fix(control_plane): serialize wake prompt after barge-in

components/control_plane/src/multimodal/control_plane_multimodal_local_wake.c

@@ -43,6 +43,8 @@
 
 #define CONTROL_PLANE_LOCAL_WAKE_QUEUE_LEN 4U
 #define CONTROL_PLANE_LOCAL_WAKE_POLL_MS 100U
+#define CONTROL_PLANE_LOCAL_WAKE_BARGE_IN_WAIT_MS 2000U
+#define CONTROL_PLANE_LOCAL_WAKE_BARGE_IN_POLL_MS 20U
 
 #if CONFIG_TQ_LOCAL_WAKE_TASK_CORE < 0
 #define CONTROL_PLANE_LOCAL_WAKE_TASK_CORE tskNO_AFFINITY
@@ -151,6 +153,64 @@ static esp_err_t local_wake_start_listening(void)
     return ESP_OK;
 }
 
+#if CONFIG_TQ_LOCAL_WAKE_RESPONSE_PROMPT_ENABLE
+static bool local_wake_prepare_prompt_playback(void)
+{
+    control_plane_multimodal_runtime_state_t state = {0};
+    esp_err_t state_err = control_plane_multimodal_get_runtime_state(&state);
+    if (state_err != ESP_OK) {
+        ESP_LOGW(TAG,
+                 "Skip local wake prompt: runtime state unavailable: %s",
+                 esp_err_to_name(state_err));
+        return false;
+    }
+
+    if (!state.remote_responding && !state.request_to_speak_pending) {
+        return true;
+    }
+
+    esp_err_t request_err = control_plane_multimodal_send_request_to_speak();
+    if (request_err != ESP_OK && request_err != ESP_ERR_INVALID_STATE) {
+        ESP_LOGW(TAG,
+                 "Skip local wake prompt: barge-in request failed: %s",
+                 esp_err_to_name(request_err));
+        return false;
+    }
+
+    TickType_t started = xTaskGetTickCount();
+    TickType_t timeout_ticks = pdMS_TO_TICKS(
+            CONTROL_PLANE_LOCAL_WAKE_BARGE_IN_WAIT_MS);
+    if (timeout_ticks == 0U) {
+        timeout_ticks = 1U;
+    }
+
+    while ((xTaskGetTickCount() - started) < timeout_ticks) {
+        local_wake_state_t wake_state = local_wake_snapshot();
+        if (wake_state.stop_requested) {
+            return false;
+        }
+
+        state_err = control_plane_multimodal_get_runtime_state(&state);
+        if (state_err == ESP_OK) {
+            if (!state.connected || !state.has_active_task) {
+                return false;
+            }
+            if (!state.remote_responding &&
+                !state.request_to_speak_pending &&
+                state.server_listening) {
+                return true;
+            }
+        }
+        vTaskDelay(pdMS_TO_TICKS(CONTROL_PLANE_LOCAL_WAKE_BARGE_IN_POLL_MS));
+    }
+
+    ESP_LOGW(TAG,
+             "Skip local wake prompt: barge-in confirmation timed out after %ums",
+             (unsigned)CONTROL_PLANE_LOCAL_WAKE_BARGE_IN_WAIT_MS);
+    return false;
+}
+#endif
+
 static void local_wake_handle_detection(
         const domain_local_command_event_t *event)
 {
@@ -178,7 +238,9 @@ static void local_wake_handle_detection(
              (unsigned)event->detection_generation);
 
 #if CONFIG_TQ_LOCAL_WAKE_RESPONSE_PROMPT_ENABLE
-    if (CONFIG_TQ_LOCAL_WAKE_RESPONSE_PROMPT_FILE[0] != '\0') {
+    bool prompt_playback_ready = local_wake_prepare_prompt_playback();
+    if (prompt_playback_ready &&
+        CONFIG_TQ_LOCAL_WAKE_RESPONSE_PROMPT_FILE[0] != '\0') {
         esp_err_t prompt_err = domain_runtime_behavior_play_prompt_file(
                 CONFIG_TQ_LOCAL_WAKE_RESPONSE_PROMPT_FILE);
         if (prompt_err != ESP_OK) {

components/control_plane/src/multimodal/control_plane_multimodal_session_service.c

@@ -1336,6 +1336,7 @@ static void process_text_json(const char *json_text)
                 notify_event(CONTROL_PLANE_MULTIMODAL_EVENT_SERVER_NOT_LISTENING);
             }
         } else if (strcmp(output_event->valuestring, "RequestAccepted") == 0) {
+            stop_local_downstream_playback();
             if (lock_state()) {
                 s_server_listening = true;
                 s_remote_responding = false;
@@ -1346,7 +1347,6 @@ static void process_text_json(const char *json_text)
                 s_drop_downstream_audio = false;
                 unlock_state();
             }
-            stop_local_downstream_playback();
             ESP_LOGI(TAG, "PTT barge-in request accepted");
             notify_event(CONTROL_PLANE_MULTIMODAL_EVENT_SERVER_LISTENING);
         } else if (strcmp(output_event->valuestring, "RespondingStarted") == 0) {