better handling of failed speech token acquisition

workbench-app/src/components/Conversations/SpeechButton.tsx

@@ -20,7 +20,7 @@ interface SpeechButtonProps {
 export const SpeechButton: React.FC<SpeechButtonProps> = (props) => {
     const { disabled, onListeningChange, onSpeechRecognizing, onSpeechRecognized } = props;
     const [recognizer, setRecognizer] = React.useState<speechSdk.SpeechRecognizer>();
-    const [isFetching, setIsFetching] = React.useState(false);
+    const [isInitialized, setIsInitialized] = React.useState(false);
     const [isListening, setIsListening] = React.useState(false);
     const [lastSpeechResultTimestamp, setLastSpeechResultTimestamp] = React.useState(0);
 
@@ -115,15 +115,17 @@ export const SpeechButton: React.FC<SpeechButtonProps> = (props) => {
     }, [getAzureSpeechTokenAsync, onSpeechRecognized, onSpeechRecognizing]);
 
     React.useEffect(() => {
-        // If the recognizer is already available or we are fetching it, do nothing
-        if (recognizer || isFetching) return;
+        // If the recognizer is already initialized, return
+        if (isInitialized) return;
 
-        // Indicate that we are fetching the recognizer to prevent multiple fetches
-        setIsFetching(true);
+        // Set the recognizer as initialized
+        setIsInitialized(true);
 
-        // Fetch the recognizer, then indicate that we are no longer fetching even if the fetch fails
-        getRecognizer().finally(() => setIsFetching(false));
-    }, [getRecognizer, isFetching, recognizer]);
+        (async () => {
+            // Fetch the recognizer
+            await getRecognizer();
+        })();
+    }, [getRecognizer, isInitialized, recognizer]);
 
     React.useEffect(() => {
         onListeningChange(isListening);

workbench-service/semantic_workbench_service/azure_speech.py

@@ -1,14 +1,22 @@
+import logging
+
 from azure.identity import DefaultAzureCredential
 
 from . import settings
 
+logger = logging.getLogger(__name__)
+
 
 def get_token() -> dict[str, str]:
     if settings.azure_speech.resource_id == "" or settings.azure_speech.region == "":
         return {}
 
     credential = DefaultAzureCredential()
-    token = credential.get_token("https://cognitiveservices.azure.com/.default").token
+    try:
+        token = credential.get_token("https://cognitiveservices.azure.com/.default").token
+    except Exception as e:
+        logger.error(f"Failed to get token: {e}")
+        return {}
 
     return {
         "token": f"aad#{settings.azure_speech.resource_id}#{token}",