Struct ExtractionSchema 

pub struct ExtractionSchema {
    pub relations: Vec<LabelSpec>,
    pub entity_types: Option<Vec<LabelSpec>>,
    pub top_per_pair: Option<usize>,
}

The relations and entity types to extract from one request.

Schema label embeddings are computed once and cached per exact label list, so a stable schema costs nothing after the first request.

Fields

relations: Vec<LabelSpec>

Relations to score for every ordered entity pair. Empty means no facts are extracted.

entity_types: Option<Vec<LabelSpec>>

Entity types to extract. None falls back to the model’s coarse types (kniv v7: LOC, MISC, NUM, ORG, PER, TIME).

top_per_pair: Option<usize>

How many relations to keep per entity pair. None uses the model’s calibrated rel_top (kniv v7: 4).

Trait Implementations

impl Clone for ExtractionSchema

fn clone(&self) -> ExtractionSchema

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExtractionSchema

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

Formats the value using the given formatter.

impl Default for ExtractionSchema

fn default() -> ExtractionSchema

Returns the “default value” for a type.

impl Eq for ExtractionSchema

impl PartialEq for ExtractionSchema

fn eq(&self, other: &ExtractionSchema) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for ExtractionSchema