fix: defer mic permission request until audioInput is enabled

  Previously, refreshInputDevices and watchAudioInputDevices were called
  on mount regardless of audioInput value, triggering mic permission
  prompts. Now these only run when audioInput is true, allowing apps to
  start in text-only mode without permission prompts.

  Also made setAudioInput and setAudioOutput async to properly await
  the underlying SDK methods.

Author: jackbridger

.gitignore

@@ -15,4 +15,6 @@ node_modules/
 
 # OS files
 .DS_Store
-Thumbs.db 
+Thumbs.db 
+
+.npm-cache

src/index.tsx

@@ -148,17 +148,21 @@ const useLayercodeAgent = (
   }, [preferredInputDeviceId]);
 
   useEffect(() => {
-    if (typeof window === 'undefined') {
+    // Only refresh input devices if audioInput is enabled
+    // This prevents requesting microphone permissions when audio is disabled
+    if (typeof window === 'undefined' || !audioInput) {
       return;
     }
 
     refreshInputDevices().catch((error) => {
       console.warn('Layercode: failed to load microphone devices', error);
     });
-  }, [refreshInputDevices]);
+  }, [refreshInputDevices, audioInput]);
 
   useEffect(() => {
-    if (typeof window === 'undefined' || typeof navigator === 'undefined') {
+    // Only watch for device changes if audioInput is enabled
+    // This prevents requesting microphone permissions when audio is disabled
+    if (typeof window === 'undefined' || typeof navigator === 'undefined' || !audioInput) {
       return;
     }
 
@@ -176,11 +180,11 @@ const useLayercodeAgent = (
     return () => {
       unsubscribe?.();
     };
-  }, []);
+  }, [audioInput]);
 
   const createClient = useCallback(
     (initialConversationId: string | null) => {
-      console.log('Creating LayercodeClient instance');
+      console.log('Creating LayercodeClient instance with audioInput:', audioInput, 'audioOutput:', audioOutput);
       const client = new LayercodeClient({
         agentId,
         conversationId: initialConversationId,
@@ -310,19 +314,19 @@ const useLayercodeAgent = (
   }, []);
 
   const setAudioInput = useCallback(
-    (state: React.SetStateAction<boolean>) => {
+    async (state: React.SetStateAction<boolean>) => {
       _setAudioInput(state);
       const next = typeof state === 'function' ? (state as (prev: boolean) => boolean)(audioInput) : state;
-      clientRef.current?.setAudioInput(next);
+      await clientRef.current?.setAudioInput(next);
     },
     [_setAudioInput, clientRef, audioInput]
   );
 
   const setAudioOutput = useCallback(
-    (state: React.SetStateAction<boolean>) => {
+    async (state: React.SetStateAction<boolean>) => {
       _setAudioOutput(state);
       const next = typeof state === 'function' ? (state as (prev: boolean) => boolean)(audioOutput) : state;
-      clientRef.current?.setAudioOutput(next);
+      await clientRef.current?.setAudioOutput(next);
     },
     [_setAudioOutput, clientRef, audioOutput]
   );