reactor-runtime 2.6.0

Reactor runtime with public model API

Reactor Runtime

A Python framework for building real-time, interactive video models. Write your ML code — Reactor handles networking, media transport, and client connections.

Installation

pip install reactor-runtime

Quick Example

from dataclasses import dataclass
from reactor_runtime.interface import Output, Video, ReactorPipeline, InputState, InputField

@dataclass
class MyOutput(Output):
    video: Video

@dataclass
class MyState(InputState):
    prompt: str = InputField(default="a sunny meadow")

class MyModel(ReactorPipeline):
    state: MyState

    def load(self, config):
        self.pipe = load_my_model(config["checkpoint"])

    def inference(self):
        while True:
            frame = self.pipe.forward(prompt=self.state.prompt)
            yield MyOutput(video=frame)

Implement inference() as a generator that yields frames. Clients can update self.state mid-generation — no restart, no re-queue.

Key Features

• Real-time streaming — frames delivered over WebRTC as they're generated
• Live interaction — clients change inputs mid-generation via typed, validated state
• No transport code — no WebRTC, WebSocket, or video encoding to manage
• Scaffold & run — the Go reactor CLI gives you reactor init / reactor run; the Python runtime ships its own reactor-runtime script for standalone use

Standalone usage (no Go CLI)

The runtime CLI is installed as reactor-runtime so it never shadows the Go reactor binary. If you only use the Python runtime:

pip install reactor-runtime
reactor-runtime init my-model
cd my-model
python -m venv .venv && . .venv/bin/activate
pip install -r requirements.txt
reactor-runtime run

The Go reactor CLI delegates to this same reactor-runtime script under the hood; both reactor run (Go) and reactor-runtime run (standalone) execute the same Python entry point.