livekit-plugins-elevenlabs 0.8.1__py3-none-any.whl → 1.0.0.dev4__py3-none-any.whl

livekit/plugins/elevenlabs/tts.py

@@ -21,9 +21,10 @@ import json
 import os
 import weakref
 from dataclasses import dataclass
-from typing import Any, List, Optional
+from typing import Any
 
 import aiohttp
+
 from livekit.agents import (
     APIConnectionError,
     APIConnectOptions,
@@ -106,9 +107,9 @@ class TTS(tts.TTS):
         base_url: str | None = None,
         streaming_latency: int = 0,
         inactivity_timeout: int = WS_INACTIVITY_TIMEOUT,
-        word_tokenizer: Optional[tokenize.WordTokenizer] = None,
+        word_tokenizer: tokenize.WordTokenizer | None = None,
         enable_ssml_parsing: bool = False,
-        chunk_length_schedule: list[int] = [80, 120, 200, 260],  # range is [50, 500]
+        chunk_length_schedule: list[int] = None,  # range is [50, 500]
         http_session: aiohttp.ClientSession | None = None,
         # deprecated
         model_id: TTSModels | str | None = None,
@@ -131,6 +132,8 @@ class TTS(tts.TTS):
             language (str | None): Language code for the TTS model, as of 10/24/24 only valid for "eleven_turbo_v2_5". Optional.
         """
 
+        if chunk_length_schedule is None:
+            chunk_length_schedule = [80, 120, 200, 260]
         super().__init__(
             capabilities=tts.TTSCapabilities(
                 streaming=True,
@@ -171,15 +174,37 @@ class TTS(tts.TTS):
             inactivity_timeout=inactivity_timeout,
         )
         self._session = http_session
+        self._pool = utils.ConnectionPool[aiohttp.ClientWebSocketResponse](
+            connect_cb=self._connect_ws,
+            close_cb=self._close_ws,
+            max_session_duration=inactivity_timeout,
+            mark_refreshed_on_get=True,
+        )
         self._streams = weakref.WeakSet[SynthesizeStream]()
 
+    async def _connect_ws(self) -> aiohttp.ClientWebSocketResponse:
+        session = self._ensure_session()
+        return await asyncio.wait_for(
+            session.ws_connect(
+                _stream_url(self._opts),
+                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:
             self._session = utils.http_context.http_session()
 
         return self._session
 
-    async def list_voices(self) -> List[Voice]:
+    def prewarm(self) -> None:
+        self._pool.prewarm()
+
+    async def list_voices(self) -> list[Voice]:
         async with self._ensure_session().get(
             f"{self._opts.base_url}/voices",
             headers={AUTHORIZATION_HEADER: self._opts.api_key},
@@ -207,8 +232,8 @@ class TTS(tts.TTS):
         self,
         text: str,
         *,
-        conn_options: Optional[APIConnectOptions] = None,
-    ) -> "ChunkedStream":
+        conn_options: APIConnectOptions | None = None,
+    ) -> ChunkedStream:
         return ChunkedStream(
             tts=self,
             input_text=text,
@@ -217,15 +242,8 @@ class TTS(tts.TTS):
             session=self._ensure_session(),
         )
 
-    def stream(
-        self, *, conn_options: Optional[APIConnectOptions] = None
-    ) -> "SynthesizeStream":
-        stream = SynthesizeStream(
-            tts=self,
-            conn_options=conn_options,
-            opts=self._opts,
-            session=self._ensure_session(),
-        )
+    def stream(self, *, conn_options: APIConnectOptions | None = None) -> SynthesizeStream:
+        stream = SynthesizeStream(tts=self, pool=self._pool, opts=self._opts)
         self._streams.add(stream)
         return stream
 
@@ -233,6 +251,7 @@ class TTS(tts.TTS):
         for stream in list(self._streams):
             await stream.aclose()
         self._streams.clear()
+        await self._pool.aclose()
         await super().aclose()
 
 
@@ -245,7 +264,7 @@ class ChunkedStream(tts.ChunkedStream):
         tts: TTS,
         input_text: str,
         opts: _TTSOptions,
-        conn_options: Optional[APIConnectOptions] = None,
+        conn_options: APIConnectOptions | None = None,
         session: aiohttp.ClientSession,
     ) -> None:
         super().__init__(tts=tts, input_text=input_text, conn_options=conn_options)
@@ -320,12 +339,11 @@ class SynthesizeStream(tts.SynthesizeStream):
         self,
         *,
         tts: TTS,
-        session: aiohttp.ClientSession,
+        pool: utils.ConnectionPool[aiohttp.ClientWebSocketResponse],
         opts: _TTSOptions,
-        conn_options: Optional[APIConnectOptions] = None,
     ):
-        super().__init__(tts=tts, conn_options=conn_options)
-        self._opts, self._session = opts, session
+        super().__init__(tts=tts)
+        self._opts, self._pool = opts, pool
 
     async def _run(self) -> None:
         request_id = utils.shortuuid()
@@ -380,147 +398,138 @@ class SynthesizeStream(tts.SynthesizeStream):
         word_stream: tokenize.WordStream,
         request_id: str,
     ) -> None:
-        ws_conn = await self._session.ws_connect(
-            _stream_url(self._opts),
-            headers={AUTHORIZATION_HEADER: self._opts.api_key},
-        )
-
-        segment_id = utils.shortuuid()
-        decoder = utils.codecs.AudioStreamDecoder(
-            sample_rate=self._opts.sample_rate,
-            num_channels=1,
-        )
-
-        # 11labs protocol expects the first message to be an "init msg"
-        init_pkt = dict(
-            text=" ",
-            voice_settings=_strip_nones(dataclasses.asdict(self._opts.voice.settings))
-            if self._opts.voice.settings
-            else None,
-            generation_config=dict(
-                chunk_length_schedule=self._opts.chunk_length_schedule
-            ),
-        )
-        await ws_conn.send_str(json.dumps(init_pkt))
-        eos_sent = False
-
-        @utils.log_exceptions(logger=logger)
-        async def send_task():
-            nonlocal eos_sent
-            xml_content = []
-            async for data in word_stream:
-                text = data.token
-                # send the xml phoneme in one go
-                if (
-                    self._opts.enable_ssml_parsing
-                    and data.token.startswith("<phoneme")
-                    or xml_content
-                ):
-                    xml_content.append(text)
-                    if data.token.find("</phoneme>") > -1:
-                        text = self._opts.word_tokenizer.format_words(xml_content)
-                        xml_content = []
-                    else:
-                        continue
-
-                data_pkt = dict(text=f"{text} ")  # must always end with a space
-                self._mark_started()
-                await ws_conn.send_str(json.dumps(data_pkt))
-            if xml_content:
-                logger.warning("11labs stream ended with incomplete xml content")
-
-            # no more token, mark eos
-            eos_pkt = dict(text="")
-            await ws_conn.send_str(json.dumps(eos_pkt))
-            eos_sent = True
+        async with self._pool.connection() as ws_conn:
+            segment_id = utils.shortuuid()
+            expected_text = ""  # accumulate all tokens sent
 
-        # consumes from decoder and generates events
-        @utils.log_exceptions(logger=logger)
-        async def generate_task():
-            emitter = tts.SynthesizedAudioEmitter(
-                event_ch=self._event_ch,
-                request_id=request_id,
-                segment_id=segment_id,
+            decoder = utils.codecs.AudioStreamDecoder(
+                sample_rate=self._opts.sample_rate,
+                num_channels=1,
             )
-            async for frame in decoder:
-                emitter.push(frame)
-            emitter.flush()
 
-        # receives from ws and decodes audio
-        @utils.log_exceptions(logger=logger)
-        async def recv_task():
-            nonlocal eos_sent
-
-            while True:
-                msg = await ws_conn.receive()
-                if msg.type in (
-                    aiohttp.WSMsgType.CLOSED,
-                    aiohttp.WSMsgType.CLOSE,
-                    aiohttp.WSMsgType.CLOSING,
-                ):
-                    if not eos_sent:
+            # 11labs protocol expects the first message to be an "init msg"
+            init_pkt = {
+                "text": " ",
+                "voice_settings": _strip_nones(dataclasses.asdict(self._opts.voice.settings))
+                if self._opts.voice.settings
+                else None,
+                "generation_config": {"chunk_length_schedule": self._opts.chunk_length_schedule},
+            }
+            await ws_conn.send_str(json.dumps(init_pkt))
+
+            @utils.log_exceptions(logger=logger)
+            async def send_task():
+                nonlocal expected_text
+                xml_content = []
+                async for data in word_stream:
+                    text = data.token
+                    expected_text += text
+                    # send the xml phoneme in one go
+                    if (
+                        self._opts.enable_ssml_parsing
+                        and data.token.startswith("<phoneme")
+                        or xml_content
+                    ):
+                        xml_content.append(text)
+                        if text.find("</phoneme>") > -1:
+                            text = self._opts.word_tokenizer.format_words(xml_content)
+                            xml_content = []
+                        else:
+                            continue
+
+                    data_pkt = {"text": f"{text} "}  # must always end with a space
+                    self._mark_started()
+                    await ws_conn.send_str(json.dumps(data_pkt))
+                if xml_content:
+                    logger.warning("11labs stream ended with incomplete xml content")
+                await ws_conn.send_str(json.dumps({"flush": True}))
+
+            # consumes from decoder and generates events
+            @utils.log_exceptions(logger=logger)
+            async def generate_task():
+                emitter = tts.SynthesizedAudioEmitter(
+                    event_ch=self._event_ch,
+                    request_id=request_id,
+                    segment_id=segment_id,
+                )
+                async for frame in decoder:
+                    emitter.push(frame)
+                emitter.flush()
+
+            # receives from ws and decodes audio
+            @utils.log_exceptions(logger=logger)
+            async def recv_task():
+                nonlocal expected_text
+                received_text = ""
+
+                while True:
+                    msg = await ws_conn.receive()
+                    if msg.type in (
+                        aiohttp.WSMsgType.CLOSED,
+                        aiohttp.WSMsgType.CLOSE,
+                        aiohttp.WSMsgType.CLOSING,
+                    ):
                         raise APIStatusError(
                             "11labs connection closed unexpectedly, not all tokens have been consumed",
                             request_id=request_id,
                         )
-                    return
 
-                if msg.type != aiohttp.WSMsgType.TEXT:
-                    logger.warning("unexpected 11labs message type %s", msg.type)
-                    continue
-
-                data = json.loads(msg.data)
-                if data.get("audio"):
-                    b64data = base64.b64decode(data["audio"])
-                    decoder.push(b64data)
-
-                elif data.get("isFinal"):
-                    decoder.end_input()
-                    break
-                elif data.get("error"):
-                    raise APIStatusError(
-                        message=data["error"],
-                        status_code=500,
-                        request_id=request_id,
-                        body=None,
-                    )
-                else:
-                    raise APIStatusError(
-                        message=f"unexpected 11labs message {data}",
-                        status_code=500,
-                        request_id=request_id,
-                        body=None,
-                    )
+                    if msg.type != aiohttp.WSMsgType.TEXT:
+                        logger.warning("unexpected 11labs message type %s", msg.type)
+                        continue
 
-        tasks = [
-            asyncio.create_task(send_task()),
-            asyncio.create_task(recv_task()),
-            asyncio.create_task(generate_task()),
-        ]
-        try:
-            await asyncio.gather(*tasks)
-        except asyncio.TimeoutError as e:
-            raise APITimeoutError() from e
-        except aiohttp.ClientResponseError as e:
-            raise APIStatusError(
-                message=e.message,
-                status_code=e.status,
-                request_id=request_id,
-                body=None,
-            ) from e
-        except APIStatusError:
-            raise
-        except Exception as e:
-            raise APIConnectionError() from e
-        finally:
-            await utils.aio.gracefully_cancel(*tasks)
-            await decoder.aclose()
-            if ws_conn is not None:
-                await ws_conn.close()
+                    data = json.loads(msg.data)
+                    if data.get("audio"):
+                        b64data = base64.b64decode(data["audio"])
+                        decoder.push(b64data)
+
+                        if alignment := data.get("normalizedAlignment"):
+                            received_text += "".join(alignment.get("chars", [])).replace(" ", "")
+                            if received_text == expected_text:
+                                decoder.end_input()
+                                break
+                    elif data.get("error"):
+                        raise APIStatusError(
+                            message=data["error"],
+                            status_code=500,
+                            request_id=request_id,
+                            body=None,
+                        )
+                    else:
+                        raise APIStatusError(
+                            message=f"unexpected 11labs message {data}",
+                            status_code=500,
+                            request_id=request_id,
+                            body=None,
+                        )
+
+            tasks = [
+                asyncio.create_task(send_task()),
+                asyncio.create_task(recv_task()),
+                asyncio.create_task(generate_task()),
+            ]
+            try:
+                await asyncio.gather(*tasks)
+            except asyncio.TimeoutError as e:
+                raise APITimeoutError() from e
+            except aiohttp.ClientResponseError as e:
+                raise APIStatusError(
+                    message=e.message,
+                    status_code=e.status,
+                    request_id=request_id,
+                    body=None,
+                ) from e
+            except APIStatusError:
+                raise
+            except Exception as e:
+                raise APIConnectionError() from e
+            finally:
+                await utils.aio.gracefully_cancel(*tasks)
+                await decoder.aclose()
 
 
 def _dict_to_voices_list(data: dict[str, Any]):
-    voices: List[Voice] = []
+    voices: list[Voice] = []
     for voice in data["voices"]:
         voices.append(
             Voice(

livekit/plugins/elevenlabs/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.8.1"
+__version__ = "1.0.0.dev4"

{livekit_plugins_elevenlabs-0.8.1.dist-info → livekit_plugins_elevenlabs-1.0.0.dev4.dist-info}/METADATA

@@ -1,35 +1,25 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: livekit-plugins-elevenlabs
-Version: 0.8.1
+Version: 1.0.0.dev4
 Summary: Agent Framework plugin for voice synthesis with ElevenLabs' API.
-Home-page: https://github.com/livekit/agents
-License: Apache-2.0
 Project-URL: Documentation, https://docs.livekit.io
 Project-URL: Website, https://livekit.io/
 Project-URL: Source, https://github.com/livekit/agents
-Keywords: webrtc,realtime,audio,video,livekit,elevenlabs
+Author-email: LiveKit <support@livekit.io>
+License-Expression: Apache-2.0
+Keywords: audio,elevenlabs,livekit,realtime,video,webrtc
 Classifier: Intended Audience :: Developers
 Classifier: License :: OSI Approved :: Apache Software License
-Classifier: Topic :: Multimedia :: Sound/Audio
-Classifier: Topic :: Multimedia :: Video
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Topic :: Multimedia :: Sound/Audio
+Classifier: Topic :: Multimedia :: Video
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9.0
+Requires-Dist: livekit-agents>=1.0.0.dev4
 Description-Content-Type: text/markdown
-Requires-Dist: livekit-agents[codecs]<1.0.0,>=0.12.16
-Dynamic: classifier
-Dynamic: description
-Dynamic: description-content-type
-Dynamic: home-page
-Dynamic: keywords
-Dynamic: license
-Dynamic: project-url
-Dynamic: requires-dist
-Dynamic: requires-python
-Dynamic: summary
 
 # LiveKit Plugins Elevenlabs
 

livekit_plugins_elevenlabs-1.0.0.dev4.dist-info/RECORD

@@ -0,0 +1,9 @@
+livekit/plugins/elevenlabs/__init__.py,sha256=YZVadomFq3JWiZN6GWXJbuE4vaNNWq1CmdH25du8qwg,1249
+livekit/plugins/elevenlabs/log.py,sha256=hIuXqDsEB5GBa7rQY3z4Uqi1oCqc_lRmCHZEmXz0LHw,73
+livekit/plugins/elevenlabs/models.py,sha256=nB43wLS1ilzS7IxLYVSQxBjKPnbiPl4AHpHAOlG2i00,273
+livekit/plugins/elevenlabs/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+livekit/plugins/elevenlabs/tts.py,sha256=eJ66yP3ta2FH0LgQ64wHdjOHEoavwguOg6GeaMIr9IU,20394
+livekit/plugins/elevenlabs/version.py,sha256=koM_bT4QbztrKQ60Gjg7V4oe99CuxgGcpuUtWMOEKqU,605
+livekit_plugins_elevenlabs-1.0.0.dev4.dist-info/METADATA,sha256=1YSGTLIaJURkWYEOIl2LqZLdgU3y1KFM4YvGvd8s4G8,1316
+livekit_plugins_elevenlabs-1.0.0.dev4.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+livekit_plugins_elevenlabs-1.0.0.dev4.dist-info/RECORD,,

{livekit_plugins_elevenlabs-0.8.1.dist-info → livekit_plugins_elevenlabs-1.0.0.dev4.dist-info}/WHEEL

@@ -1,5 +1,4 @@
 Wheel-Version: 1.0
-Generator: setuptools (76.1.0)
+Generator: hatchling 1.27.0
 Root-Is-Purelib: true
 Tag: py3-none-any
-

livekit_plugins_elevenlabs-0.8.1.dist-info/RECORD

@@ -1,10 +0,0 @@
-livekit/plugins/elevenlabs/__init__.py,sha256=YZVadomFq3JWiZN6GWXJbuE4vaNNWq1CmdH25du8qwg,1249
-livekit/plugins/elevenlabs/log.py,sha256=hIuXqDsEB5GBa7rQY3z4Uqi1oCqc_lRmCHZEmXz0LHw,73
-livekit/plugins/elevenlabs/models.py,sha256=nB43wLS1ilzS7IxLYVSQxBjKPnbiPl4AHpHAOlG2i00,273
-livekit/plugins/elevenlabs/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-livekit/plugins/elevenlabs/tts.py,sha256=-w8IeAvyQER4PLraajJz6OWDufvKpD_fPM8oPsYtX9s,19335
-livekit/plugins/elevenlabs/version.py,sha256=PoHw-_DNE2B5SpeoQ-r6HSfVmbDgYuGamg0dN2jhayQ,600
-livekit_plugins_elevenlabs-0.8.1.dist-info/METADATA,sha256=l8gbEDr8EsedqYQiqBhx6K9XwAdTtnQWVCxmlyjVG9w,1529
-livekit_plugins_elevenlabs-0.8.1.dist-info/WHEEL,sha256=beeZ86-EfXScwlR_HKu4SllMC9wUEj_8Z_4FJ3egI2w,91
-livekit_plugins_elevenlabs-0.8.1.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
-livekit_plugins_elevenlabs-0.8.1.dist-info/RECORD,,

livekit_plugins_elevenlabs-0.8.1.dist-info/top_level.txt

@@ -1 +0,0 @@
-livekit