API

dumling exposes separate entrypoints for workflow operations, public types, and schemas.

Entrypoints

Runtime API

The root runtime entrypoint exports:

import {
	dumling,
	getLanguageApi,
	inspectId,
	supportedLanguages,
} from "dumling";

The language-bound API is the primary workflow surface:

const de = dumling.de;
const en = getLanguageApi("en");

The implemented language namespaces are:

• dumling.de
• dumling.en
• dumling.he

create

create constructors set namespace-implied fields such as language and surfaceKind.

const lemma = dumling.en.create.lemma({
	canonicalLemma: "run",
	lemmaKind: "Lexeme",
	lemmaSubKind: "VERB",
	inherentFeatures: {},
	meaningInEmojis: "🏃",
});

const surface = dumling.en.create.surface.inflection({
	lemma,
	normalizedFullSurface: "ran",
	inflectionalFeatures: {
		tense: "Past",
		verbForm: "Fin",
	},
});

const selection = dumling.en.create.selection({
	surface,
	spelledSelection: "ran",
});

convert

convert derives common linked DTOs.

const surface = dumling.de.convert.lemma.toSurface(lemma);

const selection = dumling.de.convert.surface.toSelection(surface, {
	spelledSelection: "See",
	selectionFeatures: {
		coverage: "Partial",
	},
});

Defaults for generated selections are:

• spelledSelection: surface.normalizedFullSurface
• omitted selectionFeatures.orthography means standard orthography
• omitted selectionFeatures.coverage means full coverage
• omitted selectionFeatures.spelling means canonical spelling relation

extract

extract retrieves the canonical nested entity from any hydrated value:

const lemmaFromLemma = dumling.de.extract.lemma(lemma);
const lemmaFromSurface = dumling.de.extract.lemma(surface);
const lemmaFromSelection = dumling.de.extract.lemma(selection);

parse

parse validates unknown input against the language runtime schemas.

const result = dumling.he.parse.selection(input);

if (result.success) {
	result.data.surface.lemma;
} else {
	result.error.code;
	result.error.message;
}

describe

describe.as returns descriptors for routing and indexing:

dumling.en.describe.as.lemma(lemma);
dumling.en.describe.as.surface(surface);
dumling.en.describe.as.selection(selection);

id

Language-bound ID helpers encode and decode hydrated DTOs:

const id = dumling.en.id.encode(selection);
const decoded = dumling.en.id.decodeAs("Selection", id);

decode returns both the entity kind and the decoded data. decodeAs checks that the ID contains the expected entity kind.

schemas

The schema entrypoint exports concrete and abstract schema registries:

import { abstractSchemas, getSchemaTreeFor, schemasFor } from "dumling/schema";

Concrete leaf schema getters return Zod schemas:

schemasFor.de.entity.Lemma.Lexeme.NOUN();
schemasFor.en.entity.Surface.Inflection.Lexeme.VERB();
schemasFor.he.entity.Selection.Inflection.Lexeme.NOUN();

Use getSchemaTreeFor(language) when the language is only known at runtime.