The library documents itself · Neural tier

stzSemanticIndex

13 own methods · 1 forms written with extensions, listed under their method

the source: base/neural/stzSemanticIndex.ring(opens in a new tab) · All classes · The area reference: Neural tier

Under each method, its extensions (Q CS XT Z ZZ IB W U): bold and ticked, it exists; struck, it does not. The extensions and what they mean

No method of this class is run on this page: the class reaches files, the network, processes, the clock, chance or sound, and an example run here would only be an example of its effects. Each method carries the explanation the library gives of itself.

Method

Explanation, as the source carries it

init

(no comment in the source: the name reads as a sentence) initno extension

AddText

Embeds at ADD time (the model must be loaded), so a bad text fails loudly here and not in the middle of a search.no extension

AddEmbedded

Index a text with an embedding the CALLER already has -- another model, a cache, a wire -- so the index never insists on embedding it again. The vector must be L2-normalised (the score formula assumes it) and of the index's dimension.no extension

AddTexts

feeding the indexno extension

Count

feeding the indexno extension

Texts

feeding the indexno extension

EmbeddingDim

feeding the indexno extension

Vectors

The stored embeddings, 1-based rows -- so a caller (or a guard) can recompute a cosine independently of the search path.no extension

UsesGpu

Is the corpus resident on the GPU right now? Silent seams still owe an honest answer to whoever asks.no extension

SearchByVectorXT

The same search from a query embedding the caller already holds (the companion of AddEmbedded). ONE search path serves both doors: the route -- CPU resident dataset or GPU resident corpus -- is decided here and nowhere else.no extension

Same as SearchXT: Top-n semantically closest texts: [ [cText, nCosine, nPosition], ... ] best first. nCosine is in [-1, 1]; 1 means "the same meaning" and a query that IS one of the indexed texts scores ~1 against itself.QCSXTZZZIBWU

Closest

The single best text (or "" on an empty index).no extension

Haro has no destructors (the residency lesson): free the resident dataset explicitly when done with the index.no extension