Add initial voice assistant support (#412)

aioesphomeapi/api.proto

@@ -55,6 +55,8 @@ service APIConnection {
   rpc bluetooth_gatt_notify(BluetoothGATTNotifyRequest) returns (void) {}
   rpc unsubscribe_bluetooth_le_advertisements(UnsubscribeBluetoothLEAdvertisementsRequest) returns (void) {}
 
+  rpc subscribe_voice_assistant(SubscribeVoiceAssistantRequest) returns (void) {}
+
 }
 
 
@@ -210,6 +212,8 @@ message DeviceInfoResponse {
   string manufacturer = 12;
 
   string friendly_name = 13;
+
+  uint32 voice_assistant_version = 14;
 }
 
 message ListEntitiesRequest {
@@ -1403,3 +1407,55 @@ message BluetoothDeviceClearCacheResponse {
   bool success = 2;
   int32 error = 3;
 }
+
+// ==================== VOICE ASSISTANT ====================
+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;
+}

aioesphomeapi/client.py

@@ -93,10 +93,15 @@ from .api_pb2 import (  # type: ignore
     SubscribeLogsRequest,
     SubscribeLogsResponse,
     SubscribeStatesRequest,
+    SubscribeVoiceAssistantRequest,
     SwitchCommandRequest,
     SwitchStateResponse,
     TextSensorStateResponse,
     UnsubscribeBluetoothLEAdvertisementsRequest,
+    VoiceAssistantEventData,
+    VoiceAssistantEventResponse,
+    VoiceAssistantRequest,
+    VoiceAssistantResponse,
 )
 from .connection import APIConnection, ConnectionParams
 from .core import (
@@ -164,6 +169,8 @@ from .model import (
     TextSensorState,
     UserService,
     UserServiceArgType,
+    VoiceAssistantCommand,
+    VoiceAssistantEventType,
 )
 
 _LOGGER = logging.getLogger(__name__)
@@ -1238,3 +1245,81 @@ class APIClient:
         if self._connection is None:
             return None
         return self._connection.api_version
+
+    async def subscribe_voice_assistant(
+        self,
+        handle_start: Callable[[], Coroutine[Any, Any, Optional[int]]],
+        handle_stop: Callable[[], Coroutine[Any, Any, None]],
+    ) -> Callable[[], None]:
+        """Subscribes to voice assistant messages from the device.
+
+        handle_start: called when the devices requests a server to send audio data to.
+                      This callback is asyncronous and returns the port number the server is started on.
+
+        handle_stop: called when the device has stopped sending audio data and the pipeline should be closed.
+
+        Returns a callback to unsubscribe.
+        """
+        self._check_authenticated()
+
+        t: Optional[asyncio.Task[Optional[int]]] = None
+
+        def _started(fut: asyncio.Task[Optional[int]]) -> None:
+            if self._connection is not None and not fut.cancelled():
+                port = fut.result()
+                if port is not None:
+                    self._connection.send_message(VoiceAssistantResponse(port=port))
+                else:
+                    _LOGGER.error("Server could not be started")
+                    self._connection.send_message(VoiceAssistantResponse(error=True))
+
+        def on_msg(msg: VoiceAssistantRequest) -> None:
+            command = VoiceAssistantCommand.from_pb(msg)
+            loop = asyncio.get_running_loop()
+            if command.start:
+                t = loop.create_task(handle_start())
+                t.add_done_callback(_started)
+            else:
+                loop.create_task(handle_stop())
+
+        assert self._connection is not None
+
+        self._connection.send_message(SubscribeVoiceAssistantRequest(subscribe=True))
+
+        remove_callback = self._connection.add_message_callback(
+            on_msg, (VoiceAssistantRequest,)
+        )
+
+        def unsub() -> None:
+            if self._connection is not None:
+                remove_callback()
+                self._connection.send_message(
+                    SubscribeVoiceAssistantRequest(subscribe=False)
+                )
+
+            if t is not None and not t.cancelled():
+                t.cancel()
+
+        return unsub
+
+    def send_voice_assistant_event(
+        self, event_type: VoiceAssistantEventType, data: Optional[dict[str, str]]
+    ) -> None:
+        self._check_authenticated()
+
+        req = VoiceAssistantEventResponse()
+        req.event_type = event_type
+
+        data_args = []
+        if data is not None:
+            for name, value in data.items():
+                arg = VoiceAssistantEventData()
+                arg.name = name
+                arg.value = value
+                data_args.append(arg)
+
+        # pylint: disable=no-member
+        req.data.extend(data_args)
+
+        assert self._connection is not None
+        self._connection.send_message(req)

aioesphomeapi/core.py

@@ -87,10 +87,14 @@ from .api_pb2 import (  # type: ignore
     SubscribeLogsRequest,
     SubscribeLogsResponse,
     SubscribeStatesRequest,
+    SubscribeVoiceAssistantRequest,
     SwitchCommandRequest,
     SwitchStateResponse,
     TextSensorStateResponse,
     UnsubscribeBluetoothLEAdvertisementsRequest,
+    VoiceAssistantEventResponse,
+    VoiceAssistantRequest,
+    VoiceAssistantResponse,
 )
 
 TWO_CHAR = re.compile(r".{2}")
@@ -310,4 +314,8 @@ MESSAGE_TYPE_TO_PROTO = {
     86: BluetoothDeviceUnpairingResponse,
     87: UnsubscribeBluetoothLEAdvertisementsRequest,
     88: BluetoothDeviceClearCacheResponse,
+    89: SubscribeVoiceAssistantRequest,
+    90: VoiceAssistantRequest,
+    91: VoiceAssistantResponse,
+    92: VoiceAssistantEventResponse,
 }

aioesphomeapi/model.py

@@ -119,6 +119,7 @@ class DeviceInfo(APIModelBase):
     project_version: str = ""
     webserver_port: int = 0
     bluetooth_proxy_version: int = 0
+    voice_assistant_version: int = 0
 
 
 class EntityCategory(APIIntEnum):
@@ -1005,6 +1006,11 @@ class BluetoothDeviceRequestType(APIIntEnum):
     CLEAR_CACHE = 6
 
 
+@dataclass(frozen=True)
+class VoiceAssistantCommand(APIModelBase):
+    start: bool = False
+
+
 class LogLevel(APIIntEnum):
     LOG_LEVEL_NONE = 0
     LOG_LEVEL_ERROR = 1
@@ -1014,3 +1020,15 @@ class LogLevel(APIIntEnum):
     LOG_LEVEL_DEBUG = 5
     LOG_LEVEL_VERBOSE = 6
     LOG_LEVEL_VERY_VERBOSE = 7
+
+
+class VoiceAssistantEventType(APIIntEnum):
+    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