The fastest path from an empty environment to a working notebook. Installation options, hardware requirements, a ten-minute quickstart, and pointers into the rest of the library.
pip, uv, conda, source, and Docker installs with hardware notes.
Supported Python versions, OSes, accelerators, and optional extras.
Attach to any model and record activations in ten minutes.
Compute a deterministic fingerprint over a real model and dataset.
Run a repeatable experiment and inspect the recorded artifact.
The ns command, its subcommands, and where to go from there.
The Scope, Recorder, and Probe primitives in one page.
Jupyter setup, kernel notes, and inline visualization.
The official image, mounts, GPU passthrough, and CI use.
First-run failures, common install errors, and how to file a good bug.