async def speak_sentence(self,
sentence: str,
language: typing.Optional[str] = None):
"""Speak a sentence using text to speech."""
tts_id = str(uuid4())
def handle_finished():
while True:
_, message = yield
if isinstance(message, TtsSayFinished) and (message.id
== tts_id):
return True, None
say = TtsSay(id=tts_id, text=sentence, siteId=self.siteId)
if language:
say.lang = language
messages = [say]
topics = [TtsSayFinished.topic()]
# Expecting only a single result
async for result in self.publish_wait(handle_finished(), messages,
topics):
return result
def test_http_mqtt_text_to_speech(self):
"""Test text-to-speech HTTP endpoint"""
text = "This is a test."
self.client.subscribe(TtsSay.topic())
self.client.subscribe(AudioPlayBytes.topic(site_id=self.site_id))
self.client.subscribe(TtsSayFinished.topic())
response = requests.post(
self.api_url("text-to-speech"),
data=text,
params={
"siteId": self.site_id,
"sessionId": self.session_id
},
)
self.check_status(response)
wav_data = response.content
self.assertGreater(len(wav_data), 0)
# Check tts/say
tts_say_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(TtsSay.is_topic(tts_say_msg.topic))
tts_say = TtsSay.from_dict(json.loads(tts_say_msg.payload))
self.assertEqual(tts_say.site_id, self.site_id)
self.assertEqual(tts_say.session_id, self.session_id)
self.assertEqual(tts_say.text, text)
# Check audioServer/playBytes
play_bytes_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(AudioPlayBytes.is_topic(play_bytes_msg.topic))
self.assertEqual(AudioPlayBytes.get_site_id(play_bytes_msg.topic),
self.site_id)
self.assertEqual(play_bytes_msg.payload, wav_data)
# Check tts/sayFinished
tts_finished_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(TtsSayFinished.is_topic(tts_finished_msg.topic))
tts_finished = TtsSayFinished.from_dict(
json.loads(tts_finished_msg.payload))
self.assertEqual(tts_finished.session_id, self.session_id)
# Ask for repeat
response = requests.post(self.api_url("text-to-speech"),
params={"repeat": "true"})
self.check_status(response)
self.assertEqual(wav_data, response.content)
def on_message(self, client, userdata, msg):
"""Received message from MQTT broker."""
try:
_LOGGER.debug("Received %s byte(s) on %s", len(msg.payload),
msg.topic)
if msg.topic == TtsSay.topic():
json_payload = json.loads(msg.payload or "{}")
if not self._check_siteId(json_payload):
return
say = TtsSay(**json_payload)
self.handle_say(say)
except Exception:
_LOGGER.exception("on_message")
async def handle_intent(
self, nlu_intent: NluIntent) -> typing.AsyncIterable[TtsSay]:
"""Handle intent with Home Assistant."""
try:
if self.handle_type == HandleType.EVENT:
await self.handle_home_assistant_event(nlu_intent)
elif self.handle_type == HandleType.INTENT:
response_dict = await self.handle_home_assistant_intent(
nlu_intent)
assert response_dict, f"No response from {self.url}"
# Check for speech response
tts_text = (response_dict.get("speech",
{}).get("plain",
{}).get("speech", ""))
if tts_text:
# Forward to TTS system
yield TtsSay(
text=tts_text,
id=str(uuid4()),
site_id=nlu_intent.site_id,
session_id=nlu_intent.session_id,
)
else:
raise ValueError(
f"Unsupported handle_type (got {self.handle_type})")
except Exception:
_LOGGER.exception("handle_intent")
async def say(
self,
text: str,
site_id="default",
session_id="",
request_id: typing.Optional[str] = None,
block: bool = True,
) -> typing.AsyncIterable[
typing.Union[
TtsSay, HotwordToggleOn, HotwordToggleOff, AsrToggleOn, AsrToggleOff
]
]:
"""Send text to TTS system and wait for reply."""
finished_event = asyncio.Event()
finished_id = request_id or str(uuid4())
self.message_events[TtsSayFinished][finished_id] = finished_event
# Disable ASR/hotword at site
yield HotwordToggleOff(site_id=site_id, reason=HotwordToggleReason.TTS_SAY)
yield AsrToggleOff(site_id=site_id, reason=AsrToggleReason.TTS_SAY)
# Wait for messages to be delivered
await asyncio.sleep(self.toggle_delay)
try:
# Forward to TTS
_LOGGER.debug("Say: %s", text)
yield TtsSay(
id=finished_id, site_id=site_id, session_id=session_id, text=text
)
if block:
# Wait for finished event
say_finished_timeout = 10.0
if self.say_chars_per_second > 0:
# Estimate timeout based on text length
say_finished_timeout = max(
say_finished_timeout, len(text) / self.say_chars_per_second
)
_LOGGER.debug(
"Waiting for sayFinished (id=%s, timeout=%s)",
finished_id,
say_finished_timeout,
)
await asyncio.wait_for(
finished_event.wait(), timeout=say_finished_timeout
)
except asyncio.TimeoutError:
_LOGGER.warning("Did not receive sayFinished before timeout")
except Exception:
_LOGGER.exception("say")
finally:
# Wait for audio to finish play
await asyncio.sleep(self.toggle_delay)
# Re-enable ASR/hotword at site
yield HotwordToggleOn(site_id=site_id, reason=HotwordToggleReason.TTS_SAY)
yield AsrToggleOn(site_id=site_id, reason=AsrToggleReason.TTS_SAY)
def on_connect(self, client, userdata, flags, rc):
"""Connected to MQTT broker."""
try:
topics = [TtsSay.topic()]
for topic in topics:
self.client.subscribe(topic)
_LOGGER.debug("Subscribed to %s", topic)
except Exception:
_LOGGER.exception("on_connect")
async def say_and_wait(self, text: str) -> typing.Iterable[TtsSay]:
"""Send text to TTS system and wait for reply."""
assert self.session, "No session"
self.say_finished_event.clear()
# Forward to TTS
_LOGGER.debug("Say: %s", text)
yield TtsSay(siteId=self.siteId,
sessionId=self.session.sessionId,
text=text)
# Wait for finished response (with timeout)
try:
await asyncio.wait_for(self.say_finished_event.wait(),
timeout=self.say_finished_timeout)
except asyncio.TimeoutError:
_LOGGER.exception("say_and_wait")
def test_no_play(self):
"""Test text-to-speech HTTP endpoint with play=false"""
text = "This is a test."
self.client.subscribe(TtsSay.topic())
self.client.subscribe(AudioPlayBytes.topic(site_id=self.site_id))
self.client.subscribe(TtsSayFinished.topic())
self.client.subscribe(AudioToggleOff.topic())
self.client.subscribe(AudioToggleOn.topic())
response = requests.post(
self.api_url("text-to-speech"),
data=text,
params={
"siteId": self.site_id,
"sessionId": self.session_id,
"play": "false",
},
)
self.check_status(response)
wav_data = response.content
self.assertGreater(len(wav_data), 0)
# Check audioServer/toggleOff
audio_off_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(AudioToggleOff.is_topic(audio_off_msg.topic))
audio_off = AudioToggleOff.from_dict(json.loads(audio_off_msg.payload))
self.assertEqual(audio_off.site_id, self.site_id)
# Check tts/say
tts_say_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(TtsSay.is_topic(tts_say_msg.topic))
tts_say = TtsSay.from_dict(json.loads(tts_say_msg.payload))
self.assertEqual(tts_say.site_id, self.site_id)
self.assertEqual(tts_say.session_id, self.session_id)
self.assertEqual(tts_say.text, text)
# Check audioServer/playBytes (will be ignored by audio output system)
play_bytes_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(AudioPlayBytes.is_topic(play_bytes_msg.topic))
self.assertEqual(AudioPlayBytes.get_site_id(play_bytes_msg.topic),
self.site_id)
self.assertEqual(play_bytes_msg.payload, wav_data)
# Check tts/sayFinished
tts_finished_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(TtsSayFinished.is_topic(tts_finished_msg.topic))
tts_finished = TtsSayFinished.from_dict(
json.loads(tts_finished_msg.payload))
self.assertEqual(tts_finished.site_id, self.site_id)
self.assertEqual(tts_finished.session_id, self.session_id)
# Check audioServer/toggleOn
audio_on_msg = self.mqtt_messages.get(timeout=5)
self.assertTrue(AudioToggleOn.is_topic(audio_on_msg.topic))
audio_on = AudioToggleOn.from_dict(json.loads(audio_on_msg.payload))
self.assertEqual(audio_on.site_id, self.site_id)
async def speak_sentence(
self,
sentence: str,
language: typing.Optional[str] = None,
capture_audio: bool = False,
wait_play_finished: bool = True,
site_id: typing.Optional[str] = None,
session_id: str = "",
) -> typing.Tuple[TtsSayFinished, typing.Optional[AudioPlayBytes]]:
"""Speak a sentence using text to speech."""
if (self.sound_system == "dummy") and (
not self.satellite_site_ids["text_to_speech"]
):
raise TtsException("No text to speech system configured")
site_id = site_id or self.site_id
tts_id = str(uuid4())
def handle_finished():
say_finished: typing.Optional[TtsSayFinished] = None
play_bytes: typing.Optional[
AudioPlayBytes
] = None if capture_audio else True
play_finished = not wait_play_finished
while True:
topic, message = yield
if isinstance(message, TtsSayFinished) and (message.id == tts_id):
say_finished = message
play_finished = True
elif isinstance(message, TtsError):
# Assume audio playback didn't happen
say_finished = message
play_bytes = True
play_finished = True
elif isinstance(message, AudioPlayBytes):
request_id = AudioPlayBytes.get_request_id(topic)
if request_id == tts_id:
play_bytes = message
elif isinstance(message, AudioPlayError):
play_bytes = message
if say_finished and play_bytes and play_finished:
return (say_finished, play_bytes)
say = TtsSay(id=tts_id, text=sentence, site_id=site_id, session_id=session_id)
if language:
say.lang = language
messages = [say]
message_types: typing.List[typing.Type[Message]] = [
TtsSayFinished,
TtsError,
AudioPlayBytes,
AudioPlayError,
]
# Expecting only a single result
result = None
async for response in self.publish_wait(
handle_finished(), messages, message_types
):
result = response
assert isinstance(result, tuple), f"Expected tuple, got {result}"
say_response, play_response = result
if isinstance(say_response, TtsError):
_LOGGER.error(say_response)
raise TtsException(say_response.error)
assert isinstance(say_response, TtsSayFinished), say_response
if isinstance(play_response, AudioPlayError):
_LOGGER.error(play_response)
raise AudioServerException(play_response.error)
if capture_audio:
assert isinstance(play_response, AudioPlayBytes), play_response
return typing.cast(
typing.Tuple[TtsSayFinished, typing.Optional[AudioPlayBytes]], result
)
def test_tts_say():
"""Test TtsSay."""
assert TtsSay.topic() == "hermes/tts/say"
async def handle_intent(
self, intent: NluIntent
) -> typing.AsyncIterable[typing.Union[TtsSay]]:
"""Handle intent with remote server or local command."""
try:
if not self.handle_enabled:
_LOGGER.debug("Intent handling is disabled")
return
tts_text = ""
intent_dict = intent.to_rhasspy_dict()
# Add site_id
intent_dict["site_id"] = intent.site_id
if self.handle_url:
# Remote server
_LOGGER.debug(self.handle_url)
async with self.http_session.post(
self.handle_url, json=intent_dict, ssl=self.ssl_context
) as response:
response.raise_for_status()
response_dict = await response.json()
# Check for speech response
tts_text = response_dict.get("speech", {}).get("text", "")
elif self.handle_command:
intent_json = json.dumps(intent_dict)
# Local handling command
_LOGGER.debug(self.handle_command)
proc = await asyncio.create_subprocess_exec(
*self.handle_command,
stdin=asyncio.subprocess.PIPE,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
output, error = await proc.communicate(intent_json.encode())
if error:
_LOGGER.debug(error.decode())
try:
response_dict = json.loads(output)
# Check for speech response
tts_text = response_dict.get("speech", {}).get("text", "")
except json.JSONDecodeError as e:
if output:
# Only report error if non-empty output
_LOGGER.warning("Failed to parse output as JSON: %s", e)
_LOGGER.warning("Output: %s", output)
else:
_LOGGER.warning("Can't handle intent. No handle URL or command.")
if tts_text:
# Forward to TTS system
yield TtsSay(
text=tts_text,
id=str(uuid4()),
site_id=intent.site_id,
session_id=intent.session_id,
)
except Exception:
_LOGGER.exception("handle_intent")
