PearlOS is an open-source operating system that aims for autonomous evolution through the use of swarm intelligence. The system is designed to learn, take notes, create applications, and acquire new skills over time.

Architecture and Functionality

PearlOS is based on an architecture that integrates different artificial intelligences through an OpenClaw bridge. The system connects to models via OpenRouter, allowing the use of various models such as Llama, Mistral, and Qwen, both locally via Ollama instances and via cloud APIs. Intelligent routing selects fast models for tasks such as chat and intent classification, and heavier models for code generation and complex reasoning.

Voice Interaction and Output

A distinctive feature of PearlOS is voice integration. Voice output uses PocketTTS for local execution, eliminating dependence on cloud services for core functions. The system also supports ElevenLabs for higher quality voices, but optionally. The voice pipeline is built on Pipecat (Deepgram STT → model → PocketTTS) and handles interruptions and conversation turns.

Development and Contributions

The PearlOS code is available on GitHub. The system is designed to allow the AI itself to modify its own code and propose changes to the repository, facilitating a continuous and collaborative development process.