Google Assistant SDK: support audio response playback

homeassistant/components/google_assistant_sdk/__init__.py

@@ -11,23 +11,36 @@
 from homeassistant.const import CONF_ACCESS_TOKEN, CONF_NAME, Platform
 from homeassistant.core import Context, HomeAssistant, ServiceCall
 from homeassistant.exceptions import ConfigEntryAuthFailed, ConfigEntryNotReady
-from homeassistant.helpers import discovery, intent
+from homeassistant.helpers import config_validation as cv, discovery, intent
 from homeassistant.helpers.config_entry_oauth2_flow import (
     OAuth2Session,
     async_get_config_entry_implementation,
 )
 from homeassistant.helpers.typing import ConfigType
 
-from .const import CONF_ENABLE_CONVERSATION_AGENT, CONF_LANGUAGE_CODE, DOMAIN
-from .helpers import async_send_text_commands, default_language_code
+from .const import (
+    CONF_ENABLE_CONVERSATION_AGENT,
+    CONF_LANGUAGE_CODE,
+    DATA_MEM_STORAGE,
+    DATA_SESSION,
+    DOMAIN,
+)
+from .helpers import (
+    GoogleAssistantSDKAudioView,
+    InMemoryStorage,
+    async_send_text_commands,
+    default_language_code,
+)
 
 SERVICE_SEND_TEXT_COMMAND = "send_text_command"
 SERVICE_SEND_TEXT_COMMAND_FIELD_COMMAND = "command"
+SERVICE_SEND_TEXT_COMMAND_FIELD_MEDIA_PLAYER = "media_player"
 SERVICE_SEND_TEXT_COMMAND_SCHEMA = vol.All(
     {
         vol.Required(SERVICE_SEND_TEXT_COMMAND_FIELD_COMMAND): vol.All(
             str, vol.Length(min=1)
         ),
+        vol.Optional(SERVICE_SEND_TEXT_COMMAND_FIELD_MEDIA_PLAYER): cv.comp_entity_ids,
     },
 )
 
@@ -45,6 +58,8 @@ async def async_setup(hass: HomeAssistant, config: ConfigType) -> bool:
 
 async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
     """Set up Google Assistant SDK from a config entry."""
+    hass.data.setdefault(DOMAIN, {})[entry.entry_id] = {}
+
     implementation = await async_get_config_entry_implementation(hass, entry)
     session = OAuth2Session(hass, entry, implementation)
     try:
@@ -57,7 +72,11 @@ async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
         raise ConfigEntryNotReady from err
     except aiohttp.ClientError as err:
         raise ConfigEntryNotReady from err
-    hass.data.setdefault(DOMAIN, {})[entry.entry_id] = session
+    hass.data[DOMAIN][entry.entry_id][DATA_SESSION] = session
+
+    mem_storage = InMemoryStorage(hass)
+    hass.data[DOMAIN][entry.entry_id][DATA_MEM_STORAGE] = mem_storage
+    hass.http.register_view(GoogleAssistantSDKAudioView(mem_storage))
 
     await async_setup_service(hass)
 
@@ -88,7 +107,10 @@ async def async_setup_service(hass: HomeAssistant) -> None:
     async def send_text_command(call: ServiceCall) -> None:
         """Send a text command to Google Assistant SDK."""
         command: str = call.data[SERVICE_SEND_TEXT_COMMAND_FIELD_COMMAND]
-        await async_send_text_commands([command], hass)
+        media_players: list[str] | None = call.data.get(
+            SERVICE_SEND_TEXT_COMMAND_FIELD_MEDIA_PLAYER
+        )
+        await async_send_text_commands(hass, [command], media_players)
 
     hass.services.async_register(
         DOMAIN,
@@ -136,7 +158,7 @@ async def async_process(
         if self.session:
             session = self.session
         else:
-            session = self.hass.data[DOMAIN].get(self.entry.entry_id)
+            session = self.hass.data[DOMAIN][self.entry.entry_id][DATA_SESSION]
             self.session = session
         if not session.valid_token:
             await session.async_ensure_token_valid()

homeassistant/components/google_assistant_sdk/const.py

@@ -5,8 +5,12 @@
 
 DEFAULT_NAME: Final = "Google Assistant SDK"
 
+CONF_ENABLE_CONVERSATION_AGENT: Final = "enable_conversation_agent"
 CONF_LANGUAGE_CODE: Final = "language_code"
 
+DATA_MEM_STORAGE: Final = "mem_storage"
+DATA_SESSION: Final = "session"
+
 # https://developers.google.com/assistant/sdk/reference/rpc/languages
 SUPPORTED_LANGUAGE_CODES: Final = [
     "de-DE",
@@ -24,5 +28,3 @@
     "ko-KR",
     "pt-BR",
 ]
-
-CONF_ENABLE_CONVERSATION_AGENT: Final = "enable_conversation_agent"

homeassistant/components/google_assistant_sdk/helpers.py

@@ -1,18 +1,38 @@
 """Helper classes for Google Assistant SDK integration."""
 from __future__ import annotations
 
+from http import HTTPStatus
 import logging
+from typing import Any
+import uuid
 
 import aiohttp
+from aiohttp import web
 from gassist_text import TextAssistant
 from google.oauth2.credentials import Credentials
 
+from homeassistant.components.http import HomeAssistantView
+from homeassistant.components.media_player import (
+    ATTR_MEDIA_ANNOUNCE,
+    ATTR_MEDIA_CONTENT_ID,
+    ATTR_MEDIA_CONTENT_TYPE,
+    DOMAIN as DOMAIN_MP,
+    SERVICE_PLAY_MEDIA,
+    MediaType,
+)
 from homeassistant.config_entries import ConfigEntry
-from homeassistant.const import CONF_ACCESS_TOKEN
+from homeassistant.const import ATTR_ENTITY_ID, CONF_ACCESS_TOKEN
 from homeassistant.core import HomeAssistant
 from homeassistant.helpers.config_entry_oauth2_flow import OAuth2Session
+from homeassistant.helpers.event import async_call_later
 
-from .const import CONF_LANGUAGE_CODE, DOMAIN, SUPPORTED_LANGUAGE_CODES
+from .const import (
+    CONF_LANGUAGE_CODE,
+    DATA_MEM_STORAGE,
+    DATA_SESSION,
+    DOMAIN,
+    SUPPORTED_LANGUAGE_CODES,
+)
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -28,12 +48,14 @@
 }
 
 
-async def async_send_text_commands(commands: list[str], hass: HomeAssistant) -> None:
+async def async_send_text_commands(
+    hass: HomeAssistant, commands: list[str], media_players: list[str] | None = None
+) -> None:
     """Send text commands to Google Assistant Service."""
     # There can only be 1 entry (config_flow has single_instance_allowed)
     entry: ConfigEntry = hass.config_entries.async_entries(DOMAIN)[0]
 
-    session: OAuth2Session = hass.data[DOMAIN].get(entry.entry_id)
+    session: OAuth2Session = hass.data[DOMAIN][entry.entry_id][DATA_SESSION]
     try:
         await session.async_ensure_token_valid()
     except aiohttp.ClientResponseError as err:
@@ -43,10 +65,32 @@ async def async_send_text_commands(commands: list[str], hass: HomeAssistant) ->
 
     credentials = Credentials(session.token[CONF_ACCESS_TOKEN])
     language_code = entry.options.get(CONF_LANGUAGE_CODE, default_language_code(hass))
-    with TextAssistant(credentials, language_code) as assistant:
+    with TextAssistant(
+        credentials, language_code, audio_out=bool(media_players)
+    ) as assistant:
         for command in commands:
-            text_response = assistant.assist(command)[0]
+            resp = assistant.assist(command)
+            text_response = resp[0]
             _LOGGER.debug("command: %s\nresponse: %s", command, text_response)
+            audio_response = resp[2]
+            if media_players and audio_response:
+                mem_storage: InMemoryStorage = hass.data[DOMAIN][entry.entry_id][
+                    DATA_MEM_STORAGE
+                ]
+                audio_url = GoogleAssistantSDKAudioView.url.format(
+                    filename=mem_storage.store_and_get_identifier(audio_response)
+                )
+                await hass.services.async_call(
+                    DOMAIN_MP,
+                    SERVICE_PLAY_MEDIA,
+                    {
+                        ATTR_ENTITY_ID: media_players,
+                        ATTR_MEDIA_CONTENT_ID: audio_url,
+                        ATTR_MEDIA_CONTENT_TYPE: MediaType.MUSIC,
+                        ATTR_MEDIA_ANNOUNCE: True,
+                    },
+                    blocking=True,
+                )
 
 
 def default_language_code(hass: HomeAssistant):
@@ -55,3 +99,53 @@ def default_language_code(hass: HomeAssistant):
     if language_code in SUPPORTED_LANGUAGE_CODES:
         return language_code
     return DEFAULT_LANGUAGE_CODES.get(hass.config.language, "en-US")
+
+
+class InMemoryStorage:
+    """Temporarily store and retrieve data from in memory storage."""
+
+    def __init__(self, hass: HomeAssistant) -> None:
+        """Initialize InMemoryStorage."""
+        self.hass: HomeAssistant = hass
+        self.mem: dict[str, bytes] = {}
+
+    def store_and_get_identifier(self, data: bytes) -> str:
+        """
+        Temporarily store data and return identifier to be able to retrieve it.
+
+        Data expires after 5 minutes.
+        """
+        identifier: str = uuid.uuid1().hex
+        self.mem[identifier] = data
+
+        def async_remove_from_mem(*_: Any) -> None:
+            """Cleanup memory."""
+            self.mem.pop(identifier, None)
+
+        # Remove the entry from memory 5 minutes later
+        async_call_later(self.hass, 5 * 60, async_remove_from_mem)
+
+        return identifier
+
+    def retrieve(self, identifier: str) -> bytes | None:
+        """Retrieve previously stored data."""
+        return self.mem.get(identifier)
+
+
+class GoogleAssistantSDKAudioView(HomeAssistantView):
+    """Google Assistant SDK view to serve audio responses."""
+
+    requires_auth = True
+    url = "/api/google_assistant_sdk/audio/{filename}"
+    name = "api:google_assistant_sdk:audio"
+
+    def __init__(self, mem_storage: InMemoryStorage) -> None:
+        """Initialize GoogleAssistantSDKView."""
+        self.mem_storage: InMemoryStorage = mem_storage
+
+    async def get(self, request: web.Request, filename: str) -> web.Response:
+        """Start a get request."""
+        audio = self.mem_storage.retrieve(filename)
+        if not audio:
+            return web.Response(status=HTTPStatus.NOT_FOUND)
+        return web.Response(body=audio, content_type="audio/mpeg")

homeassistant/components/google_assistant_sdk/manifest.json

@@ -2,9 +2,9 @@
   "domain": "google_assistant_sdk",
   "name": "Google Assistant SDK",
   "config_flow": true,
-  "dependencies": ["application_credentials"],
+  "dependencies": ["application_credentials", "http"],
   "documentation": "https://www.home-assistant.io/integrations/google_assistant_sdk/",
-  "requirements": ["gassist-text==0.0.7"],
+  "requirements": ["gassist-text==0.0.8"],
   "codeowners": ["@tronikos"],
   "iot_class": "cloud_polling",
   "integration_type": "service"

homeassistant/components/google_assistant_sdk/notify.py

@@ -70,4 +70,4 @@ async def async_send_message(self, message: str = "", **kwargs: Any) -> None:
                 commands.append(
                     broadcast_commands(language_code)[1].format(message, target)
                 )
-        await async_send_text_commands(commands, self.hass)
+        await async_send_text_commands(self.hass, commands)

homeassistant/components/google_assistant_sdk/services.yaml

@@ -8,3 +8,10 @@ send_text_command:
       example: turn off kitchen TV
       selector:
         text:
+    media_player:
+      name: Media Player Entity
+      description: Name(s) of media player entities to play response on
+      example: media_player.living_room_speaker
+      selector:
+        entity:
+          domain: media_player

requirements_all.txt

@@ -751,7 +751,7 @@ fritzconnection==1.10.3
 gTTS==2.2.4
 
 # homeassistant.components.google_assistant_sdk
-gassist-text==0.0.7
+gassist-text==0.0.8
 
 # homeassistant.components.google
 gcal-sync==4.1.2

requirements_test_all.txt

@@ -570,7 +570,7 @@ fritzconnection==1.10.3
 gTTS==2.2.4
 
 # homeassistant.components.google_assistant_sdk
-gassist-text==0.0.7
+gassist-text==0.0.8
 
 # homeassistant.components.google
 gcal-sync==4.1.2