paiOS (Personal AI Operating System)

The Open Source Operating System for Personal Intelligence.

Documentation • Website • Contributing

---

🔮 The Vision

paiOS is more than just an operating system; it is the foundation for a new category of Personal AI Hardware.

Our mission is to build a privacy-first "Second Brain" that lives with you, understands your context, and operates independently of the cloud. We believe your intelligence should belong to you-running on trusted hardware, physically under your control.

To achieve this, we are building a unified, modular OS that powers multiple form factors, starting with paiLink (Portable AI Server) and evolving into paiGo (standalone AI Wearable).

⚡️ Form Factors

paiOS is designed to run on diverse hardware, bringing local intelligence to where you need it most.

1. paiLink (Your Local AI Server)

Currently in Development

A powerful Portable AI Server (based on the Radxa Rock 5C) that connects via USB and acts as a private AI backend for your existing tools.

The Idea: Plug it in, and any app that supports Ollama (or a compatible API) can use paiOS for inference. Your data stays on the device – no cloud required.

Compatible Apps (Examples):

• Coding: VS Code (Continue, CodeGPT), Zed, Neovim
• Writing: Obsidian, LibreOffice, OnlyOffice
• Chat: Open WebUI, Lobe Chat

Key Features:

• 🔌 Standard API: Ollama-compatible endpoint for seamless integration.
• 🔒 Air-Gapped Privacy: Inference happens on the Portable AI Server, never in the cloud.
• ⚡ Hardware Acceleration: NPU/GPU powered for real-time responses.

2. paiGo (The AI Wearable)

The North Star

The evolution of paiLink into a standalone, battery-powered AI wearable. It includes all capabilities of paiLink (local AI server, Ollama-compatible API) plus integrated sensors for contextual awareness.

Equipped with a camera and microphone array, it passively captures context to answer questions like "Where did I leave my keys?" or "What was the action item from that meeting?".

Everything paiLink can do, plus:

• 📷 Vision: Onboard camera for multimodal AI (Vision-Language models).
• 🎙️ Always-On Audio: Continuous, private voice capture.
• 🔋 Standalone: No phone or PC required – works independently.
• 🧠 Contextual Memory: "What did I discuss with Sarah last Tuesday?"

---

🆚 Why paiOS?

Feature	☁️ Cloud AI (SaaS)	🔐 paiOS (Local)
Privacy	Your data trains their models.	Your data never leaves the device.
Ownership	Renting intelligence (Subscription).	Owning the intelligence (One-off).
Latency	Network dependency (>500ms).	Real-time Neural Engine (<100ms).
Usage	Pay per token / Rate limits.	Unlimited. Use it as much as you want.
Trust	"Trust us, we're compliant."	"Verify it yourself." (Open Source).

---

🗺️ Roadmap

We are currently in Phase 0 (Foundation).

• Phase 0: Foundation: Groundwork, Architecture Definitions, CI/CD, and Documentation infrastructure.
• Phase 1: paiLink: Launching the Portable AI Server in Developer (Radxa Rock 5C) and Professional (Aluminum Case) editions.
• Phase 2: paiGo: Evolution into a standalone, multimodal AI Wearable.

👉 View Detailed Roadmap

---

🏗️ Architecture & Tech Stack

paiOS is built on Hexagonal Architecture (Ports and Adapters) to ensure that the core logic ("The Brain") is completely decoupled from the specific hardware it runs on. This allows paiOS to run on paiLink today and paiGo tomorrow.

The Stack

• Language: Rust (Safety, Performance, Concurrency)
• Build System: Debos (Reproducible, immutable Debian-based OS images)
• Documentation: Starlight (Modern, accessible, version-controlled)

System Design

graph TD
    User(("👤 User")) -- "Voice / Buttons" --> Engine
    Engine["⚙️ paiEngine\n(Rust · Tokio)"] -- "REST / gRPC / MCP" --> Clients
    Clients["💻 Host Clients\n(OpenWebUI, Obsidian, Cursor)"] -- "Configures" --> Engine
    Engine -- "V4L2 / ALSA / RKNN" --> HW["🔧 Hardware\n(RK3588S2 NPU · Camera · Mic)"]
    Engine -. "Opt-in fallback" .-> Cloud["☁️ Cloud AI"]

Loading

• Hexagonal Architecture: The engine core (domain) is fully decoupled from hardware via Rust traits (ports). Adapters implement those traits — swap hardware without touching business logic.
• Security by Design: Hardware access (camera, microphone, NPU) is designed to flow through paiEngine, where policy and permission checks are centralized. This helps ensure recording behavior is user-controlled.
• Platform Flexibility: paiEngine is a flexible API platform. For wearables (paiGo), the smartphone app (paiApp) is the primary interface for configuration and interaction. Developers can build custom GUIs/HMIs on top via REST, gRPC, or MCP APIs.
• Single Source of Truth: For deep dives into the architecture, see the Architecture Documentation.

🏁 Quick Start for Developers

This is a monorepo containing the full stack. Dependencies include Rust (latest stable) and debos.

1. Build the Engine

cd engine
cargo build --release
# Run tests
cargo test

2. Build the OS Image

Requires Docker

# 1. Build the pai-engine .deb (placeholder binary by default)
bash os/packaging/build-deb.sh

# 2. Build the image via the godebos/debos container
docker run --rm \
  -v "$(pwd):/build" -w /build \
  --cap-add SYS_ADMIN \
  godebos/debos \
  --disable-fakemachine -t machine:radxa-rock5c os/image.yaml

Note: For flashing instructions, hardware targets, and the full OS build guide, see the Documentation.

🤝 Contributing

We are building the user-centric, privacy-first alternative to Big Tech AI.

1. Check the Contributing Guide.
2. Pick an issue from GitHub Issues.
3. Sign our CLA (one-time) - the CLAassistant bot will guide you.

📜 License

• Engine & OS: GNU AGPL v3 (Viral Open Source)
• Apps & SDK: MIT (Ecosystem Friendly)