Welcome to paiOS
paiOS is the open-source, local-first operating system powering the PAI (Personal AI) ecosystem. It provides the software stack (Engine + OS) that enables PAI hardware products to process voice, vision, context, and intelligence directly on trusted hardware, giving you full control over your data. Visit aurintex.com for more information on the PAI vision.
The Vision
Section titled “The Vision”PAI creates a bridge between powerful local hardware and your daily digital life. Our mission is to democratize access to hardware-accelerated AI while guaranteeing data sovereignty.
1. The PAI Accelerator (Current Focus)
Section titled “1. The PAI Accelerator (Current Focus)”The project starts with a USB AI Accelerator based on the Radxa Rock 5C SBC. This device connects to your PC and acts as a dedicated AI server.
- Local Inference: Runs sLLMs, STT (Speech-to-Text), embeddings, and other specialized AI models entirely offline.
- Universal Compatibility: Acts as a Linux USB gadget, compatible with any host OS (Windows, macOS, Linux).
- API-First: Exposes OpenAI-compatible endpoints (similar to Ollama) for integration with tools like Obsidian, LibreOffice, or VS Code.
- Intelligent Input & Control: Acts as a virtual keyboard/mouse (HID) to “type” text or execute commands directly into your PC, enabling seamless AI-driven interaction.
2. The PAI Wearable (Future)
Section titled “2. The PAI Wearable (Future)”The long-term vision is to shrink this technology into a standalone AI Wearable. By building a robust OS and Engine today on the Accelerator, we lay the foundation for a future device that captures context and assists you throughout your day - completely privately.
Technology Stack
Section titled “Technology Stack”paiOS is built for performance, security, and reproducibility.
- Core Engine: Built in Rust for memory safety and low-latency hardware abstraction (HAL).
- Operating System: Minimal, embedded Linux (Debian-based) built with Debos for reproducibility.
- Hardware: Optimized for NPU-based SBCs (Single Board Computers), starting with the Radxa ecosystem.
Repository Structure
Section titled “Repository Structure”This monorepo contains the entire paiOS software stack:
| Directory | Description |
|---|---|
engine/ | The paiOS Engine. The Rust-based neural orchestrator, handling HAL, Security, and IPC. |
os/ | paiOS build configurations. Reproducible embedded Linux images. |
apps/ | Modular extensions and reference implementations for the PAI ecosystem. |
docs/ | This documentation site (built with Starlight). |
Current Status
Section titled “Current Status”🚧 Phase 0: Foundation & Architecture
We are currently executing the “Great Pivot”: Transitioning from internal Proof-of-Concept to this clean, production-ready open-source architecture.
- Architecture: Clean Architecture (Rust) & Modular HAL
- License: Multi-license monorepo following “Open Engine, Proprietary Fuel” model:
os/: AGPL-3.0 (Debos recipes & RAUC configs)engine/src/: AGPL-3.0 (Core engine - all Rust code)engine/libs/: AGPL-3.0 (Hardware bindings - sys-crates)engine/proto/: MIT (gRPC API definitions)apps/: MIT (Open Source Core Apps - reference implementations)hardware/: CERN-OHL (Open Hardware License - CAD & schematics)docs/: AGPL-3.0 (This documentation site)
- Focus: Establishing the
pai-engineandpaiOSbuild pipeline.
Get Involved
Section titled “Get Involved”paiOS is an open-source project driven by the belief that AI should empower users with local-first intelligence and complete data sovereignty.
- Getting Started: Set up your development environment.
- Contributing: Read about our DCO process and how to contribute.
- Architecture: Dive deep into the system design.