Struct ExtractiveBackend 

pub struct ExtractiveBackend { /* private fields */ }

The public extractive backend. Stateless aside from the name field (configurable via the registry so a project can have multiple extractive entries — e.g. one named “default” and one named “fast”).

Implementations

impl ExtractiveBackend

pub fn new(name: impl Into<String>, tokenizer: Tokenizer) -> Self

pub fn run(&self, content: &str, opts: &CompactOpts) -> String

Run the full pipeline. Public for direct testing without the trait.

Trait Implementations

impl Clone for ExtractiveBackend

fn clone(&self) -> ExtractiveBackend

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExtractiveBackend

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

Formats the value using the given formatter.

impl SummarizerBackend for ExtractiveBackend

fn compact<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, content: &'life1 str, opts: &'life2 CompactOpts, ) -> Pin<Box<dyn Future<Output = Result<String, BackendError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

fn name(&self) -> &str

Config-key name (e.g. “default”, “fast”).

fn model_id(&self) -> &str

Resolved model identifier for params_hash. The default "" is intended only for the extractive backend (which has no model). All cloud / network-backed backends MUST override this with their resolved model id so the cache key partitions across model versions.

fn uses_model_prompt(&self) -> bool

Whether this backend feeds content into a model prompt (and is thus an injection target needing HIGH internal hardening). Prompt-free backends (e.g. extractive TextRank) return false.