Advanced Topics
What a running g.Pype system does, for the reader who installs, operates and deploys it: where a pipeline runs, what each machine installs, what a run records and reports, and what each platform asks for. How the SDK is built is Architecture & Design.
One serialized pipeline across several processes: residency, which edge runs a node, the data socket and its policy, a headless server, and the runtime and the browser at a glance.
What pip install gpype and each extra install, and why
import gpype loads none of them.
Record what every source produced with save_as, and run the same
pipeline on it again with load_from – no amplifier needed.
Where latency comes from – frames, the event margin, OSCAR, your own filters – and how to tell whether a pipeline keeps up.
The session log, what does and does not change a pipeline’s condition, and three ways to hear that a run failed.
Change a node’s values while it runs through its control surface, and watch any output port with a probe.
What Windows, macOS and Linux each need: Bluetooth and keyboard permissions, UDP ports, the GDS service.
Measured event placement, trigger accuracy, load and two-machine synchronisation on every amplifier and platform, with the method.
The layout of CSV, MATLAB, HDF5 and EDF+ files and of a recorded run, and how to read each without g.Pype.
Sending and receiving Lab Streaming Layer streams, what the stream carries, timestamps, LabRecorder, and what to check.
A result or a recording into MNE as a Raw or Epochs, by array or by EDF file, with a montage, live through LSL, and back.
A g.Pype recording into a BIDS dataset with mne-bids: what g.Pype supplies and what you add.
What is built in, why the scikit-learn bridge has an allow-list, training any library on a recording, and deploying a model live.