feat[api]: separate TTS language validation per engine (#2581)

Validate chatterbox and supertonic languages with distinct enums instead
of a shared list. Expose all 18 chatterbox multilingual languages and
restrict supertonic to its actual subset (en/es/fr/pt/ko).

Co-authored-by: ishanvohra2 <your-github-email>

Author: ishanvohra2

packages/sdk/schemas/text-to-speech.ts

@@ -1,26 +1,55 @@
 import { z } from "zod";
 import { modelSrcInputSchema } from "./model-src-utils";
 
-// TTS supported languages based on available models
-export const TTS_LANGUAGES = [
+// Chatterbox multilingual supported languages (18). The engines support
+// different language sets, so the language enum is validated per engine.
+export const TTS_CHATTERBOX_LANGUAGES = [
   "en", // English
   "es", // Spanish
+  "fr", // French
   "de", // German
   "it", // Italian
+  "pt", // Portuguese
+  "nl", // Dutch
+  "pl", // Polish
+  "tr", // Turkish
+  "sv", // Swedish
+  "da", // Danish
+  "fi", // Finnish
+  "no", // Norwegian
+  "el", // Greek
+  "ms", // Malay
+  "sw", // Swahili
+  "ar", // Arabic
+  "ko", // Korean
 ] as const;
 
-const ttsLanguageSchema = z.enum(TTS_LANGUAGES);
+// Supertonic supported languages (subset of the Chatterbox set).
+export const TTS_SUPERTONIC_LANGUAGES = [
+  "en", // English
+  "es", // Spanish
+  "fr", // French
+  "pt", // Portuguese
+  "ko", // Korean
+] as const;
+
+// Union of all TTS-supported languages across engines. Kept for backwards
+// compatibility; prefer the engine-specific lists when validating a config.
+export const TTS_LANGUAGES = [...TTS_CHATTERBOX_LANGUAGES] as const;
+
+const ttsChatterboxLanguageSchema = z.enum(TTS_CHATTERBOX_LANGUAGES);
+const ttsSupertonicLanguageSchema = z.enum(TTS_SUPERTONIC_LANGUAGES);
 
 export const ttsChatterboxRuntimeConfigSchema = z.object({
   ttsEngine: z.literal("chatterbox"),
-  language: ttsLanguageSchema,
+  language: ttsChatterboxLanguageSchema,
   voice: z.string().optional(),
   useGPU: z.boolean().optional(),
 });
 
 export const ttsSupertonicRuntimeConfigSchema = z.object({
   ttsEngine: z.literal("supertonic"),
-  language: ttsLanguageSchema,
+  language: ttsSupertonicLanguageSchema,
   voice: z.string().optional(),
   ttsSpeed: z.number().optional(),
   ttsNumInferenceSteps: z.number().optional(),
@@ -141,6 +170,8 @@ export const textToSpeechStreamResponseSchema = z.object({
 });
 
 export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
+export type TtsChatterboxLanguage = (typeof TTS_CHATTERBOX_LANGUAGES)[number];
+export type TtsSupertonicLanguage = (typeof TTS_SUPERTONIC_LANGUAGES)[number];
 export type TtsChatterboxLoadConfig = z.infer<typeof ttsChatterboxLoadConfigSchema>;
 export type TtsSupertonicLoadConfig = z.infer<typeof ttsSupertonicLoadConfigSchema>;
 export type TtsLoadConfig = z.infer<typeof ttsLoadConfigSchema>;

packages/sdk/test/unit/tts-schemas.test.ts

@@ -4,7 +4,10 @@ import {
   ttsResponseSchema,
   textToSpeechStreamResponseSchema,
   ttsConfigSchema,
+  ttsChatterboxRuntimeConfigSchema,
   ttsSupertonicRuntimeConfigSchema,
+  TTS_CHATTERBOX_LANGUAGES,
+  TTS_SUPERTONIC_LANGUAGES,
   LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS,
 } from "@/schemas/text-to-speech";
 
@@ -26,6 +29,54 @@ test("ttsConfigSchema: accepts GGML supertonic load config", (t) => {
   t.is(r.success, true);
 });
 
+test("TTS_CHATTERBOX_LANGUAGES: exposes all 18 supported languages", (t) => {
+  t.is(TTS_CHATTERBOX_LANGUAGES.length, 18);
+  const expected = [
+    "en", "es", "fr", "de", "it", "pt", "nl", "pl", "tr",
+    "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko",
+  ];
+  t.alike([...TTS_CHATTERBOX_LANGUAGES], expected);
+});
+
+test("ttsChatterboxRuntimeConfigSchema: accepts all 18 chatterbox languages", (t) => {
+  for (const language of TTS_CHATTERBOX_LANGUAGES) {
+    const r = ttsChatterboxRuntimeConfigSchema.safeParse({
+      ttsEngine: "chatterbox",
+      language,
+    });
+    t.is(r.success, true, `chatterbox should accept ${language}`);
+  }
+});
+
+test("ttsSupertonicRuntimeConfigSchema: only accepts its language subset", (t) => {
+  t.alike([...TTS_SUPERTONIC_LANGUAGES], ["en", "es", "fr", "pt", "ko"]);
+  for (const language of TTS_SUPERTONIC_LANGUAGES) {
+    const r = ttsSupertonicRuntimeConfigSchema.safeParse({
+      ttsEngine: "supertonic",
+      language,
+    });
+    t.is(r.success, true, `supertonic should accept ${language}`);
+  }
+});
+
+test("ttsSupertonicRuntimeConfigSchema: rejects chatterbox-only languages", (t) => {
+  // 'de' is supported by chatterbox but not supertonic.
+  const r = ttsSupertonicRuntimeConfigSchema.safeParse({
+    ttsEngine: "supertonic",
+    language: "de",
+  });
+  t.is(r.success, false, "supertonic must reject 'de'");
+});
+
+test("ttsConfigSchema: accepts a chatterbox-only language for chatterbox", (t) => {
+  const r = ttsConfigSchema.safeParse({
+    ttsEngine: "chatterbox",
+    language: "tr",
+    s3genModelSrc: "s3:///example/s3gen.gguf",
+  });
+  t.is(r.success, true, "chatterbox load config accepts 'tr'");
+});
+
 test("ttsSupertonicRuntimeConfigSchema: strips removed ttsSupertonicMultilingual", (t) => {
   const r = ttsSupertonicRuntimeConfigSchema.safeParse({
     ttsEngine: "supertonic",