dv-pipecat-ai 0.0.85.dev825__py3-none-any.whl → 0.0.85.dev830__py3-none-any.whl

{dv_pipecat_ai-0.0.85.dev825.dist-info → dv_pipecat_ai-0.0.85.dev830.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dv-pipecat-ai
-Version: 0.0.85.dev825
+Version: 0.0.85.dev830
 Summary: An open source framework for voice (and multimodal) assistants
 License-Expression: BSD-2-Clause
 Project-URL: Source, https://github.com/pipecat-ai/pipecat

{dv_pipecat_ai-0.0.85.dev825.dist-info → dv_pipecat_ai-0.0.85.dev830.dist-info}/RECORD

@@ -1,4 +1,4 @@
-dv_pipecat_ai-0.0.85.dev825.dist-info/licenses/LICENSE,sha256=DWY2QGf2eMCFhuu2ChairtT6CB7BEFffNVhXWc4Od08,1301
+dv_pipecat_ai-0.0.85.dev830.dist-info/licenses/LICENSE,sha256=DWY2QGf2eMCFhuu2ChairtT6CB7BEFffNVhXWc4Od08,1301
 pipecat/__init__.py,sha256=j0Xm6adxHhd7D06dIyyPV_GlBYLlBnTAERVvD_jAARQ,861
 pipecat/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/adapters/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -79,7 +79,7 @@ pipecat/extensions/voicemail/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NM
 pipecat/extensions/voicemail/voicemail_detector.py,sha256=JxmU2752iWP_1_GmzZReNESUTFAeyEa4XBPL20_C208,30004
 pipecat/frames/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/frames/frames.proto,sha256=JXZm3VXLR8zMOUcOuhVoe2mhM3MQIQGMJXLopdJO_5Y,839
-pipecat/frames/frames.py,sha256=CxlrFst5DuD6kDp2CE6kWigVezF94y-Snf6h8w1pwVU,49522
+pipecat/frames/frames.py,sha256=vuYtmyK1QSU2AWx2c_pFQhcmpXqSTnfqAXF6DXKzTG8,49605
 pipecat/frames/protobufs/frames_pb2.py,sha256=VHgGV_W7qQ4sfQK6RHb5_DggLm3PiSYMr6aBZ8_p1cQ,2590
 pipecat/metrics/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pipecat/metrics/metrics.py,sha256=bdZNciEtLTtA-xgoKDz2RJAy6fKrXkTwz3pryVHzc2M,2713
@@ -113,7 +113,7 @@ pipecat/processors/idle_frame_processor.py,sha256=z8AuhGap61lA5K35P6XCaOpn4kkmK_
 pipecat/processors/logger.py,sha256=8xa4KKekXQIETlQR7zoGnwUpLNo8CeDVm7YjyXePN-w,2385
 pipecat/processors/producer_processor.py,sha256=iIIOHZd77APvUGP7JqFbznAHUnCULcq_qYiSEjwXHcc,3265
 pipecat/processors/text_transformer.py,sha256=LnfWJYzntJhZhrQ1lgSSY4D4VbHtrQJgrC227M69ZYU,1718
-pipecat/processors/transcript_processor.py,sha256=fr5JtlTOfmKnfmYG8ZwRj4DpZWP-uuGi6aNNKtlLxRg,12491
+pipecat/processors/transcript_processor.py,sha256=TbMSeZCxXgQIdYodXuMSwLp6BvXTy7vQB9YiQZfPxc0,12488
 pipecat/processors/two_stage_user_idle_processor.py,sha256=uf2aZh_lfW-eMxmFogP3R4taAJ1yXOSqjKsR7oXtD0Y,2938
 pipecat/processors/user_idle_processor.py,sha256=Dl-Kcg0B4JZqWXXiyGuvYszGimbu2oKOyOJC92R9_hE,9140
 pipecat/processors/aggregators/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -122,7 +122,7 @@ pipecat/processors/aggregators/gated.py,sha256=tii0sRrBkRW6y9Xq5iTWPnqlOEejU4VqP
 pipecat/processors/aggregators/gated_llm_context.py,sha256=CPv6sMA8irD1zZ3fU1gSv6D7qcPvCA0MdpFhBtJ_ekI,3007
 pipecat/processors/aggregators/gated_open_ai_llm_context.py,sha256=DgqmdPj1u3fP_SVmxtfP7NjHqnyhN_RVVTDfmjbkxAs,361
 pipecat/processors/aggregators/llm_context.py,sha256=wNbZA0Vt0FzNc5cu06xiv1z7DIClIlfqR1ZD8EusbVw,11085
-pipecat/processors/aggregators/llm_response.py,sha256=cBNGU8Ld4zT36-QsE1EJemrNA12q7lc9i-vLM9qmLcQ,48075
+pipecat/processors/aggregators/llm_response.py,sha256=V6wBTzfUGLJfMuI34fkf5VTR0I66AWIW8btxKI8_3IM,48795
 pipecat/processors/aggregators/llm_response_universal.py,sha256=5PqmpATpekD8BVWyBExZgatKHsNbZem8M-A7_VwTbiQ,34334
 pipecat/processors/aggregators/openai_llm_context.py,sha256=cC8DXdVPERRN04i0i-1Ys6kusvnbMALeH-Z8Pu5K684,12999
 pipecat/processors/aggregators/sentence.py,sha256=E7e3knfQl6HEGpYMKPklF1aO_gOn-rr7SnynErwfkQk,2235
@@ -318,7 +318,7 @@ pipecat/services/riva/__init__.py,sha256=rObSsj504O_TMXhPBg_ymqKslZBhovlR-A0aaRZ
 pipecat/services/riva/stt.py,sha256=bAss4dimx8eideaSPmPHM15_rSV3tfXNf13o5n1mfv4,25146
 pipecat/services/riva/tts.py,sha256=idbqx3I2NlWCXtrIFsjEaYapxA3BLIA14ai3aMBh-2w,8158
 pipecat/services/salesforce/__init__.py,sha256=OFvYbcvCadYhcKdBAVLj3ZUXVXQ1HyVyhgxIFf6_Thg,173
-pipecat/services/salesforce/llm.py,sha256=xBy5PSftIwD0v3eXxC5QCSlb8R8HiU1KQTX4R9qEmkA,28108
+pipecat/services/salesforce/llm.py,sha256=2wh4U5mBElj5Olze2L0jC6V-UjFDC8ZEXlAKu0ORTwI,27825
 pipecat/services/sambanova/__init__.py,sha256=oTXExLic-qTcsfsiWmssf3Elclf3IIWoN41_2IpoF18,128
 pipecat/services/sambanova/llm.py,sha256=5XVfPLEk__W8ykFqLdV95ZUhlGGkAaJwmbciLdZYtTc,8976
 pipecat/services/sambanova/stt.py,sha256=ZZgEZ7WQjLFHbCko-3LNTtVajjtfUvbtVLtFcaNadVQ,2536
@@ -415,7 +415,7 @@ pipecat/utils/tracing/service_decorators.py,sha256=fwzxFpi8DJl6BJbK74G0UEB4ccMJg
 pipecat/utils/tracing/setup.py,sha256=7TEgPNpq6M8lww8OQvf0P9FzYc5A30xICGklVA-fua0,2892
 pipecat/utils/tracing/turn_context_provider.py,sha256=ikon3plFOx0XbMrH6DdeHttNpb-U0gzMZIm3bWLc9eI,2485
 pipecat/utils/tracing/turn_trace_observer.py,sha256=dma16SBJpYSOE58YDWy89QzHyQFc_9gQZszKeWixuwc,9725
-dv_pipecat_ai-0.0.85.dev825.dist-info/METADATA,sha256=9do57Wuqp0n9sMyV0YkeHfibkx3w57ANLCiGI7Pbzsk,32924
-dv_pipecat_ai-0.0.85.dev825.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-dv_pipecat_ai-0.0.85.dev825.dist-info/top_level.txt,sha256=kQzG20CxGf-nSsHmtXHx3hY2-8zHA3jYg8jk0TajqXc,8
-dv_pipecat_ai-0.0.85.dev825.dist-info/RECORD,,
+dv_pipecat_ai-0.0.85.dev830.dist-info/METADATA,sha256=wPJAPffJo_L5wKNWKbIxlaBG09JAGKUTFl_qkLwmoPw,32924
+dv_pipecat_ai-0.0.85.dev830.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+dv_pipecat_ai-0.0.85.dev830.dist-info/top_level.txt,sha256=kQzG20CxGf-nSsHmtXHx3hY2-8zHA3jYg8jk0TajqXc,8
+dv_pipecat_ai-0.0.85.dev830.dist-info/RECORD,,

pipecat/frames/frames.py

@@ -451,6 +451,7 @@ class TranscriptionMessage:
         content: The message content/text.
         user_id: Optional identifier for the user.
         timestamp: Optional timestamp when the message was created.
+        message_id: Optional unique identifier for tracking and dropping messages.
     """
 
     role: Literal["user", "assistant"]

pipecat/processors/aggregators/llm_response.py

@@ -551,7 +551,8 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
                     await self._process_aggregation()
                 else:
                     self.logger.debug("Interruption conditions not met - not pushing aggregation")
-                    # Don't process aggregation, just reset it
+                    # Don't process aggregation, discard pending transcriptions and reset
+                    await self._discard_pending_transcriptions("interruption_conditions_not_met")
                     await self.reset()
             else:
                 if trigger_interruption:
@@ -628,8 +629,17 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
         for s in self.interruption_strategies:
             await s.append_audio(frame.audio, frame.sample_rate)
 
+    async def _discard_pending_transcriptions(self, reason: str):
+        """Notify upstream processors that pending transcripts should be dropped."""
+        if self._pending_transcription_ids:
+            drop_frame = TranscriptDropFrame(transcript_ids=list(self._pending_transcription_ids))
+            await self.push_frame(drop_frame, FrameDirection.UPSTREAM)
+        self._pending_transcription_ids.clear()
+
     async def _handle_user_started_speaking(self, frame: UserStartedSpeakingFrame):
         if len(self._aggregation) > 0:
+            self.logger.debug(f"Dropping {self._aggregation}")
+            self._aggregation = ""
             await self._discard_pending_transcriptions("user_started_speaking")
         self._latest_final_transcript = ""
         self._last_user_speaking_time = time.time()
@@ -805,6 +815,7 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
             if self._bot_speaking and not self._params.enable_emulated_vad_interruptions:
                 # If emulated VAD interruptions are disabled and bot is speaking, ignore
                 logger.debug("Ignoring user speaking emulation, bot is speaking.")
+                await self._discard_pending_transcriptions("emulated_vad_ignored")
                 await self.reset()
             else:
                 # Either bot is not speaking, or emulated VAD interruptions are enabled

pipecat/processors/transcript_processor.py

@@ -64,12 +64,13 @@ class BaseTranscriptProcessor(FrameProcessor):
         if not frame.transcript_ids:
             return
 
+        await self._call_event_handler("on_transcript_drop", frame)
+
         drop_ids = set(frame.transcript_ids)
         if drop_ids:
             self._processed_messages = [
                 msg for msg in self._processed_messages if msg.message_id not in drop_ids
             ]
-            await self._call_event_handler("on_transcript_drop", frame)
 
 
 class UserTranscriptProcessor(BaseTranscriptProcessor):

pipecat/services/salesforce/llm.py

@@ -13,6 +13,7 @@ from dataclasses import dataclass
 from typing import AsyncGenerator, Dict, Optional
 
 import httpx
+from env_config import api_config
 from loguru import logger
 
 from pipecat.frames.frames import (
@@ -23,6 +24,10 @@ from pipecat.frames.frames import (
     LLMTextFrame,
     LLMUpdateSettingsFrame,
 )
+from pipecat.processors.aggregators.llm_response import (
+    LLMAssistantAggregatorParams,
+    LLMUserAggregatorParams,
+)
 from pipecat.processors.aggregators.openai_llm_context import (
     OpenAILLMContext,
     OpenAILLMContextFrame,
@@ -34,11 +39,6 @@ from pipecat.services.openai.llm import (
     OpenAIContextAggregatorPair,
     OpenAIUserContextAggregator,
 )
-from pipecat.processors.aggregators.llm_response import (
-    LLMAssistantAggregatorParams,
-    LLMUserAggregatorParams,
-)
-from env_config import api_config
 from pipecat.utils.redis import create_async_redis_client
 
 
@@ -96,12 +96,11 @@ class SalesforceAgentLLMService(LLMService):
         # Initialize parent LLM service
         super().__init__(**kwargs)
         self._agent_id = agent_id
-        self._org_domain = org_domain 
+        self._org_domain = org_domain
         self._client_id = client_id
         self._client_secret = client_secret
         self._api_host = api_host
 
-
         # Validate required environment variables
         required_vars = {
             "SALESFORCE_AGENT_ID": self._agent_id,
@@ -145,7 +144,6 @@ class SalesforceAgentLLMService(LLMService):
         )
 
         self._schedule_session_warmup()
-        
 
     async def __aenter__(self):
         """Async context manager entry."""
@@ -237,7 +235,7 @@ class SalesforceAgentLLMService(LLMService):
             return
 
         ttl_seconds = 3600  # Default fallback
-        
+
         # Try to get expiration from expires_in parameter first
         if expires_in is not None:
             try:
@@ -246,7 +244,7 @@ class SalesforceAgentLLMService(LLMService):
             except (TypeError, ValueError):
                 logger.debug("Unable to parse expires_in parameter")
                 expires_in = None
-        
+
         # If no expires_in available, use default TTL
         if expires_in is None:
             logger.debug("No expiration info found, using default TTL")
@@ -271,7 +269,7 @@ class SalesforceAgentLLMService(LLMService):
 
     async def _get_access_token(self, *, force_refresh: bool = False) -> str:
         """Get OAuth access token using client credentials.
-        
+
         Args:
             force_refresh: If True, skip cache and fetch fresh token from Salesforce.
         """
@@ -301,15 +299,15 @@ class SalesforceAgentLLMService(LLMService):
 
     async def _make_authenticated_request(self, method: str, url: str, **kwargs):
         """Make an authenticated HTTP request with automatic token refresh on auth errors.
-        
+
         Args:
             method: HTTP method (GET, POST, DELETE, etc.)
             url: Request URL
             **kwargs: Additional arguments passed to httpx request
-            
+
         Returns:
             httpx.Response: The HTTP response
-            
+
         Raises:
             Exception: If request fails after token refresh attempt
         """
@@ -318,7 +316,7 @@ class SalesforceAgentLLMService(LLMService):
         headers = kwargs.get("headers", {})
         headers["Authorization"] = f"Bearer {access_token}"
         kwargs["headers"] = headers
-        
+
         try:
             response = await self._http_client.request(method, url, **kwargs)
             response.raise_for_status()
@@ -326,14 +324,16 @@ class SalesforceAgentLLMService(LLMService):
         except httpx.HTTPStatusError as e:
             # If authentication error, clear cache and retry with fresh token
             if e.response.status_code in (401, 403):
-                logger.warning(f"Salesforce authentication error ({e.response.status_code}), refreshing token")
+                logger.warning(
+                    f"Salesforce authentication error ({e.response.status_code}), refreshing token"
+                )
                 await self._clear_cached_access_token()
-                
+
                 # Retry with fresh token
                 fresh_token = await self._get_access_token(force_refresh=True)
                 headers["Authorization"] = f"Bearer {fresh_token}"
                 kwargs["headers"] = headers
-                
+
                 response = await self._http_client.request(method, url, **kwargs)
                 response.raise_for_status()
                 return response
@@ -359,9 +359,7 @@ class SalesforceAgentLLMService(LLMService):
 
         try:
             response = await self._make_authenticated_request(
-                "POST", session_url, 
-                headers={"Content-Type": "application/json"}, 
-                json=payload
+                "POST", session_url, headers={"Content-Type": "application/json"}, json=payload
             )
             session_data = response.json()
             session_id = session_data["sessionId"]
@@ -419,8 +417,7 @@ class SalesforceAgentLLMService(LLMService):
                 # End the session via API
                 url = f"{self._api_host}/einstein/ai-agent/v1/sessions/{session_id}"
                 await self._make_authenticated_request(
-                    "DELETE", url,
-                    headers={"x-session-end-reason": "UserRequest"}
+                    "DELETE", url, headers={"x-session-end-reason": "UserRequest"}
                 )
             except Exception as e:
                 logger.warning(f"Failed to end session {session_id}: {e}")
@@ -431,32 +428,32 @@ class SalesforceAgentLLMService(LLMService):
 
     def _extract_user_message(self, context: OpenAILLMContext) -> str:
         """Extract the last user message from context.
-        
+
         Similar to Vistaar pattern - extract only the most recent user message.
-        
+
         Args:
             context: The OpenAI LLM context containing messages.
-            
+
         Returns:
             The last user message as a string.
         """
         messages = context.get_messages()
-        
+
         # Find the last user message (iterate in reverse for efficiency)
         for message in reversed(messages):
             if message.get("role") == "user":
                 content = message.get("content", "")
-                
+
                 # Handle content that might be a list (for multimodal messages)
                 if isinstance(content, list):
                     text_parts = [
                         item.get("text", "") for item in content if item.get("type") == "text"
                     ]
                     content = " ".join(text_parts)
-                
+
                 if isinstance(content, str):
                     return content.strip()
-        
+
         return ""
 
     def _generate_sequence_id(self) -> int:
@@ -464,7 +461,9 @@ class SalesforceAgentLLMService(LLMService):
         self._sequence_counter += 1
         return self._sequence_counter
 
-    async def _stream_salesforce_response(self, session_id: str, user_message: str) -> AsyncGenerator[str, None]:
+    async def _stream_salesforce_response(
+        self, session_id: str, user_message: str
+    ) -> AsyncGenerator[str, None]:
         """Stream response from Salesforce Agent API."""
         url = f"{self._api_host}/einstein/ai-agent/v1/sessions/{session_id}/messages/stream"
 
@@ -472,15 +471,9 @@ class SalesforceAgentLLMService(LLMService):
             "message": {
                 "sequenceId": self._generate_sequence_id(),
                 "type": "Text",
-                "text": user_message
+                "text": user_message,
             },
-            "variables": [
-                {
-                    "name": "$Context.EndUserLanguage",
-                    "type": "Text",
-                    "value": "en_US"
-                }
-            ]
+            "variables": [{"name": "$Context.EndUserLanguage", "type": "Text", "value": "en_US"}],
         }
 
         # First attempt with current token
@@ -493,9 +486,11 @@ class SalesforceAgentLLMService(LLMService):
 
         try:
             logger.info(f"🌐 Salesforce API request: {user_message[:50]}...")
-            async with self._http_client.stream("POST", url, headers=headers, json=message_data) as response:
+            async with self._http_client.stream(
+                "POST", url, headers=headers, json=message_data
+            ) as response:
                 response.raise_for_status()
-                
+
                 async for line in response.aiter_lines():
                     if not line:
                         continue
@@ -525,17 +520,23 @@ class SalesforceAgentLLMService(LLMService):
         except httpx.HTTPStatusError as e:
             # If authentication error, retry with fresh token
             if e.response.status_code in (401, 403):
-                logger.warning(f"Salesforce streaming authentication error ({e.response.status_code}), refreshing token")
+                logger.warning(
+                    f"Salesforce streaming authentication error ({e.response.status_code}), refreshing token"
+                )
                 await self._clear_cached_access_token()
-                
+
                 # Retry with fresh token
                 fresh_token = await self._get_access_token(force_refresh=True)
                 headers["Authorization"] = f"Bearer {fresh_token}"
-                
-                logger.info(f"🔄 Retrying Salesforce stream with fresh token: {user_message[:50]}...")
-                async with self._http_client.stream("POST", url, headers=headers, json=message_data) as response:
+
+                logger.info(
+                    f"🔄 Retrying Salesforce stream with fresh token: {user_message[:50]}..."
+                )
+                async with self._http_client.stream(
+                    "POST", url, headers=headers, json=message_data
+                ) as response:
                     response.raise_for_status()
-                    
+
                     async for line in response.aiter_lines():
                         if not line:
                             continue
@@ -576,40 +577,41 @@ class SalesforceAgentLLMService(LLMService):
             context: The OpenAI LLM context containing messages to process.
         """
         logger.info(f"🔄 Salesforce processing context with {len(context.get_messages())} messages")
-        
+
         # Extract user message from context first
         user_message = self._extract_user_message(context)
-        
+
         if not user_message:
             logger.warning("Salesforce: No user message found in context")
             return
-        
+
         try:
             logger.info(f"🎯 Salesforce extracted query: {user_message}")
-            
-            # Start response 
+
+            # Start response
             await self.push_frame(LLMFullResponseStartFrame())
-            await self.push_frame(LLMFullResponseStartFrame(),FrameDirection.UPSTREAM)
+            await self.push_frame(LLMFullResponseStartFrame(), FrameDirection.UPSTREAM)
             await self.start_processing_metrics()
             await self.start_ttfb_metrics()
-            
+
             # Get or create session
             session_id = await self._get_or_create_session()
-            
+
             first_chunk = True
-            
+
             # Stream the response
             async for text_chunk in self._stream_salesforce_response(session_id, user_message):
                 if first_chunk:
                     await self.stop_ttfb_metrics()
                     first_chunk = False
-                
+
                 # Push each text chunk as it arrives
                 await self.push_frame(LLMTextFrame(text=text_chunk))
-            
+
         except Exception as e:
             logger.error(f"Salesforce context processing error: {type(e).__name__}: {str(e)}")
             import traceback
+
             logger.error(f"Salesforce traceback: {traceback.format_exc()}")
             raise
         finally:
@@ -627,7 +629,9 @@ class SalesforceAgentLLMService(LLMService):
         context = None
         if isinstance(frame, OpenAILLMContextFrame):
             context = frame.context
-            logger.info(f"🔍 Received OpenAILLMContextFrame with {len(context.get_messages())} messages")
+            logger.info(
+                f"🔍 Received OpenAILLMContextFrame with {len(context.get_messages())} messages"
+            )
         elif isinstance(frame, LLMMessagesFrame):
             context = OpenAILLMContext.from_messages(frame.messages)
             logger.info(f"🔍 Received LLMMessagesFrame with {len(frame.messages)} messages")
@@ -680,6 +684,7 @@ class SalesforceAgentLLMService(LLMService):
     def get_llm_adapter(self):
         """Get the LLM adapter for this service."""
         from pipecat.adapters.services.open_ai_adapter import OpenAILLMAdapter
+
         return OpenAILLMAdapter()
 
     async def close(self):