Trait Extractor

Source

pub trait Extractor {
    // Required method
    fn extract(&self, text: &str) -> Result<Vec<ExtractedFact>, ExtractError>;

    // Provided method
    fn extract_graph(&self, text: &str) -> Result<Extraction, ExtractError> { ... }
}

Expand description

Turns a passage of raw text into atomic, graph-ready facts.

Implement this to plug in any model — a local LLM, a hosted API, or a deterministic rule set — and feed the result straight into crate::MemoryService::remember_extracted.

Required Methods§

Source

fn extract(&self, text: &str) -> Result<Vec<ExtractedFact>, ExtractError>

Extract the atomic facts a reader would remember from text.

§Errors

Returns ExtractError if the backend fails or its output cannot be parsed into facts.

Provided Methods§

Source

fn extract_graph(&self, text: &str) -> Result<Extraction, ExtractError>

Extract the facts and the entity→entity edges and entity attributes the passage states.

Defaults to Self::extract with no relations and no attributes, so every backend written against the fact-only contract keeps compiling and keeps working — it simply builds the bipartite fact↔topic graph it always did. A backend that can read structure overrides this.

§Errors

Returns ExtractError if the backend fails or its output cannot be parsed.

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementations on Foreign Types§

Source §

impl<T: Extractor + ?Sized> Extractor for Arc<T>

Forward Extractor through an Arc, so a shared Arc<dyn Extractor> (e.g. one held by the MCP server) satisfies the X: Extractor bound on crate::MemoryService::remember_extracted.

Both methods are forwarded. Forwarding only extract would silently route every Arc-held backend — which is every backend the MCP server and the bindings use — through the fact-only default, discarding the relations and attributes the inner extractor actually produced.

Source §