The library documents itself · NLP & linguistics

stzDLM

26 own methods · 2 forms written with extensions, listed under their method

the source: base/learning/stzDLM.ring(opens in a new tab) · All classes · The area reference: NLP & linguistics

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

Examples: 0 come from the library's tests; 1 were composed from each method's signature and run for this page, on a fresh object, for example o1 = new stzDLM("restaurant"). 25 methods have none yet. Examples run on 2026-10-04 07:31 inside the library at commit 0e72e2e2c.

Method

Explanation, as the source carries it

init

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

SetDomain

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

Domain

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

composed example

o1 = new stzDLM("restaurant")
? o1.Domain()
#--> restaurant

composed from the signature, run for this page

no extension

ForgeFrom

forgingno extension

LearnEntity

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

LearnRelation

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

LearnFact

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

LearnLaw

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

Lexicon

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

Ask

the language surface ------------------------------------------------- deterministic domain answering: a graph lookup that SPEAKS -- and REFUSES outside its domain (LAW 3)QCSXTZZZIBWU

Complete

domain-valid continuations (the constrained-grammar spirit): an entity completes with its relations; a relation with its objects; a fragment with matching entitiesno extension

GenerateCorpus

every fact rendered as a sentence -- RUNG 2's teacher-free training corpus (correct by construction: these ARE the facts)no extension

Tokenizer

R4 step 8 -- the word-level DOMAIN TOKENIZER: a closed domain earns a closed vocabulary (id 1 = <unk>, then entities, types, relations, and every corpus word). The neural rung trains over these ids; the deterministic rung already speaks them.no extension

Tokenize

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

Why

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

TrainNeuralRung

RUNG 2: the NEURAL domain LM ----------------------------------------- A neural BIGRAM language model: one-hot(current token) -> softmax over the domain vocabulary -> next token. Trained ONLY on GenerateCorpus() (sentences correct-by-construction), no remote teacher -- the whole point of the Foundry: a project's own knowledge becomes a model that GENERATES domain text. Rung 1 (deterministic Ask/Complete) still owns truth; this rung learns the corpus's word-transition distribution. The softmax + categorical-cross-entropy floor makes it possible.no extension

IsNeuralTrained

(no comment in the source: the name reads as a sentence) is neural trainedno extension

NeuralWhy

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

NextToken

the token (string) the neural rung predicts to follow a wordno extension

NextTokenConfidence

the neural rung's confidence in its predicted next token (a real probability -- softmax normalizes over the whole vocabulary)no extension

NeuralGenerate

greedy generation: a seed word + nSteps neural next-tokensno extension

ExportNeuralGguf

export the trained neural rung as a real .gguf (the Foundry artifact)no extension

AddGolden

goldensQCSXTZZZIBWU

RunGoldens

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

LoadFrom

persistence (*.zdlm -- SELF-CONTAINED) --------------------------------- (extension ruling 2026-07-14: Softanza formats are z + a short abbreviation -- .zdlm, .zknw, .zcnv, .zrlz) LOAD the bundle back INTO this DLM -- the mirror of Save(), and the object's OWN act (named LoadFrom: 'Load' collides with Haro's keyword).no extension

Save

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