Module intervention 

Intervention types — data structures for cognitive model intervention.

These types bridge cognitive state (from convergence loop) to model control signals (sampling parameters, logit biases). Part of the intervention architecture that moves Noos from text I/O wrapping to model-internal modulation.

See docs/intervention.md for the full paradigm.

Structs

CognitiveSignals

Application-facing allostatic signals.

CognitiveState

Unified cognitive state snapshot — assembled from convergence loop output.

DeltaModulation

Delta modulation parameters for SSM state injection.

ForwardResult

Result from a cognitive forward pass — logits plus modulation metadata.

HiddenStateStats

Statistics derived from SSM hidden state for cognitive signal extraction.

LayerTarget

Target layer range for SSM delta modulation.

LogitBias

A single logit bias entry — modifies probability of a specific token.

SamplingOverride

Sampling parameter overrides derived from cognitive state.

Enums

DeltaModulationSource

What cognitive signal drove this delta modulation.

InterventionDepth

Model capability levels — what intervention depth is available.