videosdk-plugins-deepgram 0.0.35__py3-none-any.whl → 0.0.37__py3-none-any.whl

videosdk/plugins/deepgram/__init__.py

@@ -1,3 +1,3 @@
 from .stt import DeepgramSTT
-
-__all__ = ["DeepgramSTT"]
+from .tts import DeepgramTTS
+__all__ = ["DeepgramSTT", "DeepgramTTS"]

videosdk/plugins/deepgram/stt.py

@@ -194,16 +194,23 @@ class DeepgramSTT(BaseSTT):
         """Cleanup resources"""
         if self._ws_task:
             self._ws_task.cancel()
+            logger.info("DeepgramSTT WebSocket task cancelled")
             try:
                 await self._ws_task
             except asyncio.CancelledError:
                 pass
             self._ws_task = None
+            logger.info("DeepgramSTT WebSocket task cleared")
 
         if self._ws:
             await self._ws.close()
+            logger.info("DeepgramSTT WebSocket closed")
             self._ws = None
 
         if self._session:
             await self._session.close()
+            logger.info("DeepgramSTT cleaned up")
             self._session = None
+        
+        # Call base class cleanup
+        await super().aclose()

videosdk/plugins/deepgram/tts.py

@@ -0,0 +1,186 @@
+from __future__ import annotations
+
+import asyncio
+import aiohttp
+import json
+from typing import Any, AsyncIterator, Union, Optional
+import os
+from videosdk.agents import TTS
+DEEPGRAM_SAMPLE_RATE = 24000
+DEEPGRAM_CHANNELS = 1
+DEFAULT_MODEL = "aura-2-thalia-en"
+DEFAULT_ENCODING = "linear16"
+API_BASE_URL = "wss://api.deepgram.com/v1/speak"
+
+
+class DeepgramTTS(TTS):
+    def __init__(
+            self,
+            *,
+            api_key: str | None = None,
+            model: str = DEFAULT_MODEL,
+            encoding: str = DEFAULT_ENCODING,
+            sample_rate: int = DEEPGRAM_SAMPLE_RATE,
+            base_url: str = API_BASE_URL,
+            **kwargs: Any,
+    ) -> None:
+        super().__init__(sample_rate=sample_rate, num_channels=DEEPGRAM_CHANNELS)
+
+        self.model = model
+        self.encoding = encoding
+        self.base_url = base_url
+        self.audio_track = None
+        self.loop = None
+        self._ws_session: aiohttp.ClientSession | None = None
+        self._ws_connection: aiohttp.ClientWebSocketResponse | None = None
+        self._send_task: asyncio.Task | None = None
+        self._recv_task: asyncio.Task | None = None
+        self._should_stop = False
+        self._first_chunk_sent = False
+        self._connection_lock = asyncio.Lock()
+
+        self.api_key = api_key or os.getenv("DEEPGRAM_API_KEY")
+        if not self.api_key:
+            raise ValueError(
+                "Deepgram API key must be provided either through the 'api_key' parameter or the DEEPGRAM_API_KEY environment variable."
+            )
+
+    def reset_first_audio_tracking(self) -> None:
+        self._first_chunk_sent = False
+
+    async def _ensure_connection(self) -> None:
+        async with self._connection_lock:
+            if self._ws_connection and not self._ws_connection.closed:
+                return
+
+            params = {
+                "model": self.model,
+                "encoding": self.encoding,
+                "sample_rate": self.sample_rate,
+            }
+            param_string = "&".join([f"{k}={v}" for k, v in params.items()])
+            full_ws_url = f"{self.base_url}?{param_string}"
+            headers = {"Authorization": f"Token {self.api_key}"}
+
+            self._ws_session = aiohttp.ClientSession()
+            self._ws_connection = await asyncio.wait_for(
+                self._ws_session.ws_connect(full_ws_url, headers=headers),
+                timeout=50.0
+            )
+            if self._recv_task and not self._recv_task.done():
+                self._recv_task.cancel()
+            self._recv_task = asyncio.create_task(self._receive_audio_task())
+
+    async def synthesize(
+            self,
+            text: AsyncIterator[str] | str,
+            **kwargs: Any,
+    ) -> None:
+        try:
+            if not self.audio_track or not self.loop:
+                self.emit("error", "Audio track or event loop not set")
+                return
+
+            await self.interrupt()
+            self._should_stop = False
+            await self._stream_synthesis(text)
+
+        except Exception as e:
+            self.emit("error", f"TTS synthesis failed: {str(e)}")
+
+    async def _stream_synthesis(self, text: Union[AsyncIterator[str], str]) -> None:
+        try:
+            await self._ensure_connection()
+            self._send_task = asyncio.create_task(self._send_text_task(text))
+            await self._send_task
+        except Exception as e:
+            self.emit("error", f"Streaming synthesis failed: {str(e)}")
+            await self.aclose()
+        finally:
+            if self._send_task and not self._send_task.done():
+                self._send_task.cancel()
+            self._send_task = None
+
+    async def _send_text_task(self, text: Union[AsyncIterator[str], str]) -> None:
+        if not self._ws_connection or self._ws_connection.closed:
+            return
+
+        try:
+            if isinstance(text, str):
+                if not self._should_stop:
+                    payload = {"type": "Speak", "text": text}
+                    await self._ws_connection.send_json(payload)
+            else:
+                async for chunk in text:
+                    if self._ws_connection.closed or self._should_stop:
+                        break
+                    payload = {"type": "Speak", "text": chunk}
+                    await self._ws_connection.send_json(payload)
+
+            if not self._ws_connection.closed and not self._should_stop:
+                await self._ws_connection.send_json({"type": "Flush"})
+        except asyncio.CancelledError:
+            pass
+        except Exception as e:
+            if not self._should_stop:
+                self.emit("error", f"Send task error: {str(e)}")
+
+    async def _receive_audio_task(self) -> None:
+        if not self._ws_connection:
+            return
+
+        try:
+            while not self._ws_connection.closed:
+                msg = await self._ws_connection.receive()
+
+                if msg.type == aiohttp.WSMsgType.BINARY:
+                    if not self._should_stop:
+                        await self._stream_audio_chunks(msg.data)
+                elif msg.type == aiohttp.WSMsgType.TEXT:
+                    data = json.loads(msg.data)
+                    if data.get('type') == 'Error' and not self._should_stop:
+                        self.emit("error", f"Deepgram error: {data.get('description', 'Unknown error')}")
+                        break
+                elif msg.type in (aiohttp.WSMsgType.CLOSED, aiohttp.WSMsgType.CLOSE, aiohttp.WSMsgType.CLOSING):
+                    break
+                elif msg.type == aiohttp.WSMsgType.ERROR:
+                    raise ConnectionError(f"WebSocket error: {self._ws_connection.exception()}")
+        except asyncio.CancelledError:
+            pass
+        except Exception as e:
+            if not self._should_stop:
+                self.emit("error", f"Receive task error: {str(e)}")
+
+    async def _stream_audio_chunks(self, audio_bytes: bytes) -> None:
+        if not audio_bytes or self._should_stop:
+            return
+
+        if self.audio_track and self.loop:
+            await self.audio_track.add_new_bytes(audio_bytes)
+
+    async def interrupt(self) -> None:
+        self._should_stop = True
+
+        if self.audio_track:
+            self.audio_track.interrupt()
+
+        if self._send_task and not self._send_task.done():
+            self._send_task.cancel()
+
+
+    async def aclose(self) -> None:
+        self._should_stop = True
+
+        for task in [self._send_task, self._recv_task]:
+            if task and not task.done():
+                task.cancel()
+
+        if self._ws_connection and not self._ws_connection.closed:
+            await self._ws_connection.close()
+        if self._ws_session and not self._ws_session.closed:
+            await self._ws_session.close()
+
+        self._ws_connection = None
+        self._ws_session = None
+
+        await super().aclose()

videosdk/plugins/deepgram/version.py

@@ -1 +1 @@
-__version__ = "0.0.35"
+__version__ = "0.0.37"

{videosdk_plugins_deepgram-0.0.35.dist-info → videosdk_plugins_deepgram-0.0.37.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: videosdk-plugins-deepgram
-Version: 0.0.35
+Version: 0.0.37
 Summary: VideoSDK Agent Framework plugin for Deepgram
 Author: videosdk
 License-Expression: Apache-2.0
@@ -12,7 +12,7 @@ Classifier: Topic :: Multimedia :: Sound/Audio
 Classifier: Topic :: Multimedia :: Video
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.11
-Requires-Dist: videosdk-agents>=0.0.35
+Requires-Dist: videosdk-agents>=0.0.37
 Description-Content-Type: text/markdown
 
 # VideoSDK Deepgram Plugin

videosdk_plugins_deepgram-0.0.37.dist-info/RECORD

@@ -0,0 +1,7 @@
+videosdk/plugins/deepgram/__init__.py,sha256=gzJRVqNliLxQHAbCSiki_YiFqicxRCSKgTk1cCLRkFg,98
+videosdk/plugins/deepgram/stt.py,sha256=FrhLg-57kqySa07Zo8yFJmMjcDcmdkihWcBMfFcxBRY,8309
+videosdk/plugins/deepgram/tts.py,sha256=hd4oifQ3lRV-Ry57EGf-8VrcBm2bM9Fj3VUKSAfmgh8,6884
+videosdk/plugins/deepgram/version.py,sha256=8S56Dzx6mwG65yKNAYNfLGkyvS8XrUW3xMrLAg1TKI0,23
+videosdk_plugins_deepgram-0.0.37.dist-info/METADATA,sha256=z0ktgPjqnrcaIeUlOtGwlqVTiKhi65W3KPcj_pgwv58,767
+videosdk_plugins_deepgram-0.0.37.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+videosdk_plugins_deepgram-0.0.37.dist-info/RECORD,,

videosdk_plugins_deepgram-0.0.35.dist-info/RECORD

@@ -1,6 +0,0 @@
-videosdk/plugins/deepgram/__init__.py,sha256=sfiNIvpVbFqbReYTskb3D--3v_U3-FujI9gdkcEGEmY,55
-videosdk/plugins/deepgram/stt.py,sha256=OXqOGMX-xRkjgju9lrnBu-iqC2ltScqFCemLS9wnoqY,8004
-videosdk/plugins/deepgram/version.py,sha256=QmUV3Ydc9xA0pOuihDI_wM1kvEzVZrdkq2eiedTW2UA,23
-videosdk_plugins_deepgram-0.0.35.dist-info/METADATA,sha256=EBV6zev_u2tdgx3AMhN97lIYhtH9-stHv4h84FO0S-A,767
-videosdk_plugins_deepgram-0.0.35.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-videosdk_plugins_deepgram-0.0.35.dist-info/RECORD,,