Alice

Say "Hi" to Alice 👋, your open-source AI companion designed to live on your desktop.

Alice brings together voice interaction, intelligent context awareness, powerful tooling, and a friendly personality to assist you with everything from daily tasks to deeper creative work. Alice is more than a chatbot; she’s built to feel present, responsive, emotionally engaging, and deeply useful.

Quick showcase

✨ Key Features

💻 Local and Cloud use

Alice is designed to work with Cloud(OpenAI / Codex subscription, OpenRouter, Z.ai, Minimax, Deepseek) and Local LLMs (Ollama/LM Studio). Has built-in speech-to-text, text-to-speech, and embedding services. While the OpenAI cloud API is preferred and provides the best user experience, Alice can also operate fully locally (experimental).

🗣️ Voice Interaction

• Fast, VAD-powered voice recognition (via gpt-4o-transcribe, google-tts-voice or whisper-large-v3)
• Natural-sounding responses with OpenAI/Google TTS and optional support for local multilingual text-to-speech via Piper TTS
• Interruptible speech and streaming response cancellation for smoother flow

🧠 Memory & Context

• Thoughts: Short-term context stored in Hnswlib vector DB
• Memories: Structured long-term facts in local DB
• Summarization: Compact message history into context prompts
• Emotion awareness: Summaries include mood estimation for more human responses
• Local RAG: Add local documents to the LLM context, chat with your docs

🎨 Vision & Visual Output

• Screenshot interpretation using Vision API
• Image generation using gpt-image-2
• Animated video states (standby/speaking/thinking)

🪄 Computer Use Tools

Alice can interact with your local system with user-approved permissions:

• 📂 File system browsing (e.g., listing folders)

• 💻 Shell command execution (ls, mv, mkdir, etc)

• 🔐 Granular command approvals:

  • One-time
  • Session-based
  • Permanent (revocable)

• 🔧 Settings tab "Permissions" lets you review and manage all approved commands

⚙️ Function Calling

• Web search (including Searxng support)
• Google Calendar & Gmail integration
• Torrent search & download (via Jackett + qBittorrent)
• Time & date awareness
• Clipboard management
• Task scheduler (reminders and command execution)
• Open applications & URLs
• Image generation
• MCP server support

💬 Wake Word Support

With the local STT model, you can set a wake-up word (like "Hey, Siri").

• Alice will always listen, but only process requests when the wake word is spoken.
• Default mode is auto language detection, but you can also select a specific language in settings.

💻 Dedicated Chrome Extension

• Ask Alice about your active Chrome tab
• Context menu for selected text on a web page
  • Fact check this
  • Summarize this
  • Tell me more about it

🎛️ Flexible Settings

Fully customizable settings interface:

• LLM provider selection between OpenAI, OpenRouter, DeepSeek, Z.ai(coding plan), Minimax(token plan), Ollama, LM Studio
• Cloud or local TTS, STT, Embeddings
• Model choice & parameters (temperature, top_p, history, etc)
• Prompt and summarization tuning
• Audio/mic toggles & hotkeys
• Available tools & MCP configuration
• Google integrations

🔨 Custom Tools

Alice supports custom tools that are defined in JSON and backed by local scripts.

1. Open Settings → Customization → Custom tools
2. Upload or drop your script (writes to custom-tool-scripts/)
3. Click Add Tool, fill in metadata, and paste the JSON schema. Saving updates custom-tools.json
4. Toggle the tool on/off in the list. Only enabled + valid entries are offered to the model.

🎭 Custom Avatars

Swap Alice's appearance with your own video loops:

1. Create a folder under user-customization/custom-avatars/<AvatarName>/.
2. Drop speaking.mp4, thinking.mp4, and standby.mp4 into that folder (all required).
3. Open Settings → Customization → Assistant Avatar, hit Refresh, and pick the new avatar.

🚀 Download

👉 Download the latest release

Follow the Setup Instructions to configure your API keys and environment.

🛠️ Technologies Used

• Frontend: Vue.js, TailwindCSS
• Desktop Shell: Electron
• State Management: Pinia
• AI APIs: OpenAI, OpenRouter, DeepSeek, Groq
• Backend: Go
• Vector search engine: hnswlib-node
• Local storage: better-sqlite3
• Voice activity detection: VAD (Web)
• Local STT & TTS: whisper.cpp & Piper
• Local Embeddings: multilingual-e5-small (ONNX, 384 dimensions)
• Animation: Kling Pro

Other tools:

• Jackett — Torrent aggregator
• qBittorrent — Torrent client
• Searxng - Self-hosted web search

🧑‍💻 Getting Started (Development)

# 1. Clone the repo
$ git clone https://github.com/pmbstyle/Alice.git

# 2. Install dependencies
$ npm install

# 3. Set up your .env file (see .env.example for reference)

Follow setup instructions to obtain required API credentials.

# 4. Download ONNX Runtime and the pinned multi-lang Memory/RAG model
npm run setup:embeddings

# 5. Compile backend
npm run build:go

# 6. Run dev environment
$ npm run dev

📦 Production Build

Optionally, create an app-config.json file in the root directory for Google integration:

{
  "VITE_GOOGLE_CLIENT_ID": "",
  "VITE_GOOGLE_CLIENT_SECRET": ""
}

# Build the app
$ npm run build

Install the output from the release/ directory.

🤝 Contributing

Ideas, bug reports, feature requests - all welcome! Open an issue or PR, or drop by to share your thoughts. Your input helps shape Alice into something wonderful 💚

A full app overview with tutorials