Package videosdk.plugins.deepgram

class DeepgramSTT(BaseSTT):
    def __init__(
        self,
        *,
        api_key: str | None = None,
        model: str = "nova-2",
        language: str = "en-US",
        interim_results: bool = True,
        punctuate: bool = True,
        smart_format: bool = True,
        sample_rate: int = 48000,
        endpointing: int = 50,
        filler_words: bool = True,
        base_url: str = "wss://api.deepgram.com/v1/listen",
    ) -> None:
        """Initialize the Deepgram STT plugin

        Args:
            api_key (str | None, optional): Deepgram API key. Uses DEEPGRAM_API_KEY environment variable if not provided. Defaults to None.
            model (str): The model to use for the STT plugin. Defaults to "nova-2".
            language (str): The language to use for the STT plugin. Defaults to "en-US".
            interim_results (bool): Whether to return interim results. Defaults to True.
            punctuate (bool): Whether to add punctuation. Defaults to True.
            smart_format (bool): Whether to use smart formatting. Defaults to True.
            sample_rate (int): Sample rate to use for the STT plugin. Defaults to 48000.
            endpointing (int): Endpointing threshold. Defaults to 50.
            filler_words (bool): Whether to include filler words. Defaults to True.
            base_url (str): The base URL to use for the STT plugin. Defaults to "wss://api.deepgram.com/v1/listen".
        """
        super().__init__()

        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 api_key parameter or DEEPGRAM_API_KEY environment variable")

        self.model = model
        self.language = language
        self.sample_rate = sample_rate
        self.interim_results = interim_results
        self.punctuate = punctuate
        self.smart_format = smart_format
        self.endpointing = endpointing
        self.filler_words = filler_words
        self.base_url = base_url
        self._session: Optional[aiohttp.ClientSession] = None
        self._ws: Optional[aiohttp.ClientWebSocketResponse] = None
        self._ws_task: Optional[asyncio.Task] = None
        self._last_speech_event_time = 0.0
        self._previous_speech_event_time = 0.0

    async def process_audio(
        self,
        audio_frames: bytes,
        language: Optional[str] = None,
        **kwargs: Any
    ) -> None:
        """Process audio frames and send to Deepgram's Streaming API"""

        if not self._ws:
            await self._connect_ws()
            self._ws_task = asyncio.create_task(self._listen_for_responses())

        try:
            await self._ws.send_bytes(audio_frames)
        except Exception as e:
            logger.error(f"Error in process_audio: {str(e)}")
            self.emit("error", str(e))
            if self._ws:
                await self._ws.close()
                self._ws = None
                if self._ws_task:
                    self._ws_task.cancel()
                    self._ws_task = None

    async def _listen_for_responses(self) -> None:
        """Background task to listen for WebSocket responses"""
        if not self._ws:
            return

        try:
            async for msg in self._ws:
                if msg.type == aiohttp.WSMsgType.TEXT:
                    data = msg.json()
                    responses = self._handle_ws_message(data)
                    for response in responses:
                        if self._transcript_callback:
                            await self._transcript_callback(response)
                elif msg.type == aiohttp.WSMsgType.ERROR:
                    logger.error(f"WebSocket error: {self._ws.exception()}")
                    self.emit(
                        "error", f"WebSocket error: {self._ws.exception()}")
                    break
        except Exception as e:
            logger.error(f"Error in WebSocket listener: {str(e)}")
            self.emit("error", f"Error in WebSocket listener: {str(e)}")
        finally:
            if self._ws:
                await self._ws.close()
                self._ws = None

    async def _connect_ws(self) -> None:
        """Establish WebSocket connection with Deepgram's Streaming API"""

        if not self._session:
            self._session = aiohttp.ClientSession()

        query_params = {
            "model": self.model,
            "language": self.language,
            "interim_results": str(self.interim_results).lower(),
            "punctuate": str(self.punctuate).lower(),
            "smart_format": str(self.smart_format).lower(),
            "encoding": "linear16",
            "sample_rate": str(self.sample_rate),
            "channels": 2,
            "endpointing": self.endpointing,
            "filler_words": str(self.filler_words).lower(),
            "vad_events": "true",
            "no_delay": "true",
        }
        headers = {
            "Authorization": f"Token {self.api_key}",
        }

        ws_url = f"{self.base_url}?{urlencode(query_params)}"

        try:
            self._ws = await self._session.ws_connect(ws_url, headers=headers)
        except Exception as e:
            logger.error(f"Error connecting to WebSocket: {str(e)}")
            raise

    def _handle_ws_message(self, msg: dict) -> list[STTResponse]:
        """Handle incoming WebSocket messages and generate STT responses"""
        responses = []
        try:
            if msg["type"] == "SpeechStarted":
                current_time = time.time()

                if self._last_speech_event_time == 0.0:
                    self._last_speech_event_time = current_time
                    return responses

                if current_time - self._last_speech_event_time < 1.0:
                    global_event_emitter.emit("speech_started")

                self._previous_speech_event_time = self._last_speech_event_time
                self._last_speech_event_time = current_time

            if msg["type"] == "Results":
                channel = msg["channel"]
                alternatives = channel["alternatives"]

                if alternatives and len(alternatives) > 0:
                    alt = alternatives[0]
                    is_final = msg["is_final"]
                    if alt["transcript"] == "":
                        return responses

                    response = STTResponse(
                        event_type=SpeechEventType.FINAL if is_final else SpeechEventType.INTERIM,
                        data=SpeechData(
                            text=alt["transcript"],
                            language=self.language,
                            confidence=alt.get("confidence", 0.0),
                            start_time=alt["words"][0]["start"] if alt["words"] else 0.0,
                            end_time=alt["words"][-1]["end"] if alt["words"] else 0.0,
                        ),
                        metadata={"model": self.model}
                    )
                    responses.append(response)

        except Exception as e:
            logger.error(f"Error handling WebSocket message: {str(e)}")

        return responses

    async def aclose(self) -> None:
        """Cleanup resources"""
        if self._ws_task:
            self._ws_task.cancel()
            try:
                await self._ws_task
            except asyncio.CancelledError:
                pass
            self._ws_task = None

        if self._ws:
            await self._ws.close()
            self._ws = None

        if self._session:
            await self._session.close()
            self._session = None

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()