Skip to main content
Reactor is infrastructure for real-time video and world models. You write a Python model and deploy it to Reactor’s GPU network. The Reactor CLI packages your code and weights into a release and deploys it. Reactor handles WebRTC streaming, session management, and delivery.
Contact us if you need deployment access.

​
How it works

1

Install the CLI

Install the CLI, then start Docker.
2

Scaffold a workspace

Create a workspace as described in Deploying models.
3

Build your model

Wrap your inference code as shown in the runtime overview.
4

Deploy

Deploy from the workspace as described in Deploying models.
5

Connect clients

Connect your application using the SDK guide.
Running the model on your own machine is optional, and it needs a local GPU for most real models. Test locally covers that loop.

​
Why Reactor

Real-time streaming

Frames delivered over WebRTC as they are generated. Client inputs received live.

Stateful sessions

Your model holds state across a whole session, with hooks for every client that joins or leaves.

Global GPU network

Choose deployment regions and capacity for your clients. Available regions are listed by reactor regions.

No transport code

You never touch WebRTC, WebSockets, or video encoding. Reactor handles it.

Inspect deployment

Check release status, running instances, and model logs from the CLI.

You own your model

Your weights, your inference logic. Reactor never accesses or trains on your data.

​
Get started

Quickstart

Install the CLI, scaffold a project, and deploy your first model.

Build a Model

Tracks, the step loop, commands, and messages.

reactor.yaml reference

Every key in the model spec, with its default.

Install the CLI

Homebrew on macOS, tarballs for Linux and CI.

Deployment

Deploy a workspace and go live on Reactor’s GPUs.