PronunciationOptions
Options for createPronunciationGuardrail.
Defined in: src/guardrails/pronunciation.ts:18
Options for createPronunciationGuardrail.
Properties
| Property | Type | Default value | Description | Defined in |
|---|---|---|---|---|
caseSensitive? | boolean | false | Require the term to be matched case-sensitively. Remarks Off by default so sql, Sql, and SQL are all fixed. Turn it on when a term collides with an ordinary word in another casing — e.g. rewriting IT but not it. | src/guardrails/pronunciation.ts:39 |
name? | string | 'pronunciation' | Guardrail name reported in events. | src/guardrails/pronunciation.ts:58 |
replacements | Readonly<Record<string, string>> | undefined | Terms to rewrite, mapped to what the synthesizer should receive. Example { SQL: 'sequel', kubectl: 'kube control', 'Ln.': 'Lane' } | src/guardrails/pronunciation.ts:27 |
stages? | readonly GuardrailStage[] | both stages | Stages at which to run. | src/guardrails/pronunciation.ts:65 |
wholeWord? | boolean | true | Require the term to stand alone rather than matching inside a longer word. Remarks On by default, so replacing AI leaves SAID and chain untouched. Terms that begin or end with a non-word character (.NET, C++) get the boundary applied only on the side where it is meaningful. | src/guardrails/pronunciation.ts:51 |