Add voice assistant announce (#7377)

esphome/components/api/api.proto

@@ -1553,6 +1553,23 @@ message VoiceAssistantTimerEventResponse {
   bool is_active = 6;
 }
 
+message VoiceAssistantAnnounceRequest {
+  option (id) = 119;
+  option (source) = SOURCE_CLIENT;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  string media_id = 1;
+  string text = 2;
+}
+
+message VoiceAssistantAnnounceFinished {
+  option (id) = 120;
+  option (source) = SOURCE_SERVER;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  bool success = 1;
+}
+
 // ==================== ALARM CONTROL PANEL ====================
 enum AlarmControlPanelState {
   ALARM_STATE_DISARMED = 0;

esphome/components/api/api_connection.cpp

@@ -1213,6 +1213,16 @@ void APIConnection::on_voice_assistant_timer_event_response(const VoiceAssistant
   }
 };
 
+void APIConnection::on_voice_assistant_announce_request(const VoiceAssistantAnnounceRequest &msg) {
+  if (voice_assistant::global_voice_assistant != nullptr) {
+    if (voice_assistant::global_voice_assistant->get_api_connection() != this) {
+      return;
+    }
+
+    voice_assistant::global_voice_assistant->on_announce(msg);
+  }
+}
+
 #endif
 
 #ifdef USE_ALARM_CONTROL_PANEL

esphome/components/api/api_connection.h

@@ -151,6 +151,7 @@ class APIConnection : public APIServerConnection {
   void on_voice_assistant_event_response(const VoiceAssistantEventResponse &msg) override;
   void on_voice_assistant_audio(const VoiceAssistantAudio &msg) override;
   void on_voice_assistant_timer_event_response(const VoiceAssistantTimerEventResponse &msg) override;
+  void on_voice_assistant_announce_request(const VoiceAssistantAnnounceRequest &msg) override;
 #endif
 
 #ifdef USE_ALARM_CONTROL_PANEL

esphome/components/api/api_pb2.cpp

@@ -7061,6 +7061,59 @@ void VoiceAssistantTimerEventResponse::dump_to(std::string &out) const {
   out.append("}");
 }
 #endif
+bool VoiceAssistantAnnounceRequest::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
+  switch (field_id) {
+    case 1: {
+      this->media_id = value.as_string();
+      return true;
+    }
+    case 2: {
+      this->text = value.as_string();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantAnnounceRequest::encode(ProtoWriteBuffer buffer) const {
+  buffer.encode_string(1, this->media_id);
+  buffer.encode_string(2, this->text);
+}
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantAnnounceRequest::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantAnnounceRequest {\n");
+  out.append("  media_id: ");
+  out.append("'").append(this->media_id).append("'");
+  out.append("\n");
+
+  out.append("  text: ");
+  out.append("'").append(this->text).append("'");
+  out.append("\n");
+  out.append("}");
+}
+#endif
+bool VoiceAssistantAnnounceFinished::decode_varint(uint32_t field_id, ProtoVarInt value) {
+  switch (field_id) {
+    case 1: {
+      this->success = value.as_bool();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantAnnounceFinished::encode(ProtoWriteBuffer buffer) const { buffer.encode_bool(1, this->success); }
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantAnnounceFinished::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantAnnounceFinished {\n");
+  out.append("  success: ");
+  out.append(YESNO(this->success));
+  out.append("\n");
+  out.append("}");
+}
+#endif
 bool ListEntitiesAlarmControlPanelResponse::decode_varint(uint32_t field_id, ProtoVarInt value) {
   switch (field_id) {
     case 6: {

esphome/components/api/api_pb2.h

@@ -1825,6 +1825,29 @@ class VoiceAssistantTimerEventResponse : public ProtoMessage {
   bool decode_length(uint32_t field_id, ProtoLengthDelimited value) override;
   bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
 };
+class VoiceAssistantAnnounceRequest : public ProtoMessage {
+ public:
+  std::string media_id{};
+  std::string text{};
+  void encode(ProtoWriteBuffer buffer) const override;
+#ifdef HAS_PROTO_MESSAGE_DUMP
+  void dump_to(std::string &out) const override;
+#endif
+
+ protected:
+  bool decode_length(uint32_t field_id, ProtoLengthDelimited value) override;
+};
+class VoiceAssistantAnnounceFinished : public ProtoMessage {
+ public:
+  bool success{false};
+  void encode(ProtoWriteBuffer buffer) const override;
+#ifdef HAS_PROTO_MESSAGE_DUMP
+  void dump_to(std::string &out) const override;
+#endif
+
+ protected:
+  bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
+};
 class ListEntitiesAlarmControlPanelResponse : public ProtoMessage {
  public:
   std::string object_id{};

esphome/components/api/api_pb2_service.cpp

@@ -486,6 +486,16 @@ bool APIServerConnectionBase::send_voice_assistant_audio(const VoiceAssistantAud
 #endif
 #ifdef USE_VOICE_ASSISTANT
 #endif
+#ifdef USE_VOICE_ASSISTANT
+#endif
+#ifdef USE_VOICE_ASSISTANT
+bool APIServerConnectionBase::send_voice_assistant_announce_finished(const VoiceAssistantAnnounceFinished &msg) {
+#ifdef HAS_PROTO_MESSAGE_DUMP
+  ESP_LOGVV(TAG, "send_voice_assistant_announce_finished: %s", msg.dump().c_str());
+#endif
+  return this->send_message_<VoiceAssistantAnnounceFinished>(msg, 120);
+}
+#endif
 #ifdef USE_ALARM_CONTROL_PANEL
 bool APIServerConnectionBase::send_list_entities_alarm_control_panel_response(
     const ListEntitiesAlarmControlPanelResponse &msg) {
@@ -1135,6 +1145,17 @@ bool APIServerConnectionBase::read_message(uint32_t msg_size, uint32_t msg_type,
       ESP_LOGVV(TAG, "on_update_command_request: %s", msg.dump().c_str());
 #endif
       this->on_update_command_request(msg);
+#endif
+      break;
+    }
+    case 119: {
+#ifdef USE_VOICE_ASSISTANT
+      VoiceAssistantAnnounceRequest msg;
+      msg.decode(msg_data, msg_size);
+#ifdef HAS_PROTO_MESSAGE_DUMP
+      ESP_LOGVV(TAG, "on_voice_assistant_announce_request: %s", msg.dump().c_str());
+#endif
+      this->on_voice_assistant_announce_request(msg);
 #endif
       break;
     }

esphome/components/api/api_pb2_service.h

@@ -247,6 +247,12 @@ class APIServerConnectionBase : public ProtoService {
 #ifdef USE_VOICE_ASSISTANT
   virtual void on_voice_assistant_timer_event_response(const VoiceAssistantTimerEventResponse &value){};
 #endif
+#ifdef USE_VOICE_ASSISTANT
+  virtual void on_voice_assistant_announce_request(const VoiceAssistantAnnounceRequest &value){};
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  bool send_voice_assistant_announce_finished(const VoiceAssistantAnnounceFinished &msg);
+#endif
 #ifdef USE_ALARM_CONTROL_PANEL
   bool send_list_entities_alarm_control_panel_response(const ListEntitiesAlarmControlPanelResponse &msg);
 #endif

esphome/components/voice_assistant/voice_assistant.cpp

@@ -396,6 +396,10 @@ void VoiceAssistant::loop() {
         this->set_timeout("playing", 2000, [this]() {
           this->cancel_timeout("speaker-timeout");
           this->set_state_(State::IDLE, State::IDLE);
+
+          api::VoiceAssistantAnnounceFinished msg;
+          msg.success = true;
+          this->api_client_->send_voice_assistant_announce_finished(msg);
         });
       }
       break;
@@ -866,6 +870,18 @@ void VoiceAssistant::timer_tick_() {
   this->timer_tick_trigger_->trigger(res);
 }
 
+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_->make_call().set_media_url(msg.media_id).set_announcement(true).perform();
+    this->set_state_(State::STREAMING_RESPONSE, State::STREAMING_RESPONSE);
+    this->tts_end_trigger_->trigger(msg.media_id);
+    this->end_trigger_->trigger();
+  }
+#endif
+}
+
 VoiceAssistant *global_voice_assistant = nullptr;  // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
 
 }  // namespace voice_assistant

esphome/components/voice_assistant/voice_assistant.h

@@ -132,6 +132,7 @@ class VoiceAssistant : public Component {
   void on_event(const api::VoiceAssistantEventResponse &msg);
   void on_audio(const api::VoiceAssistantAudio &msg);
   void on_timer_event(const api::VoiceAssistantTimerEventResponse &msg);
+  void on_announce(const api::VoiceAssistantAnnounceRequest &msg);
 
   bool is_running() const { return this->state_ != State::IDLE; }
   void set_continuous(bool continuous) { this->continuous_ = continuous; }