Struct SemanticIntentClassifier 

pub struct SemanticIntentClassifier { /* private fields */ }

Classifies query embeddings into intent categories using registered prototype embeddings and weighted cosine similarity.

Implementations

impl SemanticIntentClassifier

pub fn new(config: ClassifierConfig) -> Self

Creates a new classifier with the given configuration.

pub fn register_prototype(&mut self, prototype: IntentPrototype)

Registers a new intent prototype.

Multiple prototypes may be registered for the same IntentKind. When classifying, the maximum weighted similarity across all prototypes of a given intent is used.

pub fn classify( &mut self, query_embedding: &[f32], ) -> Option<IntentClassification>

Classifies query_embedding against all registered prototypes.

Returns None when:

• No prototypes are registered.
• The best weighted cosine similarity is below config.min_confidence.

Otherwise returns an IntentClassification with the winning intent, its confidence score, and an optional runner-up.

pub fn remove_prototype(&mut self, intent: &IntentKind)

Removes all prototypes whose intent matches intent.

pub fn stats(&self) -> &ClassifierStats

Returns a reference to the accumulated statistics.

pub fn prototype_count(&self) -> usize

Returns the number of registered prototypes.

Trait Implementations

impl Debug for SemanticIntentClassifier

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.