Work from a client

Below, 192.168.1.20 stands for the server’s address, and gpu-box for a name its certificate carries (one of its --name values, or localhost).

Command line

export DSPK_IDENTITY=laptop.dspk-identity.pem
S="192.168.1.20:50051 --server-name gpu-box"

dspk remote $S info                                   # who it is, who you are, its runtime
dspk remote $S files                                  # the recordings it shares
dspk remote $S sort session1/rec.bin --sorter kilosort4 --wait
dspk remote $S fetch session1/rec.kilosort4.sorting.zarr

Python

import dspk

with dspk.remote.Session("192.168.1.20:50051", "laptop.dspk-identity.pem", server_name="gpu-box") as server:
    server.info()                                 # who it is, who you are, its runtime
    server.files("session1/")                     # the recordings it shares
    rec = server.open("session1/rec.bin")         # a dspk.io.Recording, read on the server
    window = rec.read(start=0, stop=30_000)
    server.sort("kilosort4", "session1/rec.bin")
    status = server.wait()                        # follows it; raises with the logs if it failed
    server.fetch(status["output"], "rec.kilosort4.sorting.zarr")

App

Add the server in Settings (Ctrl+,): its address, its name and your identity file. Then File → Open from server walks the data root. A remote recording opens in Explore like a local one, and sorts run on the server.

What runs where

• Reads of a remote recording are for windows and snippets. Work on a whole recording belongs on the server, as a sort.
• One job runs at a time, and starting or stopping one needs a client enrolled with --control.
• A sorting goes next to the recording unless you name an output, and never overwrites anything.
• fetch writes to a new local path, and refuses one that exists.

See the dspk.remote reference and the CLI reference for every option.