dv-pipecat-ai 0.0.82.dev881__py3-none-any.whl → 0.0.85.dev1__py3-none-any.whl

pipecat/services/vistaar/llm.py

@@ -2,10 +2,11 @@
 
 import asyncio
 import json
+import random
 import time
 import uuid
 from dataclasses import dataclass
-from typing import Any, AsyncGenerator, Dict, Optional
+from typing import Any, AsyncGenerator, Dict, List, Optional
 from urllib.parse import urlencode
 
 import httpx
@@ -13,6 +14,8 @@ from loguru import logger
 from pydantic import BaseModel, Field
 
 from pipecat.frames.frames import (
+    EndFrame,
+    CancelFrame,
     Frame,
     LLMFullResponseEndFrame,
     LLMFullResponseStartFrame,
@@ -53,12 +56,13 @@ class VistaarLLMService(LLMService):
             source_lang: Source language code (e.g., 'mr' for Marathi, 'hi' for Hindi).
             target_lang: Target language code for responses.
             session_id: Session ID for maintaining conversation context.
-            extra: Additional model-specific parameters.
+            extra: Additional model-specific parameters
         """
 
         source_lang: Optional[str] = Field(default="mr")
         target_lang: Optional[str] = Field(default="mr")
         session_id: Optional[str] = Field(default=None)
+        pre_query_response_phrases: Optional[List[str]] = Field(default_factory=list)
         extra: Optional[Dict[str, Any]] = Field(default_factory=dict)
 
     def __init__(
@@ -68,7 +72,6 @@ class VistaarLLMService(LLMService):
         params: Optional[InputParams] = None,
         timeout: float = 30.0,
         interim_timeout: float = 5.0,
-        interim_message: str = "एक क्षण थांबा, मी बघतो. ",
         **kwargs,
     ):
         """Initialize Vistaar LLM service.
@@ -77,8 +80,7 @@ class VistaarLLMService(LLMService):
             base_url: The base URL for Vistaar API. Defaults to "https://vistaar.kenpath.ai/api".
             params: Input parameters for model configuration and behavior.
             timeout: Request timeout in seconds. Defaults to 30.0 seconds.
-            interim_timeout: Time in seconds before sending interim message. Defaults to 3.0 seconds.
-            interim_message: Message to send if API takes longer than interim_timeout. Defaults to "एक क्षण थांबा, मी बघतो. ".
+            interim_timeout: Time in seconds before sending interim message. Defaults to 5.0 seconds.
             **kwargs: Additional arguments passed to the parent LLMService.
         """
         super().__init__(**kwargs)
@@ -89,10 +91,10 @@ class VistaarLLMService(LLMService):
         self._source_lang = params.source_lang
         self._target_lang = params.target_lang
         self._session_id = params.session_id or str(uuid.uuid4())
+        self._pre_query_response_phrases = params.pre_query_response_phrases or []
         self._extra = params.extra if isinstance(params.extra, dict) else {}
         self._timeout = timeout
         self._interim_timeout = interim_timeout
-        self._interim_message = interim_message
 
         # Create an async HTTP client
         self._client = httpx.AsyncClient(timeout=httpx.Timeout(self._timeout), verify=False)
@@ -103,6 +105,8 @@ class VistaarLLMService(LLMService):
         self._partial_response = []  # Track what was actually sent before interruption
         self._interim_sent = False  # Track if interim message was sent
         self._interim_task = None  # Track interim message task
+        self._interim_completion_event = asyncio.Event()  # Track interim message completion
+        self._interim_in_progress = False  # Track if interim message is being spoken
 
         logger.info(
             f"Vistaar LLM initialized - Base URL: {self._base_url}, Session ID: {self._session_id}, Source Lang: {self._source_lang}, Target Lang: {self._target_lang}, Timeout: {self._timeout}s"
@@ -161,6 +165,10 @@ class VistaarLLMService(LLMService):
         # Set interruption flag
         self._is_interrupted = True
 
+        # Reset interim state on interruption
+        self._interim_in_progress = False
+        self._interim_completion_event.set()  # Unblock any waiting LLM responses
+
         # Cancel interim message task if active
         await self._cancel_interim_message_task(
             "Cancelled interim message task - handling interruption"
@@ -193,11 +201,28 @@ class VistaarLLMService(LLMService):
             if not self._is_interrupted and not self._interim_sent:
                 logger.info(f"Sending interim message after {self._interim_timeout}s timeout")
                 self._interim_sent = True
-                await self.push_frame(LLMTextFrame(text=self._interim_message))
+                self._interim_in_progress = True
+
+                # Use random selection from pre_query_response_phrases if available, otherwise fallback to default
+                if self._pre_query_response_phrases:
+                    message = random.choice(self._pre_query_response_phrases)
+                else:
+                    message = "एक क्षण थांबा, मी बघतो. "
+
+                await self.push_frame(LLMTextFrame(text=message))
+
+                # Wait for estimated TTS duration before marking as complete
+                estimated_tts_duration = max(2.0, len(message) * 0.08)  # ~80ms per character
+                logger.info(f"Waiting {estimated_tts_duration:.2f}s for interim TTS completion")
+                await asyncio.sleep(estimated_tts_duration)
         except asyncio.CancelledError:
             logger.debug("Interim message task cancelled")
         except Exception as e:
             logger.error(f"Error sending interim message: {e}")
+        finally:
+            # Signal that interim message handling is complete
+            self._interim_completion_event.set()
+            self._interim_in_progress = False
 
     async def _stream_response(self, query: str) -> AsyncGenerator[str, None]:
         """Stream response from Vistaar API using Server-Sent Events.
@@ -231,6 +256,8 @@ class VistaarLLMService(LLMService):
         self._is_interrupted = False
         self._partial_response = []
         self._interim_sent = False
+        self._interim_in_progress = False
+        self._interim_completion_event.clear()  # Reset the event for new request
 
         try:
             # Use httpx to handle SSE streaming
@@ -291,6 +318,7 @@ class VistaarLLMService(LLMService):
 
             # Start response
             await self.push_frame(LLMFullResponseStartFrame())
+            await self.push_frame(LLMFullResponseStartFrame(), FrameDirection.UPSTREAM)
             await self.start_processing_metrics()
             await self.start_ttfb_metrics()
 
@@ -307,6 +335,15 @@ class VistaarLLMService(LLMService):
                 if first_chunk:
                     await self.stop_ttfb_metrics()
                     first_chunk = False
+
+                    # Wait for interim message to complete if it was sent and is in progress
+                    if self._interim_sent:
+                        logger.debug(
+                            "Waiting for interim message completion before sending LLM response"
+                        )
+                        await self._interim_completion_event.wait()
+                        logger.debug("Interim message completed, proceeding with LLM response")
+
                     # Cancel interim message task since we got first response
                     await self._cancel_interim_message_task(
                         "Cancelled interim message task - got first response"
@@ -334,6 +371,7 @@ class VistaarLLMService(LLMService):
             )
             await self.stop_processing_metrics()
             await self.push_frame(LLMFullResponseEndFrame())
+            await self.push_frame(LLMFullResponseEndFrame(), FrameDirection.UPSTREAM)
 
     async def process_frame(self, frame: Frame, direction: FrameDirection):
         """Process frames for LLM completion requests.

pipecat/serializers/genesys.py

@@ -1,95 +0,0 @@
-import base64
-import json
-from typing import Optional
-
-from pydantic import BaseModel
-
-from pipecat.audio.utils import create_default_resampler, pcm_to_ulaw, ulaw_to_pcm
-from pipecat.frames.frames import (
-    AudioRawFrame,
-    Frame,
-    InputAudioRawFrame,
-    InputDTMFFrame,
-    KeypadEntry,
-    StartFrame,
-    StartInterruptionFrame,
-    TransportMessageFrame,
-    TransportMessageUrgentFrame,
-)
-from pipecat.serializers.base_serializer import FrameSerializer, FrameSerializerType
-
-
-class GenesysFrameSerializer(FrameSerializer):
-    class InputParams(BaseModel):
-        genesys_sample_rate: int = 8000  # Default Genesys rate (8kHz)
-        sample_rate: Optional[int] = None  # Pipeline input rate
-
-    def __init__(self, session_id: str, params: InputParams = InputParams()):
-        self._session_id = session_id
-        self._params = params
-        self._genesys_sample_rate = self._params.genesys_sample_rate
-        self._sample_rate = 0  # Pipeline input rate
-        self._resampler = create_default_resampler()
-        self._seq = 1  # Sequence number for outgoing messages
-
-    @property
-    def type(self) -> FrameSerializerType:
-        return FrameSerializerType.TEXT
-
-    async def setup(self, frame: StartFrame):
-        self._sample_rate = self._params.sample_rate or frame.audio_in_sample_rate
-
-    async def serialize(self, frame: Frame) -> str | bytes | None:
-        if isinstance(frame, StartInterruptionFrame):
-            answer = {
-                "version": "2",
-                "type": "clearAudio",  # Or appropriate event for interruption
-                "seq": self._seq,
-                "id": self._session_id,
-            }
-            self._seq += 1
-            return json.dumps(answer)
-        elif isinstance(frame, AudioRawFrame):
-            data = frame.audio
-            # Convert PCM to 8kHz μ-law for Genesys
-            serialized_data = await pcm_to_ulaw(
-                data, frame.sample_rate, self._genesys_sample_rate, self._resampler
-            )
-            payload = base64.b64encode(serialized_data).decode("utf-8")
-            answer = {
-                "version": "2",
-                "type": "audio",
-                "seq": self._seq,
-                "id": self._session_id,
-                "media": {
-                    "payload": payload,
-                    "format": "PCMU",
-                    "rate": self._genesys_sample_rate,
-                },
-            }
-            self._seq += 1
-            return json.dumps(answer)
-        elif isinstance(frame, (TransportMessageFrame, TransportMessageUrgentFrame)):
-            return json.dumps(frame.message)
-
-    async def deserialize(self, data: str | bytes) -> Frame | None:
-        message = json.loads(data)
-        if message.get("type") == "audio":
-            payload_base64 = message["media"]["payload"]
-            payload = base64.b64decode(payload_base64)
-            # Convert Genesys 8kHz μ-law to PCM at pipeline input rate
-            deserialized_data = await ulaw_to_pcm(
-                payload, self._genesys_sample_rate, self._sample_rate, self._resampler
-            )
-            audio_frame = InputAudioRawFrame(
-                audio=deserialized_data, num_channels=1, sample_rate=self._sample_rate
-            )
-            return audio_frame
-        elif message.get("type") == "dtmf":
-            digit = message.get("dtmf", {}).get("digit")
-            try:
-                return InputDTMFFrame(KeypadEntry(digit))
-            except ValueError:
-                return None
-        else:
-            return None

pipecat/services/google/test-google-chirp.py

@@ -1,45 +0,0 @@
-import asyncio
-import os
-
-from pipecat.frames.frames import TTSAudioRawFrame
-from pipecat.services.google.tts import GoogleTTSService
-
-
-async def test_chirp_tts():
-    # Get credentials from environment variable
-    credentials_path = (
-        "/Users/kalicharanvemuru/Documents/Code/pipecat/examples/ringg-chatbot/creds.json"
-    )
-
-    if not credentials_path or not os.path.exists(credentials_path):
-        raise ValueError(
-            "Please set GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account key file"
-        )
-
-    # Initialize the TTS service with Chirp voice
-    tts = GoogleTTSService(
-        credentials_path=credentials_path,
-        voice_id="en-US-Chirp3-HD-Charon",  # Using Chirp3 HD Charon voice
-        sample_rate=24000,
-    )
-
-    # Test text
-    test_text = "Hello, this is a test of the Google TTS service with Chirp voice."
-
-    print(f"Testing TTS with text: {test_text}")
-
-    # Generate speech
-    try:
-        async for frame in tts.run_tts(test_text):
-            if isinstance(frame, TTSAudioRawFrame):
-                print(f"Received audio chunk of size: {len(frame.audio)} bytes")
-            else:
-                print(f"Received frame: {frame.__class__.__name__}")
-
-        print("TTS generation completed successfully!")
-    except Exception as e:
-        print(f"Error during TTS generation: {str(e)}")
-
-
-if __name__ == "__main__":
-    asyncio.run(test_chirp_tts())