Run your Nu program across local processes.
Wrap any Nu tree in Teleport. It runs in a spawned child process. Zero-dep, single-host, all cores.
See it.
Provision workers by tag. Send any subtree to a tagged worker. Stdlib multiprocessing under the hood.
Your program, running across local processes.
nu.mp spawns real child processes and lets you teleport any part of your Nu tree onto them. The body does not know or care where it runs.
Wrap it with the worker provider, point at a worker, done. Zero-dep. No cluster, no Ray, no infra.
What you can do with it.
One import turns a single-process program into a multi-process one.
parallel
Use all your cores.
One driver, many workers. Split the work by tag and let each core chew through its slice. No GIL in the way.
isolation
Crashes stay contained.
A worker blows up? The driver survives. Restart the worker, keep going. Real process boundaries between the parts.
no infra
Prototype the shape you will ship.
Same tree, same Teleport, same tags as nu.cluster. Design your topology on your laptop, swap in Ray when you outgrow one box.
Combines well with.
mp runs the local processes. Other fabrics hold the state and wire the processes together.
nu.cluster
Ray cluster fabric.
Same tree, same Teleport, cluster-scale compute. Move from one box to many without rewriting the topology.
nu.proxy
Fabrics on the wire.
Host a fabric inside an mp worker, reach it from the driver or from other workers. Turns a worker into a shared service.
nu.kv
Durable state fabric.
Persist worker results, weights, metrics. Workers share the same tree of Refs, no manual handoff.
Try Nu.
One command gets you the wheel with every fabric. Then follow the movies tutorial to build a real app in an afternoon.
01 Install
pip install "nustack-py[all]"02 Run the demo
nu demo movies03 Build your app
Browse examplesLike what you see?
The project is young. Star it, join the room, watch what we ship next.