pmbstyle/Alice

What it solves

Alice provides a centralized, voice-enabled AI companion for the desktop that bridges the gap between a simple chatbot and a functional OS assistant. It solves the problem of fragmented AI tools by integrating LLMs, system-level computer control, and personal memory into a single, emotionally aware interface.

How it works

Alice uses a combination of cloud APIs (like OpenAI and DeepSeek) or local LLMs (via Ollama or LM Studio) to process requests. It employs a Go-based backend and an Electron frontend to interact with the local operating system. For memory, it uses Hnswlib for short-term "thoughts" and a local SQLite database for long-term facts. Voice interaction is handled through VAD-powered recognition and TTS engines like Piper or Google TTS.

Who it’s for

It is designed for users who want a persistent AI assistant on their desktop that can perform actual tasks—such as managing files, executing shell commands, and integrating with Google Calendar—rather than just generating text.

Highlights

• Computer Use: Can browse the file system and execute shell commands with granular user permission levels.
• Multimodal Capabilities: Supports screenshot interpretation via Vision APIs and image generation.
• Hybrid Deployment: Works with cloud providers or can run fully locally (experimental) for privacy.
• Memory & RAG: Features local RAG for chatting with documents and a memory system that tracks mood and facts.
• Extensibility: Supports custom tools defined via JSON and local scripts, as well as custom video avatars.