Merge branch 'dev' into always_run_clang_tidy_no_cpp_files_platformio_change

.clang-tidy.hash

@@ -1 +1 @@
-07f621354fe1350ba51953c80273cd44a04aa44f15cc30bd7b8fe2a641427b7a
+0c2acbc16bfb7d63571dbe7042f94f683be25e4ca8a0f158a960a94adac4b931

esphome/components/esp32_camera/__init__.py

@@ -119,6 +119,12 @@ ENUM_SPECIAL_EFFECT = {
     "SEPIA": ESP32SpecialEffect.ESP32_SPECIAL_EFFECT_SEPIA,
 }
 
+camera_fb_location_t = cg.global_ns.enum("camera_fb_location_t")
+ENUM_FB_LOCATION = {
+    "PSRAM": cg.global_ns.CAMERA_FB_IN_PSRAM,
+    "DRAM": cg.global_ns.CAMERA_FB_IN_DRAM,
+}
+
 # pin assignment
 CONF_HREF_PIN = "href_pin"
 CONF_PIXEL_CLOCK_PIN = "pixel_clock_pin"
@@ -149,6 +155,7 @@ CONF_MAX_FRAMERATE = "max_framerate"
 CONF_IDLE_FRAMERATE = "idle_framerate"
 # frame buffer
 CONF_FRAME_BUFFER_COUNT = "frame_buffer_count"
+CONF_FRAME_BUFFER_LOCATION = "frame_buffer_location"
 
 # stream trigger
 CONF_ON_STREAM_START = "on_stream_start"
@@ -230,6 +237,9 @@ CONFIG_SCHEMA = cv.All(
                 cv.framerate, cv.Range(min=0, max=1)
             ),
             cv.Optional(CONF_FRAME_BUFFER_COUNT, default=1): cv.int_range(min=1, max=2),
+            cv.Optional(CONF_FRAME_BUFFER_LOCATION, default="PSRAM"): cv.enum(
+                ENUM_FB_LOCATION, upper=True
+            ),
             cv.Optional(CONF_ON_STREAM_START): automation.validate_automation(
                 {
                     cv.GenerateID(CONF_TRIGGER_ID): cv.declare_id(
@@ -301,6 +311,7 @@ SETTERS = {
     CONF_WB_MODE: "set_wb_mode",
     # test pattern
     CONF_TEST_PATTERN: "set_test_pattern",
+    CONF_FRAME_BUFFER_LOCATION: "set_frame_buffer_location",
 }
 
 
@@ -328,6 +339,7 @@ async def to_code(config):
     else:
         cg.add(var.set_idle_update_interval(1000 / config[CONF_IDLE_FRAMERATE]))
     cg.add(var.set_frame_buffer_count(config[CONF_FRAME_BUFFER_COUNT]))
+    cg.add(var.set_frame_buffer_location(config[CONF_FRAME_BUFFER_LOCATION]))
     cg.add(var.set_frame_size(config[CONF_RESOLUTION]))
 
     cg.add_define("USE_CAMERA")

esphome/components/esp32_camera/esp32_camera.cpp

@@ -133,6 +133,7 @@ void ESP32Camera::dump_config() {
   ESP_LOGCONFIG(TAG,
                 "  JPEG Quality: %u\n"
                 "  Framebuffer Count: %u\n"
+                "  Framebuffer Location: %s\n"
                 "  Contrast: %d\n"
                 "  Brightness: %d\n"
                 "  Saturation: %d\n"
@@ -140,8 +141,9 @@ void ESP32Camera::dump_config() {
                 "  Horizontal Mirror: %s\n"
                 "  Special Effect: %u\n"
                 "  White Balance Mode: %u",
-                st.quality, conf.fb_count, st.contrast, st.brightness, st.saturation, ONOFF(st.vflip),
+                st.quality, conf.fb_count, this->config_.fb_location == CAMERA_FB_IN_PSRAM ? "PSRAM" : "DRAM",
-                ONOFF(st.hmirror), st.special_effect, st.wb_mode);
+                st.contrast, st.brightness, st.saturation, ONOFF(st.vflip), ONOFF(st.hmirror), st.special_effect,
+                st.wb_mode);
   // ESP_LOGCONFIG(TAG, "  Auto White Balance: %u", st.awb);
   // ESP_LOGCONFIG(TAG, "  Auto White Balance Gain: %u", st.awb_gain);
   ESP_LOGCONFIG(TAG,
@@ -350,6 +352,9 @@ void ESP32Camera::set_frame_buffer_count(uint8_t fb_count) {
   this->config_.fb_count = fb_count;
   this->set_frame_buffer_mode(fb_count > 1 ? CAMERA_GRAB_LATEST : CAMERA_GRAB_WHEN_EMPTY);
 }
+void ESP32Camera::set_frame_buffer_location(camera_fb_location_t fb_location) {
+  this->config_.fb_location = fb_location;
+}
 
 /* ---------------- public API (specific) ---------------- */
 void ESP32Camera::add_image_callback(std::function<void(std::shared_ptr<camera::CameraImage>)> &&callback) {

esphome/components/esp32_camera/esp32_camera.h

@@ -152,6 +152,7 @@ class ESP32Camera : public camera::Camera {
   /* -- frame buffer */
   void set_frame_buffer_mode(camera_grab_mode_t mode);
   void set_frame_buffer_count(uint8_t fb_count);
+  void set_frame_buffer_location(camera_fb_location_t fb_location);
 
   /* public API (derivated) */
   void setup() override;

esphome/components/speaker/media_player/audio_pipeline.cpp

@@ -200,7 +200,7 @@ AudioPipelineState AudioPipeline::process_state() {
       if ((this->read_task_handle_ != nullptr) || (this->decode_task_handle_ != nullptr)) {
         this->delete_tasks_();
         if (this->hard_stop_) {
-          // Stop command was sent, so immediately end of the playback
+          // Stop command was sent, so immediately end the playback
           this->speaker_->stop();
           this->hard_stop_ = false;
         } else {
@@ -210,13 +210,25 @@ AudioPipelineState AudioPipeline::process_state() {
       }
     }
     this->is_playing_ = false;
-    return AudioPipelineState::STOPPED;
+    if (!this->speaker_->is_running()) {
+      return AudioPipelineState::STOPPED;
+    } else {
+      this->is_finishing_ = true;
+    }
   }
 
   if (this->pause_state_) {
     return AudioPipelineState::PAUSED;
   }
 
+  if (this->is_finishing_) {
+    if (!this->speaker_->is_running()) {
+      this->is_finishing_ = false;
+    } else {
+      return AudioPipelineState::PLAYING;
+    }
+  }
+
   if ((this->read_task_handle_ == nullptr) && (this->decode_task_handle_ == nullptr)) {
     // No tasks are running, so the pipeline is stopped.
     xEventGroupClearBits(this->event_group_, EventGroupBits::PIPELINE_COMMAND_STOP);

esphome/components/speaker/media_player/audio_pipeline.h

@@ -114,6 +114,7 @@ class AudioPipeline {
 
   bool hard_stop_{false};
   bool is_playing_{false};
+  bool is_finishing_{false};
   bool pause_state_{false};
   bool task_stack_in_psram_;
 

esphome/components/voice_assistant/voice_assistant.cpp

@@ -35,6 +35,27 @@ void VoiceAssistant::setup() {
       temp_ring_buffer->write((void *) data.data(), data.size());
     }
   });
+
+#ifdef USE_MEDIA_PLAYER
+  if (this->media_player_ != nullptr) {
+    this->media_player_->add_on_state_callback([this]() {
+      switch (this->media_player_->state) {
+        case media_player::MediaPlayerState::MEDIA_PLAYER_STATE_ANNOUNCING:
+          if (this->media_player_response_state_ == MediaPlayerResponseState::URL_SENT) {
+            // State changed to announcing after receiving the url
+            this->media_player_response_state_ = MediaPlayerResponseState::PLAYING;
+          }
+          break;
+        default:
+          if (this->media_player_response_state_ == MediaPlayerResponseState::PLAYING) {
+            // No longer announcing the TTS response
+            this->media_player_response_state_ = MediaPlayerResponseState::FINISHED;
+          }
+          break;
+      }
+    });
+  }
+#endif
 }
 
 float VoiceAssistant::get_setup_priority() const { return setup_priority::AFTER_CONNECTION; }
@@ -223,6 +244,13 @@ void VoiceAssistant::loop() {
       msg.wake_word_phrase = this->wake_word_;
       this->wake_word_ = "";
 
+      // Reset media player state tracking
+#ifdef USE_MEDIA_PLAYER
+      if (this->media_player_ != nullptr) {
+        this->media_player_response_state_ = MediaPlayerResponseState::IDLE;
+      }
+#endif
+
       if (this->api_client_ == nullptr ||
           !this->api_client_->send_message(msg, api::VoiceAssistantRequest::MESSAGE_TYPE)) {
         ESP_LOGW(TAG, "Could not request start");
@@ -315,17 +343,10 @@ void VoiceAssistant::loop() {
 #endif
 #ifdef USE_MEDIA_PLAYER
       if (this->media_player_ != nullptr) {
-        playing = (this->media_player_->state == media_player::MediaPlayerState::MEDIA_PLAYER_STATE_ANNOUNCING);
+        playing = (this->media_player_response_state_ == MediaPlayerResponseState::PLAYING);
 
-        if (playing && this->media_player_wait_for_announcement_start_) {
+        if (this->media_player_response_state_ == MediaPlayerResponseState::FINISHED) {
-          // Announcement has started playing, wait for it to finish
+          this->media_player_response_state_ = MediaPlayerResponseState::IDLE;
-          this->media_player_wait_for_announcement_start_ = false;
-          this->media_player_wait_for_announcement_end_ = true;
-        }
-
-        if (!playing && this->media_player_wait_for_announcement_end_) {
-          // Announcement has finished playing
-          this->media_player_wait_for_announcement_end_ = false;
           this->cancel_timeout("playing");
           ESP_LOGD(TAG, "Announcement finished playing");
           this->set_state_(State::RESPONSE_FINISHED, State::RESPONSE_FINISHED);
@@ -556,7 +577,7 @@ void VoiceAssistant::request_stop() {
       break;
     case State::AWAITING_RESPONSE:
       this->signal_stop_();
-      // Fallthrough intended to stop a streaming TTS announcement that has potentially started
+      break;
     case State::STREAMING_RESPONSE:
 #ifdef USE_MEDIA_PLAYER
       // Stop any ongoing media player announcement
@@ -566,6 +587,10 @@ void VoiceAssistant::request_stop() {
             .set_announcement(true)
             .perform();
       }
+      if (this->started_streaming_tts_) {
+        // Haven't reached the TTS_END stage, so send the stop signal to HA.
+        this->signal_stop_();
+      }
 #endif
       break;
     case State::RESPONSE_FINISHED:
@@ -649,13 +674,16 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
       if (this->media_player_ != nullptr) {
         for (const auto &arg : msg.data) {
           if ((arg.name == "tts_start_streaming") && (arg.value == "1") && !this->tts_response_url_.empty()) {
+            this->media_player_response_state_ = MediaPlayerResponseState::URL_SENT;
+
             this->media_player_->make_call().set_media_url(this->tts_response_url_).set_announcement(true).perform();
 
-            this->media_player_wait_for_announcement_start_ = true;
-            this->media_player_wait_for_announcement_end_ = false;
             this->started_streaming_tts_ = true;
+            this->start_playback_timeout_();
+
             tts_url_for_trigger = this->tts_response_url_;
             this->tts_response_url_.clear();  // Reset streaming URL
+            this->set_state_(State::STREAMING_RESPONSE, State::STREAMING_RESPONSE);
           }
         }
       }
@@ -714,18 +742,22 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
       this->defer([this, url]() {
 #ifdef USE_MEDIA_PLAYER
         if ((this->media_player_ != nullptr) && (!this->started_streaming_tts_)) {
+          this->media_player_response_state_ = MediaPlayerResponseState::URL_SENT;
+
           this->media_player_->make_call().set_media_url(url).set_announcement(true).perform();
 
-          this->media_player_wait_for_announcement_start_ = true;
-          this->media_player_wait_for_announcement_end_ = false;
-          // Start the playback timeout, as the media player state isn't immediately updated
           this->start_playback_timeout_();
         }
+        this->started_streaming_tts_ = false;  // Helps indicate reaching the TTS_END stage
 #endif
         this->tts_end_trigger_->trigger(url);
       });
       State new_state = this->local_output_ ? State::STREAMING_RESPONSE : State::IDLE;
-      this->set_state_(new_state, new_state);
+      if (new_state != this->state_) {
+        // Don't needlessly change the state. The intent progress stage may have already changed the state to streaming
+        // response.
+        this->set_state_(new_state, new_state);
+      }
       break;
     }
     case api::enums::VOICE_ASSISTANT_RUN_END: {
@@ -876,6 +908,9 @@ void VoiceAssistant::on_announce(const api::VoiceAssistantAnnounceRequest &msg)
 #ifdef USE_MEDIA_PLAYER
   if (this->media_player_ != nullptr) {
     this->tts_start_trigger_->trigger(msg.text);
+
+    this->media_player_response_state_ = MediaPlayerResponseState::URL_SENT;
+
     if (!msg.preannounce_media_id.empty()) {
       this->media_player_->make_call().set_media_url(msg.preannounce_media_id).set_announcement(true).perform();
     }
@@ -887,9 +922,6 @@ void VoiceAssistant::on_announce(const api::VoiceAssistantAnnounceRequest &msg)
         .perform();
     this->continue_conversation_ = msg.start_conversation;
 
-    this->media_player_wait_for_announcement_start_ = true;
-    this->media_player_wait_for_announcement_end_ = false;
-    // Start the playback timeout, as the media player state isn't immediately updated
     this->start_playback_timeout_();
 
     if (this->continuous_) {

esphome/components/voice_assistant/voice_assistant.h

@@ -90,6 +90,15 @@ struct Configuration {
   uint32_t max_active_wake_words;
 };
 
+#ifdef USE_MEDIA_PLAYER
+enum class MediaPlayerResponseState {
+  IDLE,
+  URL_SENT,
+  PLAYING,
+  FINISHED,
+};
+#endif
+
 class VoiceAssistant : public Component {
  public:
   VoiceAssistant();
@@ -272,8 +281,8 @@ class VoiceAssistant : public Component {
   media_player::MediaPlayer *media_player_{nullptr};
   std::string tts_response_url_{""};
   bool started_streaming_tts_{false};
-  bool media_player_wait_for_announcement_start_{false};
+
-  bool media_player_wait_for_announcement_end_{false};
+  MediaPlayerResponseState media_player_response_state_{MediaPlayerResponseState::IDLE};
 #endif
 
   bool local_output_{false};

platformio.ini

@@ -138,7 +138,7 @@ lib_deps =
     WiFi                                 ; wifi,web_server_base,ethernet (Arduino built-in)
     Update                               ; ota,web_server_base (Arduino built-in)
     ${common:arduino.lib_deps}
-    ESP32Async/AsyncTCP@3.4.4            ; async_tcp
+    ESP32Async/AsyncTCP@3.4.5            ; async_tcp
     NetworkClientSecure                  ; http_request,nextion (Arduino built-in)
     HTTPClient                           ; http_request,nextion (Arduino built-in)
     ESPmDNS                              ; mdns (Arduino built-in)

requirements.txt

@@ -12,7 +12,7 @@ platformio==6.1.18  # When updating platformio, also update /docker/Dockerfile
 esptool==4.9.0
 click==8.1.7
 esphome-dashboard==20250514.0
-aioesphomeapi==36.0.1
+aioesphomeapi==37.0.0
 zeroconf==0.147.0
 puremagic==1.30
 ruamel.yaml==0.18.14 # dashboard_import

tests/components/esp32_camera/common.yaml

@@ -22,6 +22,7 @@ esp32_camera:
   power_down_pin: 1
   resolution: 640x480
   jpeg_quality: 10
+  frame_buffer_location: PSRAM
   on_image:
     then:
       - lambda: |-