Struct SharedClassifier 

pub struct SharedClassifier(/* private fields */);

Shared, lockable handle to a single loaded IntentClassifier.

One handle = one loaded model session. Cloning is O(1) and recirculates the same underlying session; it does NOT duplicate model weights or fan out to a parallel session. To serve N concurrent inferences, allocate N independent SharedClassifier instances (one IntentClassifier::load each) — see NL07’s ClassifierPool.

Implementations

impl SharedClassifier

pub fn new(classifier: IntentClassifier) -> Self

Wrap an already-loaded IntentClassifier for shared, lock-mediated access.

pub fn lock(&self) -> MutexGuard<'_, IntentClassifier>

Acquire the lock for &mut access (required by ort::session::Session::run).

Use the returned guard within a small scope; do NOT hold it across .await points (the wrapping crate is sync-only anyway, but this rule still applies to any future async wrapper).

The returned parking_lot::MutexGuard is itself #[must_use], so dropping it on the floor is a clippy warning at the call site.

pub fn identity(&self) -> usize

Stable identity for this slot’s underlying Arc<Mutex<_>>.

Distinct loaded sessions = distinct identities. Used by the NL07 integration test (sqry-nl/tests/pool_concurrent_load.rs) to assert that the pool’s N slots map to N independent IntentClassifier instances. Cloning a SharedClassifier re-circulates the same Arc and therefore returns the same identity.

Trait Implementations

impl Clone for SharedClassifier

fn clone(&self) -> SharedClassifier

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.