Add push to talk voice assistant (#4648)

* Add push to talk voice assistant

* Refactor most code into voice_assistant

* Make voice_assistant the component and remove push_to_talk (can be done in yaml)

* Fix component setup

* Always AF_INET to match serverside

* Fix microphone and media player co-existence

* Format

* Update codeowners

* Update test file

* Fix endifs

* nullptr not NULL

* clang-tidy

* Format

* fixup: Add VA event data

* Generate proto

* Parse and log events

* Add default to switch

* Fix

* Add mic/va to test5

esphome/components/api/api.proto

@@ -55,6 +55,7 @@ service APIConnection {
   rpc subscribe_bluetooth_connections_free(SubscribeBluetoothConnectionsFreeRequest) returns (BluetoothConnectionsFreeResponse) {}
   rpc unsubscribe_bluetooth_le_advertisements(UnsubscribeBluetoothLEAdvertisementsRequest) returns (void) {}
 
+  rpc subscribe_voice_assistant(SubscribeVoiceAssistantRequest) returns (void) {}
 }
 
 
@@ -210,6 +211,8 @@ message DeviceInfoResponse {
   string manufacturer = 12;
 
   string friendly_name = 13;
+
+  uint32 voice_assistant_version = 14;
 }
 
 message ListEntitiesRequest {
@@ -1379,3 +1382,55 @@ message BluetoothDeviceClearCacheResponse {
   bool success = 2;
   int32 error = 3;
 }
+
+// ==================== PUSH TO TALK ====================
+message SubscribeVoiceAssistantRequest {
+  option (id) = 89;
+  option (source) = SOURCE_CLIENT;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  bool subscribe = 1;
+}
+
+message VoiceAssistantRequest {
+  option (id) = 90;
+  option (source) = SOURCE_SERVER;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  bool start = 1;
+}
+
+message VoiceAssistantResponse {
+  option (id) = 91;
+  option (source) = SOURCE_CLIENT;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  uint32 port = 1;
+  bool error = 2;
+}
+
+enum VoiceAssistantEvent {
+  VOICE_ASSISTANT_ERROR = 0;
+  VOICE_ASSISTANT_RUN_START = 1;
+  VOICE_ASSISTANT_RUN_END = 2;
+  VOICE_ASSISTANT_STT_START = 3;
+  VOICE_ASSISTANT_STT_END = 4;
+  VOICE_ASSISTANT_INTENT_START = 5;
+  VOICE_ASSISTANT_INTENT_END = 6;
+  VOICE_ASSISTANT_TTS_START = 7;
+  VOICE_ASSISTANT_TTS_END = 8;
+}
+
+message VoiceAssistantEventData {
+  string name = 1;
+  string value = 2;
+}
+
+message VoiceAssistantEventResponse {
+  option (id) = 92;
+  option (source) = SOURCE_CLIENT;
+  option (ifdef) = "USE_VOICE_ASSISTANT";
+
+  VoiceAssistantEvent event_type = 1;
+  repeated VoiceAssistantEventData data = 2;
+}

esphome/components/api/api_connection.cpp

@@ -16,6 +16,9 @@
 #ifdef USE_BLUETOOTH_PROXY
 #include "esphome/components/bluetooth_proxy/bluetooth_proxy.h"
 #endif
+#ifdef USE_VOICE_ASSISTANT
+#include "esphome/components/voice_assistant/voice_assistant.h"
+#endif
 
 namespace esphome {
 namespace api {
@@ -893,6 +896,30 @@ BluetoothConnectionsFreeResponse APIConnection::subscribe_bluetooth_connections_
 }
 #endif
 
+#ifdef USE_VOICE_ASSISTANT
+bool APIConnection::request_voice_assistant(bool start) {
+  if (!this->voice_assistant_subscription_)
+    return false;
+  VoiceAssistantRequest msg;
+  msg.start = start;
+  return this->send_voice_assistant_request(msg);
+}
+void APIConnection::on_voice_assistant_response(const VoiceAssistantResponse &msg) {
+  if (voice_assistant::global_voice_assistant != nullptr) {
+    struct sockaddr_storage storage;
+    socklen_t len = sizeof(storage);
+    this->helper_->getpeername((struct sockaddr *) &storage, &len);
+    voice_assistant::global_voice_assistant->start(&storage, msg.port);
+  }
+};
+void APIConnection::on_voice_assistant_event_response(const VoiceAssistantEventResponse &msg) {
+  if (voice_assistant::global_voice_assistant != nullptr) {
+    voice_assistant::global_voice_assistant->on_event(msg);
+  }
+}
+
+#endif
+
 bool APIConnection::send_log_message(int level, const char *tag, const char *line) {
   if (this->log_subscription_ < level)
     return false;
@@ -970,6 +997,9 @@ DeviceInfoResponse APIConnection::device_info(const DeviceInfoRequest &msg) {
   resp.bluetooth_proxy_version = bluetooth_proxy::global_bluetooth_proxy->has_active()
                                      ? bluetooth_proxy::ACTIVE_CONNECTIONS_VERSION
                                      : bluetooth_proxy::PASSIVE_ONLY_VERSION;
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  resp.voice_assistant_version = 1;
 #endif
   return resp;
 }

esphome/components/api/api_connection.h

@@ -6,6 +6,7 @@
 #include "api_server.h"
 #include "esphome/core/application.h"
 #include "esphome/core/component.h"
+#include "esphome/core/defines.h"
 
 #include <vector>
 
@@ -123,6 +124,15 @@ class APIConnection : public APIServerConnection {
   }
 #endif
 
+#ifdef USE_VOICE_ASSISTANT
+  void subscribe_voice_assistant(const SubscribeVoiceAssistantRequest &msg) override {
+    this->voice_assistant_subscription_ = msg.subscribe;
+  }
+  bool request_voice_assistant(bool start);
+  void on_voice_assistant_response(const VoiceAssistantResponse &msg) override;
+  void on_voice_assistant_event_response(const VoiceAssistantEventResponse &msg) override;
+#endif
+
   void on_disconnect_response(const DisconnectResponse &value) override;
   void on_ping_response(const PingResponse &value) override {
     // we initiated ping
@@ -203,6 +213,9 @@ class APIConnection : public APIServerConnection {
   bool service_call_subscription_{false};
 #ifdef USE_BLUETOOTH_PROXY
   bool bluetooth_le_advertisement_subscription_{false};
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  bool voice_assistant_subscription_{false};
 #endif
   bool next_close_ = false;
   APIServer *parent_;

esphome/components/api/api_frame_helper.h

@@ -10,8 +10,8 @@
 #include "noise/protocol.h"
 #endif
 
-#include "esphome/components/socket/socket.h"
 #include "api_noise_context.h"
+#include "esphome/components/socket/socket.h"
 
 namespace esphome {
 namespace api {
@@ -67,6 +67,7 @@ class APIFrameHelper {
   virtual bool can_write_without_blocking() = 0;
   virtual APIError write_packet(uint16_t type, const uint8_t *data, size_t len) = 0;
   virtual std::string getpeername() = 0;
+  virtual int getpeername(struct sockaddr *addr, socklen_t *addrlen) = 0;
   virtual APIError close() = 0;
   virtual APIError shutdown(int how) = 0;
   // Give this helper a name for logging
@@ -84,7 +85,10 @@ class APINoiseFrameHelper : public APIFrameHelper {
   APIError read_packet(ReadPacketBuffer *buffer) override;
   bool can_write_without_blocking() override;
   APIError write_packet(uint16_t type, const uint8_t *payload, size_t len) override;
-  std::string getpeername() override { return socket_->getpeername(); }
+  std::string getpeername() override { return this->socket_->getpeername(); }
+  int getpeername(struct sockaddr *addr, socklen_t *addrlen) override {
+    return this->socket_->getpeername(addr, addrlen);
+  }
   APIError close() override;
   APIError shutdown(int how) override;
   // Give this helper a name for logging
@@ -144,7 +148,10 @@ class APIPlaintextFrameHelper : public APIFrameHelper {
   APIError read_packet(ReadPacketBuffer *buffer) override;
   bool can_write_without_blocking() override;
   APIError write_packet(uint16_t type, const uint8_t *payload, size_t len) override;
-  std::string getpeername() override { return socket_->getpeername(); }
+  std::string getpeername() override { return this->socket_->getpeername(); }
+  int getpeername(struct sockaddr *addr, socklen_t *addrlen) override {
+    return this->socket_->getpeername(addr, addrlen);
+  }
   APIError close() override;
   APIError shutdown(int how) override;
   // Give this helper a name for logging

esphome/components/api/api_pb2.cpp

@@ -407,6 +407,32 @@ const char *proto_enum_to_string<enums::BluetoothDeviceRequestType>(enums::Bluet
   }
 }
 #endif
+#ifdef HAS_PROTO_MESSAGE_DUMP
+template<> const char *proto_enum_to_string<enums::VoiceAssistantEvent>(enums::VoiceAssistantEvent value) {
+  switch (value) {
+    case enums::VOICE_ASSISTANT_ERROR:
+      return "VOICE_ASSISTANT_ERROR";
+    case enums::VOICE_ASSISTANT_RUN_START:
+      return "VOICE_ASSISTANT_RUN_START";
+    case enums::VOICE_ASSISTANT_RUN_END:
+      return "VOICE_ASSISTANT_RUN_END";
+    case enums::VOICE_ASSISTANT_STT_START:
+      return "VOICE_ASSISTANT_STT_START";
+    case enums::VOICE_ASSISTANT_STT_END:
+      return "VOICE_ASSISTANT_STT_END";
+    case enums::VOICE_ASSISTANT_INTENT_START:
+      return "VOICE_ASSISTANT_INTENT_START";
+    case enums::VOICE_ASSISTANT_INTENT_END:
+      return "VOICE_ASSISTANT_INTENT_END";
+    case enums::VOICE_ASSISTANT_TTS_START:
+      return "VOICE_ASSISTANT_TTS_START";
+    case enums::VOICE_ASSISTANT_TTS_END:
+      return "VOICE_ASSISTANT_TTS_END";
+    default:
+      return "UNKNOWN";
+  }
+}
+#endif
 bool HelloRequest::decode_varint(uint32_t field_id, ProtoVarInt value) {
   switch (field_id) {
     case 2: {
@@ -594,6 +620,10 @@ bool DeviceInfoResponse::decode_varint(uint32_t field_id, ProtoVarInt value) {
       this->bluetooth_proxy_version = value.as_uint32();
       return true;
     }
+    case 14: {
+      this->voice_assistant_version = value.as_uint32();
+      return true;
+    }
     default:
       return false;
   }
@@ -654,6 +684,7 @@ void DeviceInfoResponse::encode(ProtoWriteBuffer buffer) const {
   buffer.encode_uint32(11, this->bluetooth_proxy_version);
   buffer.encode_string(12, this->manufacturer);
   buffer.encode_string(13, this->friendly_name);
+  buffer.encode_uint32(14, this->voice_assistant_version);
 }
 #ifdef HAS_PROTO_MESSAGE_DUMP
 void DeviceInfoResponse::dump_to(std::string &out) const {
@@ -712,6 +743,11 @@ void DeviceInfoResponse::dump_to(std::string &out) const {
   out.append("  friendly_name: ");
   out.append("'").append(this->friendly_name).append("'");
   out.append("\n");
+
+  out.append("  voice_assistant_version: ");
+  sprintf(buffer, "%u", this->voice_assistant_version);
+  out.append(buffer);
+  out.append("\n");
   out.append("}");
 }
 #endif
@@ -6111,6 +6147,155 @@ void BluetoothDeviceClearCacheResponse::dump_to(std::string &out) const {
   out.append("}");
 }
 #endif
+bool SubscribeVoiceAssistantRequest::decode_varint(uint32_t field_id, ProtoVarInt value) {
+  switch (field_id) {
+    case 1: {
+      this->subscribe = value.as_bool();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void SubscribeVoiceAssistantRequest::encode(ProtoWriteBuffer buffer) const { buffer.encode_bool(1, this->subscribe); }
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void SubscribeVoiceAssistantRequest::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("SubscribeVoiceAssistantRequest {\n");
+  out.append("  subscribe: ");
+  out.append(YESNO(this->subscribe));
+  out.append("\n");
+  out.append("}");
+}
+#endif
+bool VoiceAssistantRequest::decode_varint(uint32_t field_id, ProtoVarInt value) {
+  switch (field_id) {
+    case 1: {
+      this->start = value.as_bool();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantRequest::encode(ProtoWriteBuffer buffer) const { buffer.encode_bool(1, this->start); }
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantRequest::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantRequest {\n");
+  out.append("  start: ");
+  out.append(YESNO(this->start));
+  out.append("\n");
+  out.append("}");
+}
+#endif
+bool VoiceAssistantResponse::decode_varint(uint32_t field_id, ProtoVarInt value) {
+  switch (field_id) {
+    case 1: {
+      this->port = value.as_uint32();
+      return true;
+    }
+    case 2: {
+      this->error = value.as_bool();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantResponse::encode(ProtoWriteBuffer buffer) const {
+  buffer.encode_uint32(1, this->port);
+  buffer.encode_bool(2, this->error);
+}
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantResponse::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantResponse {\n");
+  out.append("  port: ");
+  sprintf(buffer, "%u", this->port);
+  out.append(buffer);
+  out.append("\n");
+
+  out.append("  error: ");
+  out.append(YESNO(this->error));
+  out.append("\n");
+  out.append("}");
+}
+#endif
+bool VoiceAssistantEventData::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
+  switch (field_id) {
+    case 1: {
+      this->name = value.as_string();
+      return true;
+    }
+    case 2: {
+      this->value = value.as_string();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantEventData::encode(ProtoWriteBuffer buffer) const {
+  buffer.encode_string(1, this->name);
+  buffer.encode_string(2, this->value);
+}
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantEventData::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantEventData {\n");
+  out.append("  name: ");
+  out.append("'").append(this->name).append("'");
+  out.append("\n");
+
+  out.append("  value: ");
+  out.append("'").append(this->value).append("'");
+  out.append("\n");
+  out.append("}");
+}
+#endif
+bool VoiceAssistantEventResponse::decode_varint(uint32_t field_id, ProtoVarInt value) {
+  switch (field_id) {
+    case 1: {
+      this->event_type = value.as_enum<enums::VoiceAssistantEvent>();
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+bool VoiceAssistantEventResponse::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
+  switch (field_id) {
+    case 2: {
+      this->data.push_back(value.as_message<VoiceAssistantEventData>());
+      return true;
+    }
+    default:
+      return false;
+  }
+}
+void VoiceAssistantEventResponse::encode(ProtoWriteBuffer buffer) const {
+  buffer.encode_enum<enums::VoiceAssistantEvent>(1, this->event_type);
+  for (auto &it : this->data) {
+    buffer.encode_message<VoiceAssistantEventData>(2, it, true);
+  }
+}
+#ifdef HAS_PROTO_MESSAGE_DUMP
+void VoiceAssistantEventResponse::dump_to(std::string &out) const {
+  __attribute__((unused)) char buffer[64];
+  out.append("VoiceAssistantEventResponse {\n");
+  out.append("  event_type: ");
+  out.append(proto_enum_to_string<enums::VoiceAssistantEvent>(this->event_type));
+  out.append("\n");
+
+  for (const auto &it : this->data) {
+    out.append("  data: ");
+    it.dump_to(out);
+    out.append("\n");
+  }
+  out.append("}");
+}
+#endif
 
 }  // namespace api
 }  // namespace esphome

esphome/components/api/api_pb2.h

@@ -165,6 +165,17 @@ enum BluetoothDeviceRequestType : uint32_t {
   BLUETOOTH_DEVICE_REQUEST_TYPE_CONNECT_V3_WITHOUT_CACHE = 5,
   BLUETOOTH_DEVICE_REQUEST_TYPE_CLEAR_CACHE = 6,
 };
+enum VoiceAssistantEvent : uint32_t {
+  VOICE_ASSISTANT_ERROR = 0,
+  VOICE_ASSISTANT_RUN_START = 1,
+  VOICE_ASSISTANT_RUN_END = 2,
+  VOICE_ASSISTANT_STT_START = 3,
+  VOICE_ASSISTANT_STT_END = 4,
+  VOICE_ASSISTANT_INTENT_START = 5,
+  VOICE_ASSISTANT_INTENT_END = 6,
+  VOICE_ASSISTANT_TTS_START = 7,
+  VOICE_ASSISTANT_TTS_END = 8,
+};
 
 }  // namespace enums
 
@@ -279,6 +290,7 @@ class DeviceInfoResponse : public ProtoMessage {
   uint32_t bluetooth_proxy_version{0};
   std::string manufacturer{};
   std::string friendly_name{};
+  uint32_t voice_assistant_version{0};
   void encode(ProtoWriteBuffer buffer) const override;
 #ifdef HAS_PROTO_MESSAGE_DUMP
   void dump_to(std::string &out) const override;
@@ -1577,6 +1589,65 @@ class BluetoothDeviceClearCacheResponse : public ProtoMessage {
  protected:
   bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
 };
+class SubscribeVoiceAssistantRequest : public ProtoMessage {
+ public:
+  bool subscribe{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 VoiceAssistantRequest : public ProtoMessage {
+ public:
+  bool start{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 VoiceAssistantResponse : public ProtoMessage {
+ public:
+  uint32_t port{0};
+  bool error{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 VoiceAssistantEventData : public ProtoMessage {
+ public:
+  std::string name{};
+  std::string value{};
+  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 VoiceAssistantEventResponse : public ProtoMessage {
+ public:
+  enums::VoiceAssistantEvent event_type{};
+  std::vector<VoiceAssistantEventData> data{};
+  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;
+  bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
+};
 
 }  // namespace api
 }  // namespace esphome

esphome/components/api/api_pb2_service.cpp

@@ -453,6 +453,20 @@ bool APIServerConnectionBase::send_bluetooth_device_clear_cache_response(const B
   return this->send_message_<BluetoothDeviceClearCacheResponse>(msg, 88);
 }
 #endif
+#ifdef USE_VOICE_ASSISTANT
+#endif
+#ifdef USE_VOICE_ASSISTANT
+bool APIServerConnectionBase::send_voice_assistant_request(const VoiceAssistantRequest &msg) {
+#ifdef HAS_PROTO_MESSAGE_DUMP
+  ESP_LOGVV(TAG, "send_voice_assistant_request: %s", msg.dump().c_str());
+#endif
+  return this->send_message_<VoiceAssistantRequest>(msg, 90);
+}
+#endif
+#ifdef USE_VOICE_ASSISTANT
+#endif
+#ifdef USE_VOICE_ASSISTANT
+#endif
 bool APIServerConnectionBase::read_message(uint32_t msg_size, uint32_t msg_type, uint8_t *msg_data) {
   switch (msg_type) {
     case 1: {
@@ -827,6 +841,39 @@ bool APIServerConnectionBase::read_message(uint32_t msg_size, uint32_t msg_type,
       ESP_LOGVV(TAG, "on_unsubscribe_bluetooth_le_advertisements_request: %s", msg.dump().c_str());
 #endif
       this->on_unsubscribe_bluetooth_le_advertisements_request(msg);
+#endif
+      break;
+    }
+    case 89: {
+#ifdef USE_VOICE_ASSISTANT
+      SubscribeVoiceAssistantRequest msg;
+      msg.decode(msg_data, msg_size);
+#ifdef HAS_PROTO_MESSAGE_DUMP
+      ESP_LOGVV(TAG, "on_subscribe_voice_assistant_request: %s", msg.dump().c_str());
+#endif
+      this->on_subscribe_voice_assistant_request(msg);
+#endif
+      break;
+    }
+    case 91: {
+#ifdef USE_VOICE_ASSISTANT
+      VoiceAssistantResponse msg;
+      msg.decode(msg_data, msg_size);
+#ifdef HAS_PROTO_MESSAGE_DUMP
+      ESP_LOGVV(TAG, "on_voice_assistant_response: %s", msg.dump().c_str());
+#endif
+      this->on_voice_assistant_response(msg);
+#endif
+      break;
+    }
+    case 92: {
+#ifdef USE_VOICE_ASSISTANT
+      VoiceAssistantEventResponse msg;
+      msg.decode(msg_data, msg_size);
+#ifdef HAS_PROTO_MESSAGE_DUMP
+      ESP_LOGVV(TAG, "on_voice_assistant_event_response: %s", msg.dump().c_str());
+#endif
+      this->on_voice_assistant_event_response(msg);
 #endif
       break;
     }
@@ -1226,6 +1273,19 @@ void APIServerConnection::on_unsubscribe_bluetooth_le_advertisements_request(
   this->unsubscribe_bluetooth_le_advertisements(msg);
 }
 #endif
+#ifdef USE_VOICE_ASSISTANT
+void APIServerConnection::on_subscribe_voice_assistant_request(const SubscribeVoiceAssistantRequest &msg) {
+  if (!this->is_connection_setup()) {
+    this->on_no_setup_connection();
+    return;
+  }
+  if (!this->is_authenticated()) {
+    this->on_unauthenticated_access();
+    return;
+  }
+  this->subscribe_voice_assistant(msg);
+}
+#endif
 
 }  // namespace api
 }  // namespace esphome

esphome/components/api/api_pb2_service.h

@@ -224,6 +224,18 @@ class APIServerConnectionBase : public ProtoService {
 #endif
 #ifdef USE_BLUETOOTH_PROXY
   bool send_bluetooth_device_clear_cache_response(const BluetoothDeviceClearCacheResponse &msg);
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  virtual void on_subscribe_voice_assistant_request(const SubscribeVoiceAssistantRequest &value){};
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  bool send_voice_assistant_request(const VoiceAssistantRequest &msg);
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  virtual void on_voice_assistant_response(const VoiceAssistantResponse &value){};
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  virtual void on_voice_assistant_event_response(const VoiceAssistantEventResponse &value){};
 #endif
  protected:
   bool read_message(uint32_t msg_size, uint32_t msg_type, uint8_t *msg_data) override;
@@ -306,6 +318,9 @@ class APIServerConnection : public APIServerConnectionBase {
 #endif
 #ifdef USE_BLUETOOTH_PROXY
   virtual void unsubscribe_bluetooth_le_advertisements(const UnsubscribeBluetoothLEAdvertisementsRequest &msg) = 0;
+#endif
+#ifdef USE_VOICE_ASSISTANT
+  virtual void subscribe_voice_assistant(const SubscribeVoiceAssistantRequest &msg) = 0;
 #endif
  protected:
   void on_hello_request(const HelloRequest &msg) override;
@@ -384,6 +399,9 @@ class APIServerConnection : public APIServerConnectionBase {
   void on_unsubscribe_bluetooth_le_advertisements_request(
       const UnsubscribeBluetoothLEAdvertisementsRequest &msg) override;
 #endif
+#ifdef USE_VOICE_ASSISTANT
+  void on_subscribe_voice_assistant_request(const SubscribeVoiceAssistantRequest &msg) override;
+#endif
 };
 
 }  // namespace api

esphome/components/api/api_server.cpp

@@ -427,5 +427,18 @@ void APIServer::on_shutdown() {
   delay(10);
 }
 
+#ifdef USE_VOICE_ASSISTANT
+void APIServer::start_voice_assistant() {
+  for (auto &c : this->clients_) {
+    c->request_voice_assistant(true);
+  }
+}
+void APIServer::stop_voice_assistant() {
+  for (auto &c : this->clients_) {
+    c->request_voice_assistant(false);
+  }
+}
+#endif
+
 }  // namespace api
 }  // namespace esphome

esphome/components/api/api_server.h

@@ -95,6 +95,11 @@ class APIServer : public Component, public Controller {
   void request_time();
 #endif
 
+#ifdef USE_VOICE_ASSISTANT
+  void start_voice_assistant();
+  void stop_voice_assistant();
+#endif
+
   bool is_connected() const;
 
   struct HomeAssistantStateSubscription {