pub struct Extraction {
pub media: MediaRef,
pub extractor: String,
pub deterministic: bool,
}Expand description
A searchable text extraction of an opaque artifact, plus the provenance a
reader needs to weigh it (doc 18 §7).
When a snapshot’s bytes are a binary the substrate can read deterministically
— a PDF’s embedded text layer, an HTML document’s text — mint extracts them to
this text blob so read/search/coverage work over the artifact directly.
The extraction is signed with the core, so the token carries the searchable
text instead of leaving it as a loose file the next agent must locate.
deterministic is the load-bearing field. A text-layer extraction is a pure
function of the bytes and reproduces exactly (true). A model-produced
transcription — OCR, ASR — is an opinion that drifts (false), and stamping it
as such is what lets a coverage receipt over the extraction be read at its true
worth: the substrate refuses to let a hallucinated transcript pass for the
artifact. The substrate never defaults to a non-deterministic extractor.
Fields§
§media: MediaRefThe extracted text, content-addressed (text/plain).
extractor: StringWhich extractor produced it — e.g. pdf-textlayer, html-strip.
deterministic: booltrue when the extraction is a pure, reproducible function of the bytes;
false when a model produced it and the text is an opinion.
Trait Implementations§
Source§impl Clone for Extraction
impl Clone for Extraction
Source§fn clone(&self) -> Extraction
fn clone(&self) -> Extraction
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source. Read more