Struct AcgLearner | NVIDIA NeMo Relay

Generated from cargo doc --no-deps -p nemo-relay -p nemo-relay-adaptive -p nemo-relay-ffi.

1 pub struct AcgLearner { /* private fields */ }

Learner that derives prompt stability state for ACG.

This learner groups annotated LLM requests by derived ACG profile key, builds prompt IR observations, persists a bounded observation window, and updates the hot cache with the latest stability results.

Implementations

`impl AcgLearner`

impl AcgLearner

`new`

pub fn new(
    agent_id: impl Into<String>,
    observation_window: usize,
    thresholds: StabilityThresholds,
) -> Self

Create a new ACG learner.

Parameters

agent_id: Agent identifier whose observations should be updated.
observation_window: Maximum number of observations to retain per profile.
thresholds: Stability thresholds used during analysis.

Returns

A configured AcgLearner.

Trait Implementations

`impl Learner for AcgLearner`

impl Learner for AcgLearner

`process_run`

fn process_run<'a>(
    &'a self,
    run: &'a RunRecord,
    backend: &'a dyn StorageBackendDyn,
    hot_cache: &'a Arc<RwLock<HotCache>>,
) -> Pin<Box<dyn Future<Output = Result<()>> + Send + 'a>>