Add preannounce media id support for ESPHome (#141474)

* Working on preannounce media id support for ESPHome

* Fix test

* Update tests

homeassistant/components/esphome/assist_satellite.py

@@ -370,8 +370,10 @@ class EsphomeAssistSatellite(
             announcement.media_id,
         )
         media_id = announcement.media_id
-        if announcement.media_id_source != "tts":
+        is_media_tts = announcement.media_id_source == "tts"
-            # Route non-TTS media through the proxy
+        preannounce_media_id = announcement.preannounce_media_id
+        if (not is_media_tts) or preannounce_media_id:
+            # Route media through the proxy
             format_to_use: MediaPlayerSupportedFormat | None = None
             for supported_format in chain(
                 *self.entry_data.media_player_formats.values()
@@ -384,22 +386,33 @@ class EsphomeAssistSatellite(
                 assert (self.registry_entry is not None) and (
                     self.registry_entry.device_id is not None
                 )
-                proxy_url = async_create_proxy_url(
+
-                    self.hass,
+                make_proxy_url = partial(
-                    self.registry_entry.device_id,
+                    async_create_proxy_url,
-                    media_id,
+                    hass=self.hass,
+                    device_id=self.registry_entry.device_id,
                     media_format=format_to_use.format,
                     rate=format_to_use.sample_rate or None,
                     channels=format_to_use.num_channels or None,
                     width=format_to_use.sample_bytes or None,
                 )
-                media_id = async_process_play_media_url(self.hass, proxy_url)
+
+                if not is_media_tts:
+                    media_id = async_process_play_media_url(
+                        self.hass, make_proxy_url(media_url=media_id)
+                    )
+
+                if preannounce_media_id:
+                    preannounce_media_id = async_process_play_media_url(
+                        self.hass, make_proxy_url(media_url=preannounce_media_id)
+                    )
 
         await self.cli.send_voice_assistant_announcement_await_response(
             media_id,
             _ANNOUNCEMENT_TIMEOUT_SEC,
             announcement.message,
             start_conversation=run_pipeline_after,
+            preannounce_media_id=preannounce_media_id or "",
         )
 
     async def handle_pipeline_start(

tests/components/esphome/test_assist_satellite.py

@@ -1212,12 +1212,17 @@ async def test_announce_message(
     done = asyncio.Event()
 
     async def send_voice_assistant_announcement_await_response(
-        media_id: str, timeout: float, text: str, start_conversation: bool
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str | None = None,
     ):
         assert satellite.state == AssistSatelliteState.RESPONDING
         assert media_id == "http://10.10.10.10:8123/api/tts_proxy/test-token"
         assert text == "test-text"
         assert not start_conversation
+        assert not preannounce_media_id
 
         done.set()
 
@@ -1302,11 +1307,16 @@ async def test_announce_media_id(
     done = asyncio.Event()
 
     async def send_voice_assistant_announcement_await_response(
-        media_id: str, timeout: float, text: str, start_conversation: bool
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str | None = None,
     ):
         assert satellite.state == AssistSatelliteState.RESPONDING
         assert media_id == "https://www.home-assistant.io/proxied.flac"
         assert not start_conversation
+        assert not preannounce_media_id
 
         done.set()
 
@@ -1335,9 +1345,9 @@ async def test_announce_media_id(
             assert satellite.state == AssistSatelliteState.IDLE
 
         mock_async_create_proxy_url.assert_called_once_with(
-            hass,
+            hass=hass,
-            dev.id,
+            device_id=dev.id,
-            "https://www.home-assistant.io/resolved.mp3",
+            media_url="https://www.home-assistant.io/resolved.mp3",
             media_format="flac",
             rate=48000,
             channels=2,
@@ -1345,6 +1355,83 @@ async def test_announce_media_id(
         )
 
 
+async def test_announce_message_with_preannounce(
+    hass: HomeAssistant,
+    mock_client: APIClient,
+    mock_esphome_device: Callable[
+        [APIClient, list[EntityInfo], list[UserService], list[EntityState]],
+        Awaitable[MockESPHomeDevice],
+    ],
+) -> None:
+    """Test announcement with message and preannounce media id."""
+    mock_device: MockESPHomeDevice = await mock_esphome_device(
+        mock_client=mock_client,
+        entity_info=[],
+        user_service=[],
+        states=[],
+        device_info={
+            "voice_assistant_feature_flags": VoiceAssistantFeature.VOICE_ASSISTANT
+            | VoiceAssistantFeature.SPEAKER
+            | VoiceAssistantFeature.API_AUDIO
+            | VoiceAssistantFeature.ANNOUNCE
+        },
+    )
+    await hass.async_block_till_done()
+
+    satellite = get_satellite_entity(hass, mock_device.device_info.mac_address)
+    assert satellite is not None
+
+    done = asyncio.Event()
+
+    async def send_voice_assistant_announcement_await_response(
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str | None = None,
+    ):
+        assert satellite.state == AssistSatelliteState.RESPONDING
+        assert media_id == "http://10.10.10.10:8123/api/tts_proxy/test-token"
+        assert text == "test-text"
+        assert not start_conversation
+        assert preannounce_media_id == "test-preannounce"
+
+        done.set()
+
+    with (
+        patch(
+            "homeassistant.components.tts.generate_media_source_id",
+            return_value="media-source://bla",
+        ),
+        patch(
+            "homeassistant.components.tts.async_resolve_engine",
+            return_value="tts.cloud_tts",
+        ),
+        patch(
+            "homeassistant.components.tts.async_create_stream",
+            return_value=MockResultStream(hass, "wav", b""),
+        ),
+        patch.object(
+            mock_client,
+            "send_voice_assistant_announcement_await_response",
+            new=send_voice_assistant_announcement_await_response,
+        ),
+    ):
+        async with asyncio.timeout(1):
+            await hass.services.async_call(
+                assist_satellite.DOMAIN,
+                "announce",
+                {
+                    "entity_id": satellite.entity_id,
+                    "message": "test-text",
+                    "preannounce_media_id": "test-preannounce",
+                },
+                blocking=True,
+            )
+            await done.wait()
+            assert satellite.state == AssistSatelliteState.IDLE
+
+
 async def test_start_conversation_supported_features(
     hass: HomeAssistant,
     mock_client: APIClient,
@@ -1417,12 +1504,17 @@ async def test_start_conversation_message(
     done = asyncio.Event()
 
     async def send_voice_assistant_announcement_await_response(
-        media_id: str, timeout: float, text: str, start_conversation: bool
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str,
     ):
         assert satellite.state == AssistSatelliteState.RESPONDING
         assert media_id == "http://10.10.10.10:8123/api/tts_proxy/test-token"
         assert text == "test-text"
         assert start_conversation
+        assert not preannounce_media_id
 
         done.set()
 
@@ -1526,11 +1618,16 @@ async def test_start_conversation_media_id(
     done = asyncio.Event()
 
     async def send_voice_assistant_announcement_await_response(
-        media_id: str, timeout: float, text: str, start_conversation: bool
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str,
     ):
         assert satellite.state == AssistSatelliteState.RESPONDING
         assert media_id == "https://www.home-assistant.io/proxied.flac"
         assert start_conversation
+        assert not preannounce_media_id
 
         done.set()
 
@@ -1563,9 +1660,9 @@ async def test_start_conversation_media_id(
             assert satellite.state == AssistSatelliteState.IDLE
 
         mock_async_create_proxy_url.assert_called_once_with(
-            hass,
+            hass=hass,
-            dev.id,
+            device_id=dev.id,
-            "https://www.home-assistant.io/resolved.mp3",
+            media_url="https://www.home-assistant.io/resolved.mp3",
             media_format="flac",
             rate=48000,
             channels=2,
@@ -1573,6 +1670,102 @@ async def test_start_conversation_media_id(
         )
 
 
+async def test_start_conversation_message_with_preannounce(
+    hass: HomeAssistant,
+    mock_client: APIClient,
+    mock_esphome_device: Callable[
+        [APIClient, list[EntityInfo], list[UserService], list[EntityState]],
+        Awaitable[MockESPHomeDevice],
+    ],
+) -> None:
+    """Test start conversation with message and preannounce media id."""
+    mock_device: MockESPHomeDevice = await mock_esphome_device(
+        mock_client=mock_client,
+        entity_info=[],
+        user_service=[],
+        states=[],
+        device_info={
+            "voice_assistant_feature_flags": VoiceAssistantFeature.VOICE_ASSISTANT
+            | VoiceAssistantFeature.SPEAKER
+            | VoiceAssistantFeature.API_AUDIO
+            | VoiceAssistantFeature.ANNOUNCE
+            | VoiceAssistantFeature.START_CONVERSATION
+        },
+    )
+    await hass.async_block_till_done()
+
+    satellite = get_satellite_entity(hass, mock_device.device_info.mac_address)
+    assert satellite is not None
+
+    pipeline = assist_pipeline.Pipeline(
+        conversation_engine="test engine",
+        conversation_language="en",
+        language="en",
+        name="test pipeline",
+        stt_engine="test stt",
+        stt_language="en",
+        tts_engine="test tts",
+        tts_language="en",
+        tts_voice=None,
+        wake_word_entity=None,
+        wake_word_id=None,
+    )
+
+    done = asyncio.Event()
+
+    async def send_voice_assistant_announcement_await_response(
+        media_id: str,
+        timeout: float,
+        text: str,
+        start_conversation: bool,
+        preannounce_media_id: str,
+    ):
+        assert satellite.state == AssistSatelliteState.RESPONDING
+        assert media_id == "http://10.10.10.10:8123/api/tts_proxy/test-token"
+        assert text == "test-text"
+        assert start_conversation
+        assert preannounce_media_id == "test-preannounce"
+
+        done.set()
+
+    with (
+        patch(
+            "homeassistant.components.tts.generate_media_source_id",
+            return_value="media-source://bla",
+        ),
+        patch(
+            "homeassistant.components.tts.async_resolve_engine",
+            return_value="tts.cloud_tts",
+        ),
+        patch(
+            "homeassistant.components.tts.async_create_stream",
+            return_value=MockResultStream(hass, "wav", b""),
+        ),
+        patch.object(
+            mock_client,
+            "send_voice_assistant_announcement_await_response",
+            new=send_voice_assistant_announcement_await_response,
+        ),
+        patch(
+            "homeassistant.components.assist_satellite.entity.async_get_pipeline",
+            return_value=pipeline,
+        ),
+    ):
+        async with asyncio.timeout(1):
+            await hass.services.async_call(
+                assist_satellite.DOMAIN,
+                "start_conversation",
+                {
+                    "entity_id": satellite.entity_id,
+                    "start_message": "test-text",
+                    "preannounce_media_id": "test-preannounce",
+                },
+                blocking=True,
+            )
+            await done.wait()
+            assert satellite.state == AssistSatelliteState.IDLE
+
+
 async def test_satellite_unloaded_on_disconnect(
     hass: HomeAssistant,
     mock_client: APIClient,