Merge pull request #6863 from esphome/bump-2024.5.5

2024.5.5

esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp

@@ -19,10 +19,27 @@ void I2SAudioSpeaker::setup() {
   ESP_LOGCONFIG(TAG, "Setting up I2S Audio Speaker...");
 
   this->buffer_queue_ = xQueueCreate(BUFFER_COUNT, sizeof(DataEvent));
-  this->event_queue_ = xQueueCreate(BUFFER_COUNT, sizeof(TaskEvent));
+  if (this->buffer_queue_ == nullptr) {
+    ESP_LOGE(TAG, "Failed to create buffer queue");
+    this->mark_failed();
+    return;
   }
 
-void I2SAudioSpeaker::start() { this->state_ = speaker::STATE_STARTING; }
+  this->event_queue_ = xQueueCreate(BUFFER_COUNT, sizeof(TaskEvent));
+  if (this->event_queue_ == nullptr) {
+    ESP_LOGE(TAG, "Failed to create event queue");
+    this->mark_failed();
+    return;
+  }
+}
+
+void I2SAudioSpeaker::start() {
+  if (this->is_failed()) {
+    ESP_LOGE(TAG, "Cannot start audio, speaker failed to setup");
+    return;
+  }
+  this->state_ = speaker::STATE_STARTING;
+}
 void I2SAudioSpeaker::start_() {
   if (!this->parent_->try_lock()) {
     return;  // Waiting for another i2s component to return lock
@@ -141,6 +158,8 @@ void I2SAudioSpeaker::player_task(void *params) {
 }
 
 void I2SAudioSpeaker::stop() {
+  if (this->is_failed())
+    return;
   if (this->state_ == speaker::STATE_STOPPED)
     return;
   if (this->state_ == speaker::STATE_STARTING) {
@@ -200,6 +219,10 @@ void I2SAudioSpeaker::loop() {
 }
 
 size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length) {
+  if (this->is_failed()) {
+    ESP_LOGE(TAG, "Cannot play audio, speaker failed to setup");
+    return 0;
+  }
   if (this->state_ != speaker::STATE_RUNNING && this->state_ != speaker::STATE_STARTING) {
     this->start();
   }

esphome/components/improv_serial/improv_serial_component.cpp

@@ -109,6 +109,7 @@ void ImprovSerialComponent::write_data_(std::vector<uint8_t> &data) {
 #if defined(USE_ESP32_VARIANT_ESP32C3) || defined(USE_ESP32_VARIANT_ESP32C6) || defined(USE_ESP32_VARIANT_ESP32S3)
     case logger::UART_SELECTION_USB_SERIAL_JTAG:
       usb_serial_jtag_write_bytes((char *) data.data(), data.size(), 20 / portTICK_PERIOD_MS);
+      usb_serial_jtag_ll_txfifo_flush();  // fixes for issue in IDF 4.4.7
       break;
 #endif  // USE_ESP32_VARIANT_ESP32C3 || USE_ESP32_VARIANT_ESP32S3
     default:

esphome/components/improv_serial/improv_serial_component.h

@@ -17,6 +17,7 @@
 #if defined(USE_ESP32_VARIANT_ESP32C3) || defined(USE_ESP32_VARIANT_ESP32C6) || defined(USE_ESP32_VARIANT_ESP32S3) || \
     defined(USE_ESP32_VARIANT_ESP32H2)
 #include <driver/usb_serial_jtag.h>
+#include <hal/usb_serial_jtag_ll.h>
 #endif
 #if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
 #include <esp_private/usb_console.h>

esphome/components/voice_assistant/voice_assistant.cpp

@@ -17,7 +17,7 @@ static const char *const TAG = "voice_assistant";
 
 static const size_t SAMPLE_RATE_HZ = 16000;
 static const size_t INPUT_BUFFER_SIZE = 32 * SAMPLE_RATE_HZ / 1000;  // 32ms * 16kHz / 1000ms
-static const size_t BUFFER_SIZE = 1024 * SAMPLE_RATE_HZ / 1000;
+static const size_t BUFFER_SIZE = 512 * SAMPLE_RATE_HZ / 1000;
 static const size_t SEND_BUFFER_SIZE = INPUT_BUFFER_SIZE * sizeof(int16_t);
 static const size_t RECEIVE_SIZE = 1024;
 static const size_t SPEAKER_BUFFER_SIZE = 16 * RECEIVE_SIZE;
@@ -71,6 +71,12 @@ void VoiceAssistant::setup() {
   ESP_LOGCONFIG(TAG, "Setting up Voice Assistant...");
 
   global_voice_assistant = this;
+}
+
+bool VoiceAssistant::allocate_buffers_() {
+  if (this->send_buffer_ != nullptr) {
+    return true;  // Already allocated
+  }
 
 #ifdef USE_SPEAKER
   if (this->speaker_ != nullptr) {
@@ -78,8 +84,7 @@ void VoiceAssistant::setup() {
     this->speaker_buffer_ = speaker_allocator.allocate(SPEAKER_BUFFER_SIZE);
     if (this->speaker_buffer_ == nullptr) {
       ESP_LOGW(TAG, "Could not allocate speaker buffer");
-      this->mark_failed();
+      return false;
-      return;
     }
   }
 #endif
@@ -88,8 +93,7 @@ void VoiceAssistant::setup() {
   this->input_buffer_ = allocator.allocate(INPUT_BUFFER_SIZE);
   if (this->input_buffer_ == nullptr) {
     ESP_LOGW(TAG, "Could not allocate input buffer");
-    this->mark_failed();
+    return false;
-    return;
   }
 
 #ifdef USE_ESP_ADF
@@ -99,17 +103,71 @@ void VoiceAssistant::setup() {
   this->ring_buffer_ = RingBuffer::create(BUFFER_SIZE * sizeof(int16_t));
   if (this->ring_buffer_ == nullptr) {
     ESP_LOGW(TAG, "Could not allocate ring buffer");
-    this->mark_failed();
+    return false;
-    return;
   }
 
   ExternalRAMAllocator<uint8_t> send_allocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
   this->send_buffer_ = send_allocator.allocate(SEND_BUFFER_SIZE);
   if (send_buffer_ == nullptr) {
     ESP_LOGW(TAG, "Could not allocate send buffer");
-    this->mark_failed();
+    return false;
-    return;
   }
+
+  return true;
+}
+
+void VoiceAssistant::clear_buffers_() {
+  if (this->send_buffer_ != nullptr) {
+    memset(this->send_buffer_, 0, SEND_BUFFER_SIZE);
+  }
+
+  if (this->input_buffer_ != nullptr) {
+    memset(this->input_buffer_, 0, INPUT_BUFFER_SIZE * sizeof(int16_t));
+  }
+
+  if (this->ring_buffer_ != nullptr) {
+    this->ring_buffer_->reset();
+  }
+
+#ifdef USE_SPEAKER
+  if (this->speaker_buffer_ != nullptr) {
+    memset(this->speaker_buffer_, 0, SPEAKER_BUFFER_SIZE);
+
+    this->speaker_buffer_size_ = 0;
+    this->speaker_buffer_index_ = 0;
+    this->speaker_bytes_received_ = 0;
+  }
+#endif
+}
+
+void VoiceAssistant::deallocate_buffers_() {
+  ExternalRAMAllocator<uint8_t> send_deallocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
+  send_deallocator.deallocate(this->send_buffer_, SEND_BUFFER_SIZE);
+  this->send_buffer_ = nullptr;
+
+  if (this->ring_buffer_ != nullptr) {
+    this->ring_buffer_.reset();
+    this->ring_buffer_ = nullptr;
+  }
+
+#ifdef USE_ESP_ADF
+  if (this->vad_instance_ != nullptr) {
+    vad_destroy(this->vad_instance_);
+    this->vad_instance_ = nullptr;
+  }
+#endif
+
+  ExternalRAMAllocator<int16_t> input_deallocator(ExternalRAMAllocator<int16_t>::ALLOW_FAILURE);
+  input_deallocator.deallocate(this->input_buffer_, INPUT_BUFFER_SIZE);
+  this->input_buffer_ = nullptr;
+
+#ifdef USE_SPEAKER
+  if (this->speaker_buffer_ != nullptr) {
+    ExternalRAMAllocator<uint8_t> speaker_deallocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
+    speaker_deallocator.deallocate(this->speaker_buffer_, SPEAKER_BUFFER_SIZE);
+    this->speaker_buffer_ = nullptr;
+  }
+#endif
 }
 
 int VoiceAssistant::read_microphone_() {
@@ -138,14 +196,13 @@ void VoiceAssistant::loop() {
     }
     this->continuous_ = false;
     this->signal_stop_();
+    this->clear_buffers_();
     return;
   }
   switch (this->state_) {
     case State::IDLE: {
       if (this->continuous_ && this->desired_state_ == State::IDLE) {
         this->idle_trigger_->trigger();
-
-        this->ring_buffer_->reset();
 #ifdef USE_ESP_ADF
         if (this->use_wake_word_) {
           this->set_state_(State::START_MICROPHONE, State::WAIT_FOR_VAD);
@@ -161,8 +218,15 @@ void VoiceAssistant::loop() {
     }
     case State::START_MICROPHONE: {
       ESP_LOGD(TAG, "Starting Microphone");
-      memset(this->send_buffer_, 0, SEND_BUFFER_SIZE);
+      if (!this->allocate_buffers_()) {
-      memset(this->input_buffer_, 0, INPUT_BUFFER_SIZE * sizeof(int16_t));
+        this->status_set_error("Failed to allocate buffers");
+        return;
+      }
+      if (this->status_has_error()) {
+        this->status_clear_error();
+      }
+      this->clear_buffers_();
+
       this->mic_->start();
       this->high_freq_.start();
       this->set_state_(State::STARTING_MICROPHONE);
@@ -343,10 +407,9 @@ void VoiceAssistant::loop() {
         this->speaker_->stop();
         this->cancel_timeout("speaker-timeout");
         this->cancel_timeout("playing");
-        this->speaker_buffer_size_ = 0;
+
-        this->speaker_buffer_index_ = 0;
+        this->clear_buffers_();
-        this->speaker_bytes_received_ = 0;
+
-        memset(this->speaker_buffer_, 0, SPEAKER_BUFFER_SIZE);
         this->wait_for_stream_end_ = false;
         this->stream_ended_ = false;
 
@@ -507,7 +570,6 @@ void VoiceAssistant::request_start(bool continuous, bool silence_detection) {
   if (this->state_ == State::IDLE) {
     this->continuous_ = continuous;
     this->silence_detection_ = silence_detection;
-    this->ring_buffer_->reset();
 #ifdef USE_ESP_ADF
     if (this->use_wake_word_) {
       this->set_state_(State::START_MICROPHONE, State::WAIT_FOR_VAD);

esphome/components/voice_assistant/voice_assistant.h

@@ -151,6 +151,10 @@ class VoiceAssistant : public Component {
   void set_wake_word(const std::string &wake_word) { this->wake_word_ = wake_word; }
 
  protected:
+  bool allocate_buffers_();
+  void clear_buffers_();
+  void deallocate_buffers_();
+
   int read_microphone_();
   void set_state_(State state);
   void set_state_(State state, State desired_state);

esphome/components/web_server_base/__init__.py

@@ -37,4 +37,4 @@ async def to_code(config):
             cg.add_library("FS", None)
             cg.add_library("Update", None)
         # https://github.com/esphome/ESPAsyncWebServer/blob/master/library.json
-        cg.add_library("esphome/ESPAsyncWebServer-esphome", "3.2.0")
+        cg.add_library("esphome/ESPAsyncWebServer-esphome", "3.2.2")

esphome/const.py

@@ -1,6 +1,6 @@
 """Constants used by esphome."""
 
-__version__ = "2024.5.2"
+__version__ = "2024.5.5"
 
 ALLOWED_NAME_CHARS = "abcdefghijklmnopqrstuvwxyz0123456789-_"
 VALID_SUBSTITUTIONS_CHARACTERS = (

esphome/core/helpers.cpp

@@ -433,6 +433,10 @@ int8_t step_to_accuracy_decimals(float step) {
   return str.length() - dot_pos - 1;
 }
 
+static const std::string BASE64_CHARS = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+                                        "abcdefghijklmnopqrstuvwxyz"
+                                        "0123456789+/";
+
 static inline bool is_base64(char c) { return (isalnum(c) || (c == '+') || (c == '/')); }
 
 std::string base64_encode(const std::vector<uint8_t> &buf) { return base64_encode(buf.data(), buf.size()); }

esphome/core/helpers.h

@@ -435,10 +435,6 @@ std::string value_accuracy_to_string(float value, int8_t accuracy_decimals);
 /// Derive accuracy in decimals from an increment step.
 int8_t step_to_accuracy_decimals(float step);
 
-static const std::string BASE64_CHARS = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
-                                        "abcdefghijklmnopqrstuvwxyz"
-                                        "0123456789+/";
-
 std::string base64_encode(const uint8_t *buf, size_t buf_len);
 std::string base64_encode(const std::vector<uint8_t> &buf);
 

platformio.ini

@@ -58,7 +58,7 @@ lib_deps =
     SPI                                                   ; spi (Arduino built-in)
     Wire                                                  ; i2c (Arduino built-int)
     heman/AsyncMqttClient-esphome@1.0.0                   ; mqtt
-    esphome/ESPAsyncWebServer-esphome@3.2.0               ; web_server_base
+    esphome/ESPAsyncWebServer-esphome@3.2.2               ; web_server_base
     fastled/FastLED@3.3.2                                 ; fastled_base
     mikalhart/TinyGPSPlus@1.0.2                           ; gps
     freekode/TM1651@1.0.1                                 ; tm1651