fix: disable SSML when not support

Author: luoling8192

packages/stage-ui/src/stores/modules/speech.test.ts

@@ -126,6 +126,39 @@ describe('speech store helpers', () => {
     expect(request.input).toContain('volume="-5%"')
   })
 
+  /**
+   * @example
+   * speechStore.resolveVoicePackSpeechInput({ text, voice, forceSSML: true, supportsSSML: false, supportsAdapterProsody: true })
+   */
+  it('keeps official adapter-backed speech input as plain text when global SSML is enabled', () => {
+    const speechStore = useSpeechStore()
+    const voice = {
+      id: 'voice-1',
+      name: 'Voice 1',
+      provider: OFFICIAL_SPEECH_PROVIDER_ID,
+      languages: [{ code: 'en-US', title: 'English' }],
+      gender: 'neutral',
+    }
+
+    // ROOT CAUSE:
+    //
+    // Auto TTS can enable global SSML before the server routes the official
+    // speech provider to DashScope CosyVoice. DashScope rejects `<speak>...`
+    // payloads with `SSML text is not supported at the moment!`, so providers
+    // that apply prosody through adapter options must keep the text field plain.
+    const request = speechStore.resolveVoicePackSpeechInput({
+      text: 'hello',
+      voice,
+      providerConfig: { pitch: 0 },
+      forceSSML: true,
+      supportsSSML: false,
+      supportsAdapterProsody: true,
+    })
+
+    expect(request.input).toBe('hello')
+    expect(request.input).not.toContain('<speak')
+  })
+
   /**
    * @example
    * speechStore.resolveVoicePackSpeechInput({ text, voice, params: { pitch: '+20%' }, supportsAdapterProsody: true })

packages/stage-ui/src/stores/modules/speech.ts

@@ -473,10 +473,11 @@ export const useSpeechStore = defineStore('speech', () => {
 
   function resolveVoicePackSpeechInput(options: VoicePackSpeechInputOptions): VoicePackSpeechInput {
     const providerConfig = { ...options.providerConfig }
+    const canUseSSML = options.supportsSSML === true
 
     if (!options.params) {
       return {
-        input: options.forceSSML
+        input: options.forceSSML === true && canUseSSML
           ? generateSSML(options.text, options.voice, providerConfig)
           : options.text,
         providerConfig,
@@ -493,6 +494,8 @@ export const useSpeechStore = defineStore('speech', () => {
     if (speed != null)
       providerConfig.speed = speed
 
+    const shouldUseSSML = canUseSSML && (options.forceSSML === true || (needsProsody && !options.supportsAdapterProsody))
+
     if (options.voicePack) {
       providerConfig.extraBody = {
         ...(providerConfig.extraBody as Record<string, unknown> | undefined),
@@ -511,11 +514,11 @@ export const useSpeechStore = defineStore('speech', () => {
         voice_pack: { pitch, volume },
       }
     }
-    else if (needsProsody && !options.forceSSML && !options.supportsSSML) {
+    else if (needsProsody && !options.supportsAdapterProsody && !shouldUseSSML) {
       throw new Error('Voice Pack pitch and volume parameters require an SSML-capable speech provider.')
     }
 
-    if (!options.forceSSML && (!needsProsody || options.supportsAdapterProsody)) {
+    if (!shouldUseSSML && (!needsProsody || options.supportsAdapterProsody)) {
       return {
         input: options.text,
         providerConfig,