Google Assistant SDK: support audio response playback

homeassistant/components/google_assistant_sdk/__init__.py

@@ -11,23 +11,35 @@
 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_AUDIO_VIEW,
+    DATA_SESSION,
+    DOMAIN,
+)
+from .helpers import (
+    GoogleAssistantSDKAudioView,
+    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 +57,8 @@ 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 +71,11 @@ 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
+
+    audio_view = GoogleAssistantSDKAudioView(hass)
+    hass.http.register_view(audio_view)
+    hass.data[DOMAIN][entry.entry_id][DATA_AUDIO_VIEW] = audio_view
 
     await async_setup_service(hass)
 
@@ -88,7 +106,10 @@ 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 +157,7 @@ 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_AUDIO_VIEW: Final = "audio_view"
+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 +1,36 @@
 """Helper classes for Google Assistant SDK integration."""
 from __future__ import annotations
 
+from http import HTTPStatus
 import logging
+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 .const import CONF_LANGUAGE_CODE, DOMAIN, SUPPORTED_LANGUAGE_CODES
+from .const import (
+    CONF_LANGUAGE_CODE,
+    DATA_AUDIO_VIEW,
+    DATA_SESSION,
+    DOMAIN,
+    SUPPORTED_LANGUAGE_CODES,
+)
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -28,12 +46,14 @@
 }
 
 
-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 +63,31 @@ 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:
+                audio_view: GoogleAssistantSDKAudioView = hass.data[DOMAIN][
+                    entry.entry_id
+                ][DATA_AUDIO_VIEW]
+                await hass.services.async_call(
+                    DOMAIN_MP,
+                    SERVICE_PLAY_MEDIA,
+                    {
+                        ATTR_ENTITY_ID: media_players,
+                        ATTR_MEDIA_CONTENT_ID: audio_view.memcache_store_and_get_url(
+                            audio_response
+                        ),
+                        ATTR_MEDIA_CONTENT_TYPE: MediaType.MUSIC,
+                        ATTR_MEDIA_ANNOUNCE: True,
+                    },
+                    blocking=True,
+                )
 
 
 def default_language_code(hass: HomeAssistant):
@@ -55,3 +96,37 @@ 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 GoogleAssistantSDKAudioView(HomeAssistantView):
+    """Google Assistant SDK view to serve audio responses."""
+
+    requires_auth = False
+    url = "/api/google_assistant_sdk/audio/{filename}"
+    name = "api:google_assistant_sdk:audio"
+
+    def __init__(self, hass: HomeAssistant) -> None:
+        """Initialize GoogleAssistantSDKView."""
+        self.hass: HomeAssistant = hass
+        self.mem_cache: dict[str, bytes] = {}
+
+    def memcache_store_and_get_url(self, audio: bytes) -> str:
+        """Write audio to memcache and return URL to serve it."""
+        filename: str = uuid.uuid1().hex
+        self.mem_cache[filename] = audio
+
+        def async_remove_from_mem() -> None:
+            """Cleanup memcache."""
+            self.mem_cache.pop(filename, None)
+
+        # Remove the entry from memcache 5 minutes later
+        self.hass.loop.call_later(5 * 60, async_remove_from_mem)
+
+        return self.url.format(filename=filename)
+
+    async def get(self, request: web.Request, filename: str) -> web.Response:
+        """Start a get request."""
+        audio = self.mem_cache.get(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

homeassistant/components/media_player/browse_media.py

@@ -22,7 +22,10 @@
 from .const import CONTENT_AUTH_EXPIRY_TIME, MediaClass, MediaType
 
 # Paths that we don't need to sign
-PATHS_WITHOUT_AUTH = ("/api/tts_proxy/",)
+PATHS_WITHOUT_AUTH = (
+    "/api/google_assistant_sdk/audio/",
+    "/api/tts_proxy/",
+)
 
 
 @callback

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

tests/components/google_assistant_sdk/test_init.py

@@ -13,9 +13,18 @@
 
 from .conftest import ComponentSetup, ExpectedCredentials
 
+from tests.common import async_mock_service
 from tests.test_util.aiohttp import AiohttpClientMocker
 
 
+async def fetch_api_url(hass_client, url):
+    """Fetch an API URL and return HTTP status and contents."""
+    client = await hass_client()
+    response = await client.get(url)
+    contents = await response.read()
+    return response.status, contents
+
+
 async def test_setup_success(
     hass: HomeAssistant, setup_integration: ComponentSetup
 ) -> None:
@@ -129,7 +138,7 @@ async def test_send_text_command(
             blocking=True,
         )
     mock_text_assistant.assert_called_once_with(
-        ExpectedCredentials(), expected_language_code
+        ExpectedCredentials(), expected_language_code, audio_out=False
     )
     mock_text_assistant.assert_has_calls([call().__enter__().assist(command)])
 
@@ -180,6 +189,78 @@ async def test_send_text_command_expired_token_refresh_failure(
     assert any(entry.async_get_active_flows(hass, {"reauth"})) == requires_reauth
 
 
+async def test_send_text_command_media_player(
+    hass: HomeAssistant, setup_integration: ComponentSetup, hass_client
+) -> None:
+    """Test send_text_command with media_player."""
+    await setup_integration()
+
+    play_media_calls = async_mock_service(hass, "media_player", "play_media")
+
+    command = "tell me a joke"
+    media_player = "media_player.office_speaker"
+    audio_response1 = b"joke1 audio response bytes"
+    audio_response2 = b"joke2 audio response bytes"
+    with patch(
+        "homeassistant.components.google_assistant_sdk.helpers.TextAssistant.assist",
+        side_effect=[
+            ("joke1 text", None, audio_response1),
+            ("joke2 text", None, audio_response2),
+        ],
+    ) as mock_assist_call:
+        # Run the same command twice, getting different audio response each time.
+        await hass.services.async_call(
+            DOMAIN,
+            "send_text_command",
+            {
+                "command": command,
+                "media_player": media_player,
+            },
+            blocking=True,
+        )
+        await hass.services.async_call(
+            DOMAIN,
+            "send_text_command",
+            {
+                "command": command,
+                "media_player": media_player,
+            },
+            blocking=True,
+        )
+
+    mock_assist_call.assert_has_calls([call(command), call(command)])
+    assert len(play_media_calls) == 2
+    for play_media_call in play_media_calls:
+        assert play_media_call.data["entity_id"] == [media_player]
+        assert play_media_call.data["media_content_id"].startswith(
+            "/api/google_assistant_sdk/audio/"
+        )
+
+    audio_url1 = play_media_calls[0].data["media_content_id"]
+    audio_url2 = play_media_calls[1].data["media_content_id"]
+    assert audio_url1 != audio_url2
+
+    # Assert that both audio responses can be served multiple times.
+    status, response = await fetch_api_url(hass_client, audio_url1)
+    assert status == http.HTTPStatus.OK
+    assert response == audio_response1
+    status, response = await fetch_api_url(hass_client, audio_url1)
+    assert status == http.HTTPStatus.OK
+    assert response == audio_response1
+    status, response = await fetch_api_url(hass_client, audio_url2)
+    assert status == http.HTTPStatus.OK
+    assert response == audio_response2
+    status, response = await fetch_api_url(hass_client, audio_url2)
+    assert status == http.HTTPStatus.OK
+    assert response == audio_response2
+
+    # Assert a nonexistent URL returns 404
+    status, _ = await fetch_api_url(
+        hass_client, "/api/google_assistant_sdk/audio/nonexistent"
+    )
+    assert status == http.HTTPStatus.NOT_FOUND
+
+
 async def test_conversation_agent(
     hass: HomeAssistant,
     setup_integration: ComponentSetup,