Command-line interface

Run a common analysis with beginner-friendly defaults

Validate and describe a self-contained VRAVEN model bundle

Create vraven.json for model and input files already in a directory

Run the legacy activation-trace demonstration

Explain one saved PyTorch model invocation and write one report

Run finite-budget empirical adversarial stress testing

Search for a bounded continuous-input counterfactual

Measure score stability under controlled input perturbations

Run epsilon-rule Layer-wise Relevance Propagation

Run formal local L-infinity verification for supported operators

Explain a bounded sample cohort and build a reasoning-fingerprint atlas

Compare two saved models on the same bounded sample cohort

Track one input through two or more ordered model checkpoints

Inspect measured local anatomy of Linear layers for one sample

Run a deterministic compiler-assisted causal decompilation demonstration

Compile a PyTorch model into the architecture-neutral VRAVEN Mechanism IR

Recover concepts, causal mechanism, executable programme and Decision Proof

Inspect a saved legacy VRAVEN trace

Render one legacy standalone figure

Display the optional blinking or dancing terminal raven

Report the local VRAVEN and PyTorch capabilities

List supported commands, modes, formats and visualisations