│   Logic     │───▶│   Engine            │ │
│  │  (Pluggable)│    │   (Bun)    │    │   (Pluggable)       │ │
│  └─────────────┘    └──────┬──────┘    └─────────────────────┘ │
│                            │                                    │
│                            ▼                                    │
│                   ┌─────────────────┐                          │
│                   │  OpenCode API   │                          │
│                   │  Client         │                          │
│                   └────────┬────────┘                          │
│                            │                                    │
│                            ▼                                    │
│                   ┌─────────────────┐                          │
│                   │  OpenCode       │                          │
│                   │  Server         │                          │
│                   └─────────────────┘                          │
└─────────────────────────────────────────────────────────────────┘

Components

STT Engine (Speech-to-Text)

• Local: faster-whisper (CPU/GPU)
• Cloud: OpenAI Whisper API, Google STT
• Interface: Pluggable adapter pattern

Core Logic (Bun)

• Event-driven conversation flow
• Audio buffer management
• Session state handling
• Command parsing

TTS Engine (Text-to-Speech)

• Local: Coqui XTTS, Piper
• Cloud: Edge TTS, ElevenLabs, OpenAI TTS
• Interface: Pluggable adapter pattern

OpenCode API Client

• HTTP client for OpenCode server
• SSE streaming support
• Session management

Installation

Prerequisites

• Bun runtime
• OpenCode server running (opencode serve)
• Audio input/output device

Quick Start

# Clone the repository
git clone https://github.com/jp-cruz/voice-opencode.git
cd voice-opencode

# Install dependencies
bun install

# Configure (copy and edit)
cp .env.example .env

# Run
bun run start

Configuration

See .env.example for configuration options:

# OpenCode Server
OPENCODE_URL=http://localhost:4096

# STT Configuration
STT_ENGINE=whisper          # whisper, openai, google
STT_MODEL=base              # model size for local

# TTS Configuration
TTS_ENGINE=edge             # edge, elevenlabs, coqui
TTS_VOICE=en-US-AriaNeural  # voice for Edge TTS

# Audio Configuration
AUDIO_INPUT_DEVICE=default
AUDIO_OUTPUT_DEVICE=default

Usage

CLI Mode

# Start voice interaction
bun run start

# Test STT only
bun run test:stt

# Test TTS only
bun run test:tts

# Run with specific config
bun run start --tts=elevenlabs --stt=openai

Keyboard Controls

Key	Action
Space	Push-to-talk (start/stop recording)
Ctrl+C	Exit
M	Mute/unmute TTS
R	Reset session

Development

Project Structure

voice-opencode/
├── src/
│   ├── cli/              # CLI entry point
│   ├── core/             # Core conversation logic
│   ├── engines/          # STT/TTS engines
│   │   ├── stt/         # Speech-to-text adapters
│   │   └── tts/         # Text-to-speech adapters
│   ├── opencode/         # OpenCode API client
│   └── audio/           # Audio utilities
├── scripts/              # Build/dev scripts
├── package.json
└── bun.lock

Running Tests

bun test

Adding New STT/TTS Engines

Implement the adapter interface:

// src/engines/stt/base.ts
interface STTEngine {
  transcribe(audioBuffer: Buffer): Promise<string>;
  start(): void;
  stop(): void;
}

// src/engines/tts/base.ts
interface TTSEngine {
  speak(text: string): Promise<void>;
  stop(): void;
  setVoice(voice: string): void;
}

Roadmap

• Phase 1: CLI with basic voice I/O
• Phase 2: Web UI (like opencode-web with voice)
• Phase 3: Mobile companion app

Related Projects

• opencode - AI coding agent
• opencode-web - Web UI for OpenCode
• Voice_opencode - Another voice wrapper

License

This project is licensed under AGPLv3. See LICENSE for details.

Attribution

This project includes attribution to John Paul "Jp" Cruz. See LICENSE for details.

Project Dependencies

OpenCode - This project interacts with and extends OpenCode, an AI coding agent built by SST. Voice OpenCode is not affiliated with or endorsed by SST.

Zen - Inspired by the Zen project from SST, which provides a beautiful terminal interface for OpenCode.

Big Pickle Model - This project was conceptualized and developed with assistance from the Big Pickle AI model. The Big Pickle model is an AI coding assistant that helps developers build software.

Third-Party Libraries

This project uses the following open-source libraries:

• Bun - JavaScript runtime
• OpenAI - Whisper API for speech-to-text
• Microsoft Edge TTS - Text-to-speech engine