Function useSpeechToText

apps/sim/hooks/use-speech-to-text.ts:41–403

({
  onTranscript,
  onUsageLimitExceeded,
  language,
  workspaceId,
}: UseSpeechToTextProps)

}

export function useSpeechToText({
  onTranscript,
  onUsageLimitExceeded,
  language,
  workspaceId,
}: UseSpeechToTextProps): UseSpeechToTextReturn {
  const [isListening, setIsListening] = useState(false)
  const [isSupported, setIsSupported] = useState(false)
  const [permissionState, setPermissionState] = useState<PermissionState>('prompt')

  const onTranscriptRef = useRef(onTranscript)
  const onUsageLimitExceededRef = useRef(onUsageLimitExceeded)
  const languageRef = useRef(language)
  const workspaceIdRef = useRef(workspaceId)
  const mountedRef = useRef(true)
  const startingRef = useRef(false)

  const wsRef = useRef<WebSocket | null>(null)
  const streamRef = useRef<MediaStream | null>(null)
  const audioContextRef = useRef<AudioContext | null>(null)
  const processorRef = useRef<ScriptProcessorNode | null>(null)

  const pcmBufferRef = useRef<Float32Array[]>([])
  const sendIntervalRef = useRef<ReturnType<typeof setInterval> | null>(null)
  const sessionTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null)
  const stopStreamingRef = useRef<() => void>(() => {})
  const isFirstChunkRef = useRef(true)
  const committedTextRef = useRef('')

  onTranscriptRef.current = onTranscript
  onUsageLimitExceededRef.current = onUsageLimitExceeded
  languageRef.current = language
  workspaceIdRef.current = workspaceId

  useEffect(() => {
    const browserOk =
      typeof window !== 'undefined' &&
      typeof AudioContext !== 'undefined' &&
      typeof WebSocket !== 'undefined' &&
      typeof navigator?.mediaDevices?.getUserMedia === 'function'

    if (!browserOk) {
      setIsSupported(false)
      return
    }

    requestJson(getVoiceSettingsContract, {})
      .then((data) => {
        if (mountedRef.current) setIsSupported(data.sttAvailable === true)
      })
      .catch(() => {
        if (mountedRef.current) setIsSupported(false)
      })
  }, [])

  const flushAudioBuffer = useCallback(() => {
    const ws = wsRef.current
    if (!ws || ws.readyState !== WebSocket.OPEN) return