Your workers, your machines,
your rules.
The open-source distributed task queue. From simple scripts to GPU-intensive inference—your workers handle it all. Runqy distributes the tasks.
Your compute belongs to you. So does your code.
Runqy is a task broker. It receives jobs, distributes them to your workers, and tracks execution. The workers run on your machines—with resources you control.
Got an RTX 4090 under your desk? Run GPU inference in production. A beefy server for CPU-heavy scraping? Add it to the pool. Need more capacity? Spin up cloud workers for peak loads. Runqy orchestrates. You stay in control.
How it works
Write your task
Pure Python. No DSL. No config files. Just your code with two decorators.
Push to GitHub
Workers pull your code automatically. They clone, install deps, and signal ready—zero manual deployment.
Enqueue jobs
Via CLI, SDK, or API. Runqy distributes to available workers. Results come back via callback.
Your infra, your way
Mix environments as you see fit. Add workers when you need them. Remove them when it's quiet.
Zero-touch deployment
Push code to GitHub. Workers deploy themselves. No SSH. No Docker builds. No CI pipelines to maintain.
Use cases
Image Generation
Stable Diffusion, SDXL, Flux. Enqueue a prompt, get an image. Workers process in parallel on your machines.
LLM Inference
vLLM, Transformers, llama.cpp. Stream tokens to your app via callbacks. The model runs on your hardware.
Video Generation
Long-running jobs on your own hardware. Get notified when it's done.
Web Scraping
Parallel crawling across workers. Scale your scrapers horizontally without managing infrastructure.
Data Pipelines
ETL jobs, data transformations, report generation. Offload heavy processing to dedicated workers.
Batch Processing
Thousands of documents to process? Enqueue them all, workers split the load.
What this changes
Free and open source
No vendor fees. No per-task pricing. Self-host the broker, run workers on your own machines.
Use hardware you already have
Your gaming GPU, your lab server, your render machine. Anything can become a Runqy worker.
Mix and match providers
Your local machine, on-prem servers, cheap GPUs on Lambda, burst capacity on AWS. Mix providers freely, find the best price.
Zero migration pain
Switching cloud providers? Moving in-house? Your workers move, your code stays the same.
Built for real workloads
Simple CPU tasks finish fast. Complex GPU jobs get the time they need. Long timeouts for slow generations. Callback support for streaming. Priority queues to separate urgent jobs from batch.
Whether your job takes 500ms or 10 minutes, Runqy handles it.
Pure Python. Code-first.
Two decorators. One function call. That's all it takes to turn your Python into a distributed worker.
from runqy_python import task, load, run
@load
def setup():
# Runs once at startup (load your model here)
return {"model": load_my_model()}
@task
def process(payload, ctx):
# Handles each task
result = ctx["model"].predict(payload["data"])
return {"prediction": result}
run() # Enters the task loopGet started
Install the SDK. Write your worker. Run it on your machine. Enqueue your first job.
Your workers, your machines, your rules.