TibiOS

AI at home, within everyone's reach.

La IA en el hogar, al alcance de todos.

The TibiBox: your personal TibiOS node

A distributed Operating System to orchestrate AI compute.

Across heterogeneous infrastructure: the application describes what should run, the Operating System decides if, where, when and how.

Read the vision (PDF)
Rusttibios-core
Python / Rayai-runtime
gRPCproto/

The vision: a distributed network

Vision material for the project — where TibiOS is headed long-term, beyond the current state of the code.

TibiOS breaks the hardware barrier to using and building AI. It's a community of owners where every machine adds up, contributing value and returns to its members and partners.
The best Artificial Intelligence — frontier models — for homes, at zero cost, on basic hardware.

Today, artificial intelligence is built collectively — millions of people contribute compute, data and ideas — but ownership of the models stays concentrated in a handful of companies. TibiOS's vision proposes a different model: a global network where any device (a TibiBox, a server, an idle GPU) contributes resources to a collective, private intelligence, with no central servers and no single points of failure.

The cycle that powers that network is simple: contribute compute, build intelligence out of that shared compute, and return the value generated to those who contributed — the more you contribute, the more you own of the ecosystem.

TibiOS vision as a decentralized network of AI nodes

Watch our vision on YouTube

What is TibiOS

The conceptual contract is simple: applications describe what should run (a Workload); the Runtime decides if it's admitted, where it runs, when resources are committed, and how it runs and is observed.

The Runtime itself is not a scheduler, nor a storage engine, nor a network stack, nor an execution engine — it's the composition of all those domains, each owned by an independent domain, wired once in a Composition Root.

Components

tibios-core

Rust

The Runtime itself: the complete architectural model, a Cargo workspace, and the Composition Root that wires domains and Worker adapters.

ai-runtime

Python / Ray

Heavy AI execution Worker — one of the two interchangeable implementations of the Worker Contract.

proto/

Protocol Buffers

The shared, language-neutral gRPC contract between the Runtime and its Workers.

Everything is a Worker

The application never changes. The same contract (Worker) covers CoreML, TensorRT, ONNX, OpenVINO, llama.cpp and more — the Runtime decides which implementation executes each Workload.

Worker trait diagram: same contract, multiple AI execution implementations

Architectural principles

Architecture before implementation

Describing a capability doesn't imply implementing it. Cost begins when real behavior exists, not when a possibility is described.

Ownership

Every piece of mutable state has exactly one authoritative owner. Coordination between domains always happens through published facts, never shared mutable state.

Architecture

Applications → Runtime API → Runtime (Rust) → Worker Contract → Workers.

TibiOS architecture diagram: applications, Runtime API, Rust Runtime and Workers