WS刚连接成功可以进行 对话，超过30sWS断开后，语音上行停止，需要修复

components/control_plane/src/multimodal/control_plane_multimodal_runtime.c

@@ -46,6 +46,7 @@
 #define CONTROL_PLANE_DOWNSTREAM_PLAYBACK_GUARD_MS 220
 #define CONTROL_PLANE_IDLE_TRANSPORT_SHUTDOWN_MS 30000
 #define CONTROL_PLANE_WATCHDOG_TASK_STACK_SIZE 6144
+#define CONTROL_PLANE_SERVER_SUCCESS_PROMPT_FILE "/littlefs/audio_prompts/server_connect_success.opus"
 
 #ifndef CONFIG_TQ_NETWORK_TASK_CORE
 #define CONFIG_TQ_NETWORK_TASK_CORE 0
@@ -295,6 +296,13 @@ static esp_err_t control_plane_start_uplink_if_ready(void)
 
 esp_err_t control_plane_multimodal_handle_ptt_press_step(void)
 {
+    esp_err_t capture_err = control_plane_multimodal_start_capture();
+    if (capture_err != ESP_OK &&
+        capture_err != ESP_ERR_INVALID_STATE &&
+        capture_err != ESP_ERR_NOT_SUPPORTED) {
+        ESP_LOGW(TAG, "PTT capture start failed: %s", esp_err_to_name(capture_err));
+    }
+
     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) {
@@ -580,6 +588,7 @@ static void control_plane_multimodal_event_bridge(control_plane_multimodal_event
 
     switch (event) {
         case CONTROL_PLANE_MULTIMODAL_EVENT_TRANSPORT_CONNECTED:
+            (void)domain_runtime_behavior_play_prompt_file(CONTROL_PLANE_SERVER_SUCCESS_PROMPT_FILE);
             if (CONTROL_PLANE_MULTIMODAL_AUTO_START_FLAG) {
                 schedule_autostart_if_enabled();
             }

components/control_plane/src/multimodal/control_plane_multimodal_session_service.c

@@ -83,6 +83,8 @@
 #define CONTROL_PLANE_LOCAL_RESPONDING_EVENT_TASK_PRIORITY 5U
 #define CONTROL_PLANE_LOCAL_RESPONDING_EVENT_QUEUE_LEN 4U
 #define CONTROL_PLANE_MULTIMODAL_BINARY_SUMMARY_INTERVAL 200U
+#define CONTROL_PLANE_UPLINK_READY_WAIT_MS 6000U
+#define CONTROL_PLANE_UPLINK_READY_POLL_MS 20U
 
 #ifndef CONFIG_TQ_LOCAL_TASK_CORE
 #define CONFIG_TQ_LOCAL_TASK_CORE 1
@@ -1322,7 +1324,12 @@ static void transport_event_bridge(const domain_multimodal_transport_event_t *ev
             break;
 
         case DOMAIN_MULTIMODAL_TRANSPORT_EVENT_DISCONNECTED:
-            reset_session_state();
+            if (lock_state()) {
+                s_started = false;
+                memset(&s_auth, 0, sizeof(s_auth));
+                reset_session_state_locked();
+                unlock_state();
+            }
             notify_event(CONTROL_PLANE_MULTIMODAL_EVENT_TRANSPORT_DISCONNECTED);
             break;
 
@@ -1426,6 +1433,30 @@ static void downstream_playback_event_bridge(
 static esp_err_t control_plane_audio_uplink_send_cb(const uint8_t *data, size_t len, void *ctx)
 {
     (void)ctx;
+    uint32_t waited_ms = 0U;
+    while (waited_ms < CONTROL_PLANE_UPLINK_READY_WAIT_MS) {
+        esp_err_t err = control_plane_multimodal_send_audio_realtime(data, len);
+        if (err != ESP_ERR_INVALID_STATE) {
+            return err;
+        }
+
+        control_plane_multimodal_runtime_state_t state = {0};
+        if (control_plane_multimodal_get_runtime_state(&state) == ESP_OK &&
+            state.connected) {
+            if (!state.has_active_task) {
+                (void)control_plane_multimodal_send_start();
+            } else if (state.server_listening &&
+                       !state.client_speech_active &&
+                       !state.ptt_release_required &&
+                       !state.remote_responding) {
+                (void)control_plane_multimodal_send_speech_start();
+            }
+        }
+
+        vTaskDelay(pdMS_TO_TICKS(CONTROL_PLANE_UPLINK_READY_POLL_MS));
+        waited_ms += CONTROL_PLANE_UPLINK_READY_POLL_MS;
+    }
+
     return control_plane_multimodal_send_audio_realtime(data, len);
 }
 
@@ -1439,7 +1470,7 @@ esp_err_t control_plane_multimodal_start(void)
     if (lock_state()) {
         bool already_started = s_started;
         unlock_state();
-        if (already_started && domain_runtime_multimodal_transport_is_connected()) {
+        if (already_started) {
             return ESP_OK;
         }
     }
@@ -2056,6 +2087,22 @@ static esp_err_t control_plane_multimodal_prepare_audio_runtime(void)
     return domain_runtime_multimodal_audio_init();
 }
 
+esp_err_t control_plane_multimodal_start_capture(void)
+{
+    if (!domain_runtime_multimodal_audio_is_initialized()) {
+        esp_err_t init_err = domain_runtime_multimodal_audio_init();
+        if (init_err != ESP_OK) {
+            return init_err;
+        }
+    }
+
+    if (domain_runtime_multimodal_audio_is_uplink_running()) {
+        return ESP_OK;
+    }
+
+    return domain_runtime_multimodal_audio_start_uplink(control_plane_audio_uplink_send_cb, NULL);
+}
+
 esp_err_t control_plane_multimodal_start_uplink(void)
 {
     esp_err_t ready_err = control_plane_multimodal_prepare_audio_runtime();
@@ -2072,7 +2119,7 @@ esp_err_t control_plane_multimodal_start_uplink(void)
 
 void control_plane_multimodal_stop_uplink(void)
 {
-    esp_err_t err = domain_runtime_multimodal_audio_stop_uplink(2000);
+    esp_err_t err = domain_runtime_multimodal_audio_stop_uplink(CONTROL_PLANE_UPLINK_READY_WAIT_MS);
     if (err != ESP_OK) {
         ESP_LOGW(TAG, "Mic uplink stop ended with %s", esp_err_to_name(err));
     }