Merge pull request #5548 from esphome/bump-2023.10.0b3

2023.10.0b3

esphome/components/api/api.proto

@@ -1459,6 +1459,8 @@ enum VoiceAssistantEvent {
   VOICE_ASSISTANT_WAKE_WORD_END = 10;
   VOICE_ASSISTANT_STT_VAD_START = 11;
   VOICE_ASSISTANT_STT_VAD_END = 12;
+  VOICE_ASSISTANT_TTS_STREAM_START = 98;
+  VOICE_ASSISTANT_TTS_STREAM_END = 99;
 }
 
 message VoiceAssistantEventData {

esphome/components/api/api_pb2.cpp

@@ -452,6 +452,10 @@ template<> const char *proto_enum_to_string<enums::VoiceAssistantEvent>(enums::V
       return "VOICE_ASSISTANT_STT_VAD_START";
     case enums::VOICE_ASSISTANT_STT_VAD_END:
       return "VOICE_ASSISTANT_STT_VAD_END";
+    case enums::VOICE_ASSISTANT_TTS_STREAM_START:
+      return "VOICE_ASSISTANT_TTS_STREAM_START";
+    case enums::VOICE_ASSISTANT_TTS_STREAM_END:
+      return "VOICE_ASSISTANT_TTS_STREAM_END";
     default:
       return "UNKNOWN";
   }

esphome/components/api/api_pb2.h

@@ -184,6 +184,8 @@ enum VoiceAssistantEvent : uint32_t {
   VOICE_ASSISTANT_WAKE_WORD_END = 10,
   VOICE_ASSISTANT_STT_VAD_START = 11,
   VOICE_ASSISTANT_STT_VAD_END = 12,
+  VOICE_ASSISTANT_TTS_STREAM_START = 98,
+  VOICE_ASSISTANT_TTS_STREAM_END = 99,
 };
 enum AlarmControlPanelState : uint32_t {
   ALARM_STATE_DISARMED = 0,

esphome/components/esp32_improv/esp32_improv_component.cpp

@@ -107,6 +107,7 @@ void ESP32ImprovComponent::loop() {
       break;
     }
     case improv::STATE_AUTHORIZED: {
+#ifdef USE_BINARY_SENSOR
       if (this->authorizer_ != nullptr) {
         if (now - this->authorized_start_ > this->authorized_duration_) {
           ESP_LOGD(TAG, "Authorization timeout");
@@ -114,6 +115,7 @@ void ESP32ImprovComponent::loop() {
           return;
         }
       }
+#endif
       if (!this->check_identify_()) {
         this->set_status_indicator_state_((now % 1000) < 500);
       }
@@ -290,8 +292,10 @@ void ESP32ImprovComponent::process_incoming_data_() {
 void ESP32ImprovComponent::on_wifi_connect_timeout_() {
   this->set_error_(improv::ERROR_UNABLE_TO_CONNECT);
   this->set_state_(improv::STATE_AUTHORIZED);
+#ifdef USE_BINARY_SENSOR
   if (this->authorizer_ != nullptr)
     this->authorized_start_ = millis();
+#endif
   ESP_LOGW(TAG, "Timed out trying to connect to given WiFi network");
   wifi::global_wifi_component->clear_sta();
 }

esphome/components/grove_tb6612fng/__init__.py

@@ -8,12 +8,15 @@ from esphome.const import (
     CONF_CHANNEL,
     CONF_SPEED,
     CONF_DIRECTION,
+    CONF_ADDRESS,
 )
 
 DEPENDENCIES = ["i2c"]
 
 CODEOWNERS = ["@max246"]
 
+MULTI_CONF = True
+
 grove_tb6612fng_ns = cg.esphome_ns.namespace("grove_tb6612fng")
 GROVE_TB6612FNG = grove_tb6612fng_ns.class_(
     "GroveMotorDriveTB6612FNG", cg.Component, i2c.I2CDevice
@@ -33,6 +36,9 @@ GROVETB6612FNGMotorStandbyAction = grove_tb6612fng_ns.class_(
 GROVETB6612FNGMotorNoStandbyAction = grove_tb6612fng_ns.class_(
     "GROVETB6612FNGMotorNoStandbyAction", automation.Action
 )
+GROVETB6612FNGMotorChangeAddressAction = grove_tb6612fng_ns.class_(
+    "GROVETB6612FNGMotorChangeAddressAction", automation.Action
+)
 
 DIRECTION_TYPE = {
     "FORWARD": 1,
@@ -150,3 +156,22 @@ async def grove_tb6612fng_no_standby_to_code(config, action_id, template_arg, ar
     await cg.register_parented(var, config[CONF_ID])
 
     return var
+
+
+@automation.register_action(
+    "grove_tb6612fng.change_address",
+    GROVETB6612FNGMotorChangeAddressAction,
+    cv.Schema(
+        {
+            cv.Required(CONF_ID): cv.use_id(GROVE_TB6612FNG),
+            cv.Required(CONF_ADDRESS): cv.i2c_address,
+        }
+    ),
+)
+async def grove_tb6612fng_change_address_to_code(config, action_id, template_arg, args):
+    var = cg.new_Pvariable(action_id, template_arg)
+    await cg.register_parented(var, config[CONF_ID])
+
+    template_channel = await cg.templatable(config[CONF_ADDRESS], args, int)
+    cg.add(var.set_address(template_channel))
+    return var

esphome/components/grove_tb6612fng/grove_tb6612fng.h

@@ -84,8 +84,7 @@ class GroveMotorDriveTB6612FNG : public Component, public i2c::I2CDevice {
   *************************************************************/
   void set_i2c_addr(uint8_t addr);
 
-  /*************************************************************
+  /***********************************change_address
-      Description
        Drive a motor.
       Parameter
        chl: MOTOR_CHA or MOTOR_CHB
@@ -204,5 +203,13 @@ class GROVETB6612FNGMotorNoStandbyAction : public Action<Ts...>, public Parented
   void play(Ts... x) override { this->parent_->not_standby(); }
 };
 
+template<typename... Ts>
+class GROVETB6612FNGMotorChangeAddressAction : public Action<Ts...>, public Parented<GroveMotorDriveTB6612FNG> {
+ public:
+  TEMPLATABLE_VALUE(uint8_t, address)
+
+  void play(Ts... x) override { this->parent_->set_i2c_addr(this->address_.value(x...)); }
+};
+
 }  // namespace grove_tb6612fng
 }  // namespace esphome

esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp

@@ -37,6 +37,8 @@ void I2SAudioMicrophone::setup() {
 void I2SAudioMicrophone::start() {
   if (this->is_failed())
     return;
+  if (this->state_ == microphone::STATE_RUNNING)
+    return;  // Already running
   this->state_ = microphone::STATE_STARTING;
 }
 void I2SAudioMicrophone::start_() {

esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp

@@ -158,8 +158,13 @@ void I2SAudioSpeaker::watch_() {
   if (xQueueReceive(this->event_queue_, &event, 0) == pdTRUE) {
     switch (event.type) {
       case TaskEventType::STARTING:
+        ESP_LOGD(TAG, "Starting I2S Audio Speaker");
+        break;
       case TaskEventType::STARTED:
+        ESP_LOGD(TAG, "Started I2S Audio Speaker");
+        break;
       case TaskEventType::STOPPING:
+        ESP_LOGD(TAG, "Stopping I2S Audio Speaker");
         break;
       case TaskEventType::PLAYING:
         this->status_clear_warning();
@@ -170,6 +175,7 @@ void I2SAudioSpeaker::watch_() {
         this->player_task_handle_ = nullptr;
         this->parent_->unlock();
         xQueueReset(this->buffer_queue_);
+        ESP_LOGD(TAG, "Stopped I2S Audio Speaker");
         break;
       case TaskEventType::WARNING:
         ESP_LOGW(TAG, "Error writing to I2S: %s", esp_err_to_name(event.err));

esphome/components/network/ip_address.h

@@ -87,7 +87,7 @@ struct IPAddress {
   bool is_ip6() { return IP_IS_V6(&ip_addr_); }
   std::string str() const { return ipaddr_ntoa(&ip_addr_); }
   bool operator==(const IPAddress &other) const { return ip_addr_cmp(&ip_addr_, &other.ip_addr_); }
-  bool operator!=(const IPAddress &other) const { return !(&ip_addr_ == &other.ip_addr_); }
+  bool operator!=(const IPAddress &other) const { return !ip_addr_cmp(&ip_addr_, &other.ip_addr_); }
   IPAddress &operator+=(uint8_t increase) {
     if (IP_IS_V4(&ip_addr_)) {
 #if LWIP_IPV6

esphome/components/prometheus/prometheus_handler.cpp

@@ -1,5 +1,3 @@
-#ifdef USE_ARDUINO
-
 #include "prometheus_handler.h"
 #include "esphome/core/application.h"
 
@@ -89,7 +87,7 @@ void PrometheusHandler::sensor_row_(AsyncResponseStream *stream, sensor::Sensor
     stream->print(obj->get_unit_of_measurement().c_str());
     stream->print(F("\"} "));
     stream->print(value_accuracy_to_string(obj->state, obj->get_accuracy_decimals()).c_str());
-    stream->print('\n');
+    stream->print(F("\n"));
   } else {
     // Invalid state
     stream->print(F("esphome_sensor_failed{id=\""));
@@ -124,7 +122,7 @@ void PrometheusHandler::binary_sensor_row_(AsyncResponseStream *stream, binary_s
     stream->print(relabel_name_(obj).c_str());
     stream->print(F("\"} "));
     stream->print(obj->state);
-    stream->print('\n');
+    stream->print(F("\n"));
   } else {
     // Invalid state
     stream->print(F("esphome_binary_sensor_failed{id=\""));
@@ -158,7 +156,7 @@ void PrometheusHandler::fan_row_(AsyncResponseStream *stream, fan::Fan *obj) {
   stream->print(relabel_name_(obj).c_str());
   stream->print(F("\"} "));
   stream->print(obj->state);
-  stream->print('\n');
+  stream->print(F("\n"));
   // Speed if available
   if (obj->get_traits().supports_speed()) {
     stream->print(F("esphome_fan_speed{id=\""));
@@ -167,7 +165,7 @@ void PrometheusHandler::fan_row_(AsyncResponseStream *stream, fan::Fan *obj) {
     stream->print(relabel_name_(obj).c_str());
     stream->print(F("\"} "));
     stream->print(obj->speed);
-    stream->print('\n');
+    stream->print(F("\n"));
   }
   // Oscillation if available
   if (obj->get_traits().supports_oscillation()) {
@@ -177,7 +175,7 @@ void PrometheusHandler::fan_row_(AsyncResponseStream *stream, fan::Fan *obj) {
     stream->print(relabel_name_(obj).c_str());
     stream->print(F("\"} "));
     stream->print(obj->oscillating);
-    stream->print('\n');
+    stream->print(F("\n"));
   }
 }
 #endif
@@ -281,7 +279,7 @@ void PrometheusHandler::cover_row_(AsyncResponseStream *stream, cover::Cover *ob
     stream->print(relabel_name_(obj).c_str());
     stream->print(F("\"} "));
     stream->print(obj->position);
-    stream->print('\n');
+    stream->print(F("\n"));
     if (obj->get_traits().get_supports_tilt()) {
       stream->print(F("esphome_cover_tilt{id=\""));
       stream->print(relabel_id_(obj).c_str());
@@ -289,7 +287,7 @@ void PrometheusHandler::cover_row_(AsyncResponseStream *stream, cover::Cover *ob
       stream->print(relabel_name_(obj).c_str());
       stream->print(F("\"} "));
       stream->print(obj->tilt);
-      stream->print('\n');
+      stream->print(F("\n"));
     }
   } else {
     // Invalid state
@@ -322,7 +320,7 @@ void PrometheusHandler::switch_row_(AsyncResponseStream *stream, switch_::Switch
   stream->print(relabel_name_(obj).c_str());
   stream->print(F("\"} "));
   stream->print(obj->state);
-  stream->print('\n');
+  stream->print(F("\n"));
 }
 #endif
 
@@ -346,11 +344,9 @@ void PrometheusHandler::lock_row_(AsyncResponseStream *stream, lock::Lock *obj)
   stream->print(relabel_name_(obj).c_str());
   stream->print(F("\"} "));
   stream->print(obj->state);
-  stream->print('\n');
+  stream->print(F("\n"));
 }
 #endif
 
 }  // namespace prometheus
 }  // namespace esphome
-
-#endif  // USE_ARDUINO

esphome/components/prometheus/prometheus_handler.h

@@ -1,14 +1,12 @@
 #pragma once
 
-#ifdef USE_ARDUINO
-
 #include <map>
 #include <utility>
 
-#include "esphome/core/entity_base.h"
 #include "esphome/components/web_server_base/web_server_base.h"
-#include "esphome/core/controller.h"
 #include "esphome/core/component.h"
+#include "esphome/core/controller.h"
+#include "esphome/core/entity_base.h"
 
 namespace esphome {
 namespace prometheus {
@@ -119,5 +117,3 @@ class PrometheusHandler : public AsyncWebHandler, public Component {
 
 }  // namespace prometheus
 }  // namespace esphome
-
-#endif  // USE_ARDUINO

esphome/components/voice_assistant/voice_assistant.cpp

@@ -281,11 +281,15 @@ void VoiceAssistant::loop() {
             memmove(this->speaker_buffer_, this->speaker_buffer_ + written, this->speaker_buffer_size_ - written);
             this->speaker_buffer_size_ -= written;
             this->speaker_buffer_index_ -= written;
-            this->set_timeout("speaker-timeout", 1000, [this]() { this->speaker_->stop(); });
+            this->set_timeout("speaker-timeout", 2000, [this]() { this->speaker_->stop(); });
           } else {
             ESP_LOGW(TAG, "Speaker buffer full.");
           }
         }
+        if (this->wait_for_stream_end_) {
+          this->cancel_timeout("playing");
+          break;  // We dont want to timeout here as the STREAM_END event will take care of that.
+        }
         playing = this->speaker_->is_running();
       }
 #endif
@@ -295,28 +299,77 @@ void VoiceAssistant::loop() {
       }
 #endif
       if (playing) {
-        this->set_timeout("playing", 100, [this]() {
+        this->set_timeout("playing", 2000, [this]() {
           this->cancel_timeout("speaker-timeout");
           this->set_state_(State::IDLE, State::IDLE);
         });
       }
       break;
     }
+    case State::RESPONSE_FINISHED: {
+#ifdef USE_SPEAKER
+      if (this->speaker_ != nullptr) {
+        this->speaker_->stop();
+        this->cancel_timeout("speaker-timeout");
+        this->cancel_timeout("playing");
+        this->speaker_buffer_size_ = 0;
+        this->speaker_buffer_index_ = 0;
+        memset(this->speaker_buffer_, 0, SPEAKER_BUFFER_SIZE);
+      }
+#endif
+      this->wait_for_stream_end_ = false;
+      this->set_state_(State::IDLE, State::IDLE);
+      break;
+    }
     default:
       break;
   }
 }
 
+static const LogString *voice_assistant_state_to_string(State state) {
+  switch (state) {
+    case State::IDLE:
+      return LOG_STR("IDLE");
+    case State::START_MICROPHONE:
+      return LOG_STR("START_MICROPHONE");
+    case State::STARTING_MICROPHONE:
+      return LOG_STR("STARTING_MICROPHONE");
+    case State::WAIT_FOR_VAD:
+      return LOG_STR("WAIT_FOR_VAD");
+    case State::WAITING_FOR_VAD:
+      return LOG_STR("WAITING_FOR_VAD");
+    case State::START_PIPELINE:
+      return LOG_STR("START_PIPELINE");
+    case State::STARTING_PIPELINE:
+      return LOG_STR("STARTING_PIPELINE");
+    case State::STREAMING_MICROPHONE:
+      return LOG_STR("STREAMING_MICROPHONE");
+    case State::STOP_MICROPHONE:
+      return LOG_STR("STOP_MICROPHONE");
+    case State::STOPPING_MICROPHONE:
+      return LOG_STR("STOPPING_MICROPHONE");
+    case State::AWAITING_RESPONSE:
+      return LOG_STR("AWAITING_RESPONSE");
+    case State::STREAMING_RESPONSE:
+      return LOG_STR("STREAMING_RESPONSE");
+    case State::RESPONSE_FINISHED:
+      return LOG_STR("RESPONSE_FINISHED");
+    default:
+      return LOG_STR("UNKNOWN");
+  }
+};
+
 void VoiceAssistant::set_state_(State state) {
   State old_state = this->state_;
   this->state_ = state;
-  ESP_LOGD(TAG, "State changed from %d to %d", static_cast<uint8_t>(old_state), static_cast<uint8_t>(state));
+  ESP_LOGD(TAG, "State changed from %s to %s", LOG_STR_ARG(voice_assistant_state_to_string(old_state)),
+           LOG_STR_ARG(voice_assistant_state_to_string(state)));
 }
 
 void VoiceAssistant::set_state_(State state, State desired_state) {
   this->set_state_(state);
   this->desired_state_ = desired_state;
-  ESP_LOGD(TAG, "Desired state set to %d", static_cast<uint8_t>(desired_state));
+  ESP_LOGD(TAG, "Desired state set to %s", LOG_STR_ARG(voice_assistant_state_to_string(desired_state)));
 }
 
 void VoiceAssistant::failed_to_start() {
@@ -400,6 +453,7 @@ void VoiceAssistant::request_stop() {
       break;
     case State::AWAITING_RESPONSE:
     case State::STREAMING_RESPONSE:
+    case State::RESPONSE_FINISHED:
       break;  // Let the incoming audio stream finish then it will go to idle.
   }
 }
@@ -531,6 +585,14 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
       this->error_trigger_->trigger(code, message);
       break;
     }
+    case api::enums::VOICE_ASSISTANT_TTS_STREAM_START: {
+      this->wait_for_stream_end_ = true;
+      break;
+    }
+    case api::enums::VOICE_ASSISTANT_TTS_STREAM_END: {
+      this->set_state_(State::RESPONSE_FINISHED, State::IDLE);
+      break;
+    }
     default:
       ESP_LOGD(TAG, "Unhandled event type: %d", msg.event_type);
       break;

esphome/components/voice_assistant/voice_assistant.h

@@ -46,6 +46,7 @@ enum class State {
   STOPPING_MICROPHONE,
   AWAITING_RESPONSE,
   STREAMING_RESPONSE,
+  RESPONSE_FINISHED,
 };
 
 class VoiceAssistant : public Component {
@@ -132,10 +133,10 @@ class VoiceAssistant : public Component {
   uint8_t *speaker_buffer_;
   size_t speaker_buffer_index_{0};
   size_t speaker_buffer_size_{0};
+  bool wait_for_stream_end_{false};
 #endif
 #ifdef USE_MEDIA_PLAYER
   media_player::MediaPlayer *media_player_{nullptr};
-  bool playing_tts_{false};
 #endif
 
   bool local_output_{false};

esphome/const.py

@@ -1,6 +1,6 @@
 """Constants used by esphome."""
 
-__version__ = "2023.10.0b2"
+__version__ = "2023.10.0b3"
 
 ALLOWED_NAME_CHARS = "abcdefghijklmnopqrstuvwxyz0123456789-_"
 VALID_SUBSTITUTIONS_CHARACTERS = (