Every runnable NeuronScope program touches exactly these symbols. They live in the top-level namespace so imports stay short.
Symbols
| Kind | Symbol | Status | Summary |
|---|---|---|---|
| class | neuronscope.Scope | Stable | The context manager that binds a model, a probe, and an output artifact for a single recording. |
| class | neuronscope.Recorder | Stable | The framework-agnostic recorder invoked by Scope. Installs hooks, streams observations, seals the artifact. |
| class | neuronscope.Probe | Stable | Base class for observations. Declares which layers to touch and how to summarise the tensor. |
| class | neuronscope.Session | Stable | Groups related experiments into a single directory with a shared plan. |
| function | neuronscope.record | Stable | One-line shorthand for Scope. Prefer this in application code. |
| function | neuronscope.seed_all | Stable | Sets seeds in every framework the recorder can see. Required for deterministic runs. |
| constant | neuronscope.__version__ | Stable | The installed library version as a string. Currently 0.9.2. |
How they compose
A Scope is a thin composition of one Recorder and one or more Probe instances. The Session exists to keep several related recordings in one place — it is a convenience, not a semantic layer.