livekit-plugins-cartesia 0.4.7__py3-none-any.whl → 0.4.9__py3-none-any.whl

livekit/plugins/cartesia/tts.py

@@ -18,13 +18,12 @@ import asyncio
 import base64
 import json
 import os
+import weakref
 from dataclasses import dataclass
-from typing import Any
+from typing import Any, Optional
 
 import aiohttp
-from livekit import rtc
 from livekit.agents import (
-    DEFAULT_API_CONNECT_OPTIONS,
     APIConnectionError,
     APIConnectOptions,
     APIStatusError,
@@ -61,6 +60,13 @@ class _TTSOptions:
     emotion: list[TTSVoiceEmotion | str] | None
     api_key: str
     language: str
+    base_url: str
+
+    def get_http_url(self, path: str) -> str:
+        return f"{self.base_url}{path}"
+
+    def get_ws_url(self, path: str) -> str:
+        return f"{self.base_url.replace('http', 'ws', 1)}{path}"
 
 
 class TTS(tts.TTS):
@@ -76,6 +82,7 @@ class TTS(tts.TTS):
         sample_rate: int = 24000,
         api_key: str | None = None,
         http_session: aiohttp.ClientSession | None = None,
+        base_url: str = "https://api.cartesia.ai",
     ) -> None:
         """
         Create a new instance of Cartesia TTS.
@@ -92,6 +99,7 @@ class TTS(tts.TTS):
             sample_rate (int, optional): The audio sample rate in Hz. Defaults to 24000.
             api_key (str, optional): The Cartesia API key. If not provided, it will be read from the CARTESIA_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 Cartesia API. Defaults to "https://api.cartesia.ai".
         """
 
         super().__init__(
@@ -113,8 +121,28 @@ class TTS(tts.TTS):
             speed=speed,
             emotion=emotion,
             api_key=api_key,
+            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=300,
+            mark_refreshed_on_get=True,
+        )
+        self._streams = weakref.WeakSet[SynthesizeStream]()
+
+    async def _connect_ws(self) -> aiohttp.ClientWebSocketResponse:
+        session = self._ensure_session()
+        url = self._opts.get_ws_url(
+            f"/tts/websocket?api_key={self._opts.api_key}&cartesia_version={API_VERSION}"
+        )
+        return await asyncio.wait_for(
+            session.ws_connect(url), 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:
@@ -122,6 +150,9 @@ class TTS(tts.TTS):
 
         return self._session
 
+    def prewarm(self) -> None:
+        self._pool.prewarm()
+
     def update_options(
         self,
         *,
@@ -155,7 +186,7 @@ class TTS(tts.TTS):
         self,
         text: str,
         *,
-        conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,
+        conn_options: Optional[APIConnectOptions] = None,
     ) -> ChunkedStream:
         return ChunkedStream(
             tts=self,
@@ -166,14 +197,22 @@ class TTS(tts.TTS):
         )
 
     def stream(
-        self, *, conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS
+        self, *, conn_options: Optional[APIConnectOptions] = None
     ) -> "SynthesizeStream":
-        return SynthesizeStream(
+        stream = SynthesizeStream(
             tts=self,
-            conn_options=conn_options,
+            pool=self._pool,
             opts=self._opts,
-            session=self._ensure_session(),
         )
+        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()
 
 
 class ChunkedStream(tts.ChunkedStream):
@@ -184,9 +223,9 @@ class ChunkedStream(tts.ChunkedStream):
         *,
         tts: TTS,
         input_text: str,
-        conn_options: APIConnectOptions,
         opts: _TTSOptions,
         session: aiohttp.ClientSession,
+        conn_options: Optional[APIConnectOptions] = None,
     ) -> None:
         super().__init__(tts=tts, input_text=input_text, conn_options=conn_options)
         self._opts, self._session = opts, session
@@ -207,7 +246,7 @@ class ChunkedStream(tts.ChunkedStream):
 
         try:
             async with self._session.post(
-                "https://api.cartesia.ai/tts/bytes",
+                self._opts.get_http_url("/tts/bytes"),
                 headers=headers,
                 json=json,
                 timeout=aiohttp.ClientTimeout(
@@ -216,19 +255,17 @@ class ChunkedStream(tts.ChunkedStream):
                 ),
             ) as resp:
                 resp.raise_for_status()
+                emitter = tts.SynthesizedAudioEmitter(
+                    event_ch=self._event_ch,
+                    request_id=request_id,
+                )
                 async for data, _ in resp.content.iter_chunks():
                     for frame in bstream.write(data):
-                        self._event_ch.send_nowait(
-                            tts.SynthesizedAudio(
-                                request_id=request_id,
-                                frame=frame,
-                            )
-                        )
+                        emitter.push(frame)
 
                 for frame in bstream.flush():
-                    self._event_ch.send_nowait(
-                        tts.SynthesizedAudio(request_id=request_id, frame=frame)
-                    )
+                    emitter.push(frame)
+                emitter.flush()
         except asyncio.TimeoutError as e:
             raise APITimeoutError() from e
         except aiohttp.ClientResponseError as e:
@@ -247,12 +284,11 @@ class SynthesizeStream(tts.SynthesizeStream):
         self,
         *,
         tts: TTS,
-        conn_options: APIConnectOptions,
         opts: _TTSOptions,
-        session: aiohttp.ClientSession,
+        pool: utils.ConnectionPool[aiohttp.ClientWebSocketResponse],
     ):
-        super().__init__(tts=tts, conn_options=conn_options)
-        self._opts, self._session = opts, session
+        super().__init__(tts=tts)
+        self._opts, self._pool = opts, pool
         self._sent_tokenizer_stream = tokenize.basic.SentenceTokenizer(
             min_sentence_len=BUFFERED_WORDS_COUNT
         ).stream()
@@ -289,22 +325,10 @@ class SynthesizeStream(tts.SynthesizeStream):
                 sample_rate=self._opts.sample_rate,
                 num_channels=NUM_CHANNELS,
             )
-
-            last_frame: rtc.AudioFrame | None = None
-
-            def _send_last_frame(*, segment_id: str, is_final: bool) -> None:
-                nonlocal last_frame
-                if last_frame is not None:
-                    self._event_ch.send_nowait(
-                        tts.SynthesizedAudio(
-                            request_id=request_id,
-                            segment_id=segment_id,
-                            frame=last_frame,
-                            is_final=is_final,
-                        )
-                    )
-
-                    last_frame = None
+            emitter = tts.SynthesizedAudioEmitter(
+                event_ch=self._event_ch,
+                request_id=request_id,
+            )
 
             while True:
                 msg = await ws.receive()
@@ -324,35 +348,23 @@ class SynthesizeStream(tts.SynthesizeStream):
 
                 data = json.loads(msg.data)
                 segment_id = data.get("context_id")
+                emitter._segment_id = segment_id
 
                 if data.get("data"):
                     b64data = base64.b64decode(data["data"])
                     for frame in audio_bstream.write(b64data):
-                        _send_last_frame(segment_id=segment_id, is_final=False)
-                        last_frame = frame
+                        emitter.push(frame)
                 elif data.get("done"):
                     for frame in audio_bstream.flush():
-                        _send_last_frame(segment_id=segment_id, is_final=False)
-                        last_frame = frame
-
-                    _send_last_frame(segment_id=segment_id, is_final=True)
-
+                        emitter.push(frame)
+                    emitter.flush()
                     if segment_id == request_id:
-                        # we're not going to receive more frames, close the connection
-                        await ws.close()
+                        # we're not going to receive more frames, end stream
                         break
                 else:
                     logger.error("unexpected Cartesia message %s", data)
 
-        url = f"wss://api.cartesia.ai/tts/websocket?api_key={self._opts.api_key}&cartesia_version={API_VERSION}"
-
-        ws: aiohttp.ClientWebSocketResponse | None = None
-
-        try:
-            ws = await asyncio.wait_for(
-                self._session.ws_connect(url), self._conn_options.timeout
-            )
-
+        async with self._pool.connection() as ws:
             tasks = [
                 asyncio.create_task(_input_task()),
                 asyncio.create_task(_sentence_stream_task(ws)),
@@ -363,9 +375,6 @@ class SynthesizeStream(tts.SynthesizeStream):
                 await asyncio.gather(*tasks)
             finally:
                 await utils.aio.gracefully_cancel(*tasks)
-        finally:
-            if ws is not None:
-                await ws.close()
 
 
 def _to_cartesia_options(opts: _TTSOptions) -> dict[str, Any]:

livekit/plugins/cartesia/version.py

@@ -12,4 +12,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.4.7"
+__version__ = "0.4.9"

{livekit_plugins_cartesia-0.4.7.dist-info → livekit_plugins_cartesia-0.4.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: livekit-plugins-cartesia
-Version: 0.4.7
+Version: 0.4.9
 Summary: LiveKit Agents Plugin for Cartesia
 Home-page: https://github.com/livekit/agents
 License: Apache-2.0
@@ -19,7 +19,7 @@ Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3 :: Only
 Requires-Python: >=3.9.0
 Description-Content-Type: text/markdown
-Requires-Dist: livekit-agents>=0.12.11
+Requires-Dist: livekit-agents<1.0.0,>=0.12.16
 Dynamic: classifier
 Dynamic: description
 Dynamic: description-content-type

livekit_plugins_cartesia-0.4.9.dist-info/RECORD

@@ -0,0 +1,10 @@
+livekit/plugins/cartesia/__init__.py,sha256=UTa6Q7IxhRBCwPftowHEUDvmBg99J_UjGS_yxTzKD7g,1095
+livekit/plugins/cartesia/log.py,sha256=4Mnhjng_DU1dIWP9IWjIQGZ67EV3LnQhWMWCHVudJbo,71
+livekit/plugins/cartesia/models.py,sha256=56CJgo7my-w-vpedir_ImV_aqKASeLihE5DbcCCgGJI,950
+livekit/plugins/cartesia/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+livekit/plugins/cartesia/tts.py,sha256=C7R_Z2nmWK3V0bQ30Nl5lgRgIZTfCzb7hiYWfBohGQs,14138
+livekit/plugins/cartesia/version.py,sha256=IgraxfFuS50xVM2gugcS0QqKE5YXL-FzFLco3Ffq6BY,600
+livekit_plugins_cartesia-0.4.9.dist-info/METADATA,sha256=P-63JHYrrl1DtEUBVVCnSPeS2vnws2dIdTPclQNG4ow,1470
+livekit_plugins_cartesia-0.4.9.dist-info/WHEEL,sha256=jB7zZ3N9hIM9adW7qlTAyycLYW9npaWKLRzaoVcLKcM,91
+livekit_plugins_cartesia-0.4.9.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
+livekit_plugins_cartesia-0.4.9.dist-info/RECORD,,

{livekit_plugins_cartesia-0.4.7.dist-info → livekit_plugins_cartesia-0.4.9.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.8.0)
+Generator: setuptools (75.8.2)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

livekit_plugins_cartesia-0.4.7.dist-info/RECORD

@@ -1,10 +0,0 @@
-livekit/plugins/cartesia/__init__.py,sha256=UTa6Q7IxhRBCwPftowHEUDvmBg99J_UjGS_yxTzKD7g,1095
-livekit/plugins/cartesia/log.py,sha256=4Mnhjng_DU1dIWP9IWjIQGZ67EV3LnQhWMWCHVudJbo,71
-livekit/plugins/cartesia/models.py,sha256=56CJgo7my-w-vpedir_ImV_aqKASeLihE5DbcCCgGJI,950
-livekit/plugins/cartesia/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-livekit/plugins/cartesia/tts.py,sha256=MKMLQBLc9A9XBkfwbFFt7PogUiWUHl-HozUR9mBmtMI,13906
-livekit/plugins/cartesia/version.py,sha256=85hVEOZ--XQ1Y7ngd1qGTZPpeywK2do8-2uhP_kdeyA,600
-livekit_plugins_cartesia-0.4.7.dist-info/METADATA,sha256=gZFU-QvodV6JmJiLDZ_OluuXviq8Zef-pYxhNEvxHho,1463
-livekit_plugins_cartesia-0.4.7.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-livekit_plugins_cartesia-0.4.7.dist-info/top_level.txt,sha256=OoDok3xUmXbZRvOrfvvXB-Juu4DX79dlq188E19YHoo,8
-livekit_plugins_cartesia-0.4.7.dist-info/RECORD,,