livekit-plugins-neuphonic 1.0.23__py3-none-any.whl → 1.1.0__py3-none-any.whl

livekit/plugins/neuphonic/__init__.py

@@ -28,7 +28,7 @@ from .log import logger
 
 
 class NeuphonicPlugin(Plugin):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__(__name__, __version__, __package__, logger)
 
 

livekit/plugins/neuphonic/models.py

@@ -1,10 +1,3 @@
 from typing import Literal
 
-TTSEncodings = Literal[
-    "pcm_linear",
-    "pcm_mulaw",
-]
-
-TTSModels = Literal["neu-fast", "neu-hq"]
-
 TTSLangCodes = Literal["en", "nl", "es", "de", "hi", "en-hi", "ar"]

livekit/plugins/neuphonic/tts.py

@@ -18,8 +18,7 @@ import asyncio
 import base64
 import json
 import os
-import weakref
-from dataclasses import dataclass
+from dataclasses import dataclass, replace
 
 import aiohttp
 
@@ -34,81 +33,31 @@ from livekit.agents import (
 from livekit.agents.types import DEFAULT_API_CONNECT_OPTIONS, NOT_GIVEN, NotGivenOr
 from livekit.agents.utils import is_given
 
-from .log import logger
-from .models import TTSEncodings, TTSLangCodes, TTSModels
+from .models import TTSLangCodes
 
 API_BASE_URL = "api.neuphonic.com"
 AUTHORIZATION_HEADER = "X-API-KEY"
-NUM_CHANNELS = 1
 
 
 @dataclass
 class _TTSOptions:
     base_url: str
-    api_key: str
-    model: TTSModels | str
     lang_code: TTSLangCodes | str
-    encoding: TTSEncodings | str
-    sampling_rate: int
+    api_key: str
+    sample_rate: int
     speed: float
-    voice_id: NotGivenOr[str] = NOT_GIVEN
-
-    @property
-    def model_params(self) -> dict:
-        """Returns a dictionary of model parameters for API requests."""
-        params = {
-            "voice_id": self.voice_id,
-            "model": self.model,
-            "lang_code": self.lang_code,
-            "encoding": self.encoding,
-            "sampling_rate": self.sampling_rate,
-            "speed": self.speed,
-        }
-        return {k: v for k, v in params.items() if is_given(v) and v is not None}
-
-    def get_query_param_string(self):
-        """Forms the query parameter string from all model parameters."""
-        queries = []
-        for key, value in self.model_params.items():
-            queries.append(f"{key}={value}")
-
-        return "?" + "&".join(queries)
-
-
-def _parse_sse_message(message: str) -> dict:
-    """
-    Parse each response from the SSE endpoint.
-
-    The message will either be a string reading:
-    - `event: error`
-    - `event: message`
-    - `data: { "status_code": 200, "data": {"audio": ... } }`
-    """
-    message = message.strip()
-
-    if not message or "data" not in message:
-        return None
-
-    _, value = message.split(": ", 1)
-    message = json.loads(value)
-
-    if message.get("errors") is not None:
-        raise Exception(f"Status {message.status_code} error received: {message.errors}.")
-
-    return message
+    voice_id: str | None
 
 
 class TTS(tts.TTS):
     def __init__(
         self,
         *,
-        model: TTSModels | str = "neu_hq",
-        voice_id: NotGivenOr[str] = NOT_GIVEN,
+        voice_id: str = "8e9c4bc8-3979-48ab-8626-df53befc2090",
+        api_key: str | None = None,
         lang_code: TTSLangCodes | str = "en",
-        encoding: TTSEncodings | str = "pcm_linear",
         speed: float = 1.0,
         sample_rate: int = 22050,
-        api_key: NotGivenOr[str] = NOT_GIVEN,
         http_session: aiohttp.ClientSession | None = None,
         base_url: str = API_BASE_URL,
     ) -> None:
@@ -118,62 +67,34 @@ class TTS(tts.TTS):
         See https://docs.neuphonic.com for more documentation on all of these options, or go to https://app.neuphonic.com/ to test out different options.
 
         Args:
-            model (TTSModels | str, optional): The Neuphonic model to use. See Defaults to "neu_hq".
             voice_id (str, optional): The voice ID for the desired voice. Defaults to None.
             lang_code (TTSLanguages | str, optional): The language code for synthesis. Defaults to "en".
             encoding (TTSEncodings | str, optional): The audio encoding format. Defaults to "pcm_mulaw".
             speed (float, optional): The audio playback speed. Defaults to 1.0.
             sample_rate (int, optional): The audio sample rate in Hz. Defaults to 22050.
-            api_key (str | None, optional): The Neuphonic API key. If not provided, it will be read from the NEUPHONIC_API_TOKEN environment variable.
+            api_key (str | None, optional): The Neuphonic API key. If not provided, it will be read from the NEUPHONIC_API_KEY environment variable.
             http_session (aiohttp.ClientSession | None, optional): An existing aiohttp ClientSession to use. If not provided, a new session will be created.
             base_url (str, optional): The base URL for the Neuphonic API. Defaults to "api.neuphonic.com".
         """  # noqa: E501
         super().__init__(
             capabilities=tts.TTSCapabilities(streaming=True),
             sample_rate=sample_rate,
-            num_channels=NUM_CHANNELS,
-        )
-
-        neuphonic_api_key = (
-            api_key
-            if is_given(api_key)
-            else os.environ.get("NEUPHONIC_API_KEY") or os.environ.get("NEUPHONIC_API_TOKEN")
+            num_channels=1,
         )
 
-        if not neuphonic_api_key:
+        api_key = api_key or os.environ.get("NEUPHONIC_API_KEY")
+        if not api_key:
             raise ValueError("API key must be provided or set in NEUPHONIC_API_KEY")
 
         self._opts = _TTSOptions(
-            model=model,
             voice_id=voice_id,
             lang_code=lang_code,
-            encoding=encoding,
+            api_key=api_key,
             speed=speed,
-            sampling_rate=sample_rate,
-            api_key=neuphonic_api_key,
+            sample_rate=sample_rate,
             base_url=base_url,
         )
-
         self._session = http_session
-        self._pool = utils.ConnectionPool[aiohttp.ClientWebSocketResponse](
-            connect_cb=self._connect_ws,
-            close_cb=self._close_ws,
-            max_session_duration=90,
-            mark_refreshed_on_get=True,
-        )
-        self._streams = weakref.WeakSet[SynthesizeStream]()
-
-    async def _connect_ws(self) -> aiohttp.ClientWebSocketResponse:
-        session = self._ensure_session()
-        url = f"wss://{self._opts.base_url}/speak/{self._opts.lang_code}{self._opts.get_query_param_string()}"
-
-        return await asyncio.wait_for(
-            session.ws_connect(url, headers={AUTHORIZATION_HEADER: self._opts.api_key}),
-            self._conn_options.timeout,
-        )
-
-    async def _close_ws(self, ws: aiohttp.ClientWebSocketResponse):
-        await ws.close()
 
     def _ensure_session(self) -> aiohttp.ClientSession:
         if not self._session:
@@ -181,16 +102,11 @@ class TTS(tts.TTS):
 
         return self._session
 
-    def prewarm(self) -> None:
-        self._pool.prewarm()
-
     def update_options(
         self,
         *,
-        model: NotGivenOr[TTSModels] = NOT_GIVEN,
         voice_id: NotGivenOr[str] = NOT_GIVEN,
         lang_code: NotGivenOr[TTSLangCodes] = NOT_GIVEN,
-        encoding: NotGivenOr[TTSEncodings] = NOT_GIVEN,
         speed: NotGivenOr[float] = NOT_GIVEN,
         sample_rate: NotGivenOr[int] = NOT_GIVEN,
     ) -> None:
@@ -209,51 +125,19 @@ class TTS(tts.TTS):
             speed (float, optional): The audio playback speed.
             sample_rate (int, optional): The audio sample rate in Hz.
         """  # noqa: E501
-        if is_given(model):
-            self._opts.model = model
         if is_given(voice_id):
             self._opts.voice_id = voice_id
         if is_given(lang_code):
             self._opts.lang_code = lang_code
-        if is_given(encoding):
-            self._opts.encoding = encoding
         if is_given(speed):
             self._opts.speed = speed
         if is_given(sample_rate):
-            self._opts.sampling_rate = sample_rate
-        self._pool.invalidate()
+            self._opts.sample_rate = sample_rate
 
     def synthesize(
-        self,
-        text: str,
-        *,
-        conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,
+        self, text: str, *, conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS
     ) -> ChunkedStream:
-        return ChunkedStream(
-            tts=self,
-            input_text=text,
-            conn_options=conn_options,
-            opts=self._opts,
-            session=self._ensure_session(),
-        )
-
-    def stream(
-        self, *, conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS
-    ) -> SynthesizeStream:
-        stream = SynthesizeStream(
-            tts=self,
-            pool=self._pool,
-            opts=self._opts,
-        )
-        self._streams.add(stream)
-        return stream
-
-    async def aclose(self) -> None:
-        for stream in list(self._streams):
-            await stream.aclose()
-        self._streams.clear()
-        await self._pool.aclose()
-        await super().aclose()
+        return ChunkedStream(tts=self, input_text=text, conn_options=conn_options)
 
 
 class ChunkedStream(tts.ChunkedStream):
@@ -264,157 +148,86 @@ class ChunkedStream(tts.ChunkedStream):
         *,
         tts: TTS,
         input_text: str,
-        opts: _TTSOptions,
-        session: aiohttp.ClientSession,
         conn_options: APIConnectOptions,
     ) -> None:
         super().__init__(tts=tts, input_text=input_text, conn_options=conn_options)
-        self._opts, self._session = opts, session
-
-    async def _run(self) -> None:
-        request_id = utils.shortuuid()
-        bstream = utils.audio.AudioByteStream(
-            sample_rate=self._opts.sampling_rate, num_channels=NUM_CHANNELS
-        )
-
-        json_data = {
-            "text": self._input_text,
-            **self._opts.model_params,
-        }
-
-        headers = {
-            AUTHORIZATION_HEADER: self._opts.api_key,
-        }
+        self._tts: TTS = tts
+        self._opts = replace(tts._opts)
 
+    async def _run(self, output_emitter: tts.AudioEmitter) -> None:
         try:
-            async with self._session.post(
+            async with self._tts._ensure_session().post(
                 f"https://{self._opts.base_url}/sse/speak/{self._opts.lang_code}",
-                headers=headers,
-                json=json_data,
+                headers={AUTHORIZATION_HEADER: self._opts.api_key},
+                json={
+                    "text": self._input_text,
+                    "voice_id": self._opts.voice_id,
+                    "lang_code": self._opts.lang_code,
+                    "encoding": "pcm_linear",
+                    "sampling_rate": self._opts.sample_rate,
+                    "speed": self._opts.speed,
+                },
                 timeout=aiohttp.ClientTimeout(
                     total=30,
                     sock_connect=self._conn_options.timeout,
                 ),
-                read_bufsize=10
-                * 1024
-                * 1024,  # large read_bufsize to avoid `ValueError: Chunk too big`
-            ) as response:
-                response.raise_for_status()
-                emitter = tts.SynthesizedAudioEmitter(
-                    event_ch=self._event_ch,
-                    request_id=request_id,
+                # large read_bufsize to avoid `ValueError: Chunk too big`
+                read_bufsize=10 * 1024 * 1024,
+            ) as resp:
+                resp.raise_for_status()
+
+                output_emitter.initialize(
+                    request_id=utils.shortuuid(),
+                    sample_rate=self._opts.sample_rate,
+                    num_channels=1,
+                    mime_type="audio/pcm",
                 )
 
-                async for line in response.content:
-                    message = line.decode("utf-8").strip()
-                    if message:
-                        parsed_message = _parse_sse_message(message)
-
-                        if (
-                            parsed_message is not None
-                            and parsed_message.get("data", {}).get("audio") is not None
-                        ):
-                            audio_bytes = base64.b64decode(parsed_message["data"]["audio"])
-
-                            for frame in bstream.write(audio_bytes):
-                                emitter.push(frame)
-
-                for frame in bstream.flush():
-                    emitter.push(frame)
-                emitter.flush()
-        except asyncio.TimeoutError as e:
-            raise APITimeoutError() from e
+                async for line in resp.content:
+                    message = line.decode("utf-8")
+                    if not message:
+                        continue
+
+                    parsed_message = _parse_sse_message(message)
+
+                    if (
+                        parsed_message is not None
+                        and parsed_message.get("data", {}).get("audio") is not None
+                    ):
+                        audio_bytes = base64.b64decode(parsed_message["data"]["audio"])
+                        output_emitter.push(audio_bytes)
+
+                output_emitter.flush()
+        except asyncio.TimeoutError:
+            raise APITimeoutError() from None
         except aiohttp.ClientResponseError as e:
             raise APIStatusError(
-                message=e.message,
-                status_code=e.status,
-                request_id=None,
-                body=None,
-            ) from e
+                message=e.message, status_code=e.status, request_id=None, body=None
+            ) from None
         except Exception as e:
             raise APIConnectionError() from e
 
 
-class SynthesizeStream(tts.SynthesizeStream):
-    def __init__(
-        self,
-        *,
-        tts: TTS,
-        opts: _TTSOptions,
-        pool: utils.ConnectionPool[aiohttp.ClientWebSocketResponse],
-    ):
-        super().__init__(tts=tts)
-        self._opts, self._pool = opts, pool
-
-    async def _run(self) -> None:
-        request_id = utils.shortuuid()
-
-        async def _send_task(ws: aiohttp.ClientWebSocketResponse):
-            """Stream text to the websocket."""
-            async for data in self._input_ch:
-                self._mark_started()
-
-                if isinstance(data, self._FlushSentinel):
-                    await ws.send_str(json.dumps({"text": "<STOP>"}))
-                    continue
-
-                await ws.send_str(json.dumps({"text": data}))
-
-        async def _recv_task(ws: aiohttp.ClientWebSocketResponse):
-            audio_bstream = utils.audio.AudioByteStream(
-                sample_rate=self._opts.sampling_rate,
-                num_channels=NUM_CHANNELS,
-            )
-            emitter = tts.SynthesizedAudioEmitter(
-                event_ch=self._event_ch,
-                request_id=request_id,
-            )
-
-            while True:
-                try:
-                    msg = await ws.receive()
-                except Exception as e:
-                    raise APIStatusError(
-                        "Neuphonic connection closed unexpectedly",
-                        request_id=request_id,
-                    ) from e
-
-                if msg.type in (
-                    aiohttp.WSMsgType.CLOSED,
-                    aiohttp.WSMsgType.CLOSE,
-                    aiohttp.WSMsgType.CLOSING,
-                ):
-                    raise APIStatusError(
-                        "Neuphonic connection closed unexpectedly",
-                        request_id=request_id,
-                    )
-
-                if msg.type != aiohttp.WSMsgType.TEXT:
-                    logger.warning("Unexpected Neuphonic message type %s", msg.type)
-                    continue
-
-                data = json.loads(msg.data)
-
-                if data.get("data"):
-                    b64data = base64.b64decode(data["data"]["audio"])
-                    for frame in audio_bstream.write(b64data):
-                        emitter.push(frame)
-
-                    if data["data"].get("stop"):  # A bool flag, is True when audio reaches "<STOP>"
-                        for frame in audio_bstream.flush():
-                            emitter.push(frame)
-                        emitter.flush()
-                        break  # we are not going to receive any more audio
-                else:
-                    logger.error("Unexpected Neuphonic message %s", data)
-
-        async with self._pool.connection() as ws:
-            tasks = [
-                asyncio.create_task(_send_task(ws)),
-                asyncio.create_task(_recv_task(ws)),
-            ]
-
-            try:
-                await asyncio.gather(*tasks)
-            finally:
-                await utils.aio.gracefully_cancel(*tasks)
+def _parse_sse_message(message: str) -> dict | None:
+    """
+    Parse each response from the SSE endpoint.
+
+    The message will either be a string reading:
+    - `event: error`
+    - `event: message`
+    - `data: { "status_code": 200, "data": {"audio": ... } }`
+    """
+    message = message.strip()
+
+    if not message or "data" not in message:
+        return None
+
+    _, value = message.split(": ", 1)
+    message_dict: dict = json.loads(value)
+
+    if message_dict.get("errors") is not None:
+        raise Exception(
+            f"received error status {message_dict['status_code']}: {message_dict['errors']}"
+        )
+
+    return message_dict

livekit/plugins/neuphonic/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "1.0.23"
+__version__ = "1.1.0"

{livekit_plugins_neuphonic-1.0.23.dist-info → livekit_plugins_neuphonic-1.1.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-neuphonic
-Version: 1.0.23
+Version: 1.1.0
 Summary: Neuphonic inference plugin for LiveKit Agents
 Project-URL: Documentation, https://docs.livekit.io
 Project-URL: Website, https://livekit.io/
@@ -16,7 +16,7 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Topic :: Multimedia :: Sound/Audio
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9.0
-Requires-Dist: livekit-agents>=1.0.23
+Requires-Dist: livekit-agents>=1.1.0
 Description-Content-Type: text/markdown
 
 # Neuphonic plugin for LiveKit Agents

livekit_plugins_neuphonic-1.1.0.dist-info/RECORD

@@ -0,0 +1,9 @@
+livekit/plugins/neuphonic/__init__.py,sha256=c2yzK8LhbqZooNlJaX8TKKIjknTrZaEv6CmU9KF6dc4,1235
+livekit/plugins/neuphonic/log.py,sha256=rAHz71IcbvPkixndXBVffPQsmWUKTLqRaYRuPIxO29w,72
+livekit/plugins/neuphonic/models.py,sha256=dn6xtU7qJOI5XvEFupyww8IbCcwt5Ki-yS7ua_v6YxM,96
+livekit/plugins/neuphonic/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+livekit/plugins/neuphonic/tts.py,sha256=q-rV5SjadtFRNr19lLCyeb4tZvAdLoUy44tz2xSv1jQ,8370
+livekit/plugins/neuphonic/version.py,sha256=7SjyflIFTjH0djSotKGIRoRykPCqMpVYetIlvHMFuh0,600
+livekit_plugins_neuphonic-1.1.0.dist-info/METADATA,sha256=9GFF2OOfCGccJAjWvZ32MBd_QE0tABpzBDklbGZEx00,1313
+livekit_plugins_neuphonic-1.1.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+livekit_plugins_neuphonic-1.1.0.dist-info/RECORD,,

livekit_plugins_neuphonic-1.0.23.dist-info/RECORD

@@ -1,9 +0,0 @@
-livekit/plugins/neuphonic/__init__.py,sha256=QtlNAqYqv0gIHcVddh5INfdC1bnGkXErplWzhThtN3Q,1227
-livekit/plugins/neuphonic/log.py,sha256=rAHz71IcbvPkixndXBVffPQsmWUKTLqRaYRuPIxO29w,72
-livekit/plugins/neuphonic/models.py,sha256=Svfn_sWA3Q2ZXsPBXY-K5hslq5FE62hvyXBES2C3aSc,201
-livekit/plugins/neuphonic/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-livekit/plugins/neuphonic/tts.py,sha256=OIi6kqAJqqbczW1etEfy2-c244RqK2agm9kKalUdO28,14577
-livekit/plugins/neuphonic/version.py,sha256=BRUqwxRBnPVqEcIODJdaZHGAanu4zkwM4NsAQjNtUEM,601
-livekit_plugins_neuphonic-1.0.23.dist-info/METADATA,sha256=mcvg57hZGjsqiVWQJzwEQ4br0cLWUF-b0vrWB8pw-xo,1315
-livekit_plugins_neuphonic-1.0.23.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-livekit_plugins_neuphonic-1.0.23.dist-info/RECORD,,