dv-pipecat-ai 0.0.85.dev10__tar.gz → 0.0.85.dev12__tar.gz

{dv_pipecat_ai-0.0.85.dev10 → dv_pipecat_ai-0.0.85.dev12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dv-pipecat-ai
-Version: 0.0.85.dev10
+Version: 0.0.85.dev12
 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.dev10 → dv_pipecat_ai-0.0.85.dev12}/src/dv_pipecat_ai.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dv-pipecat-ai
-Version: 0.0.85.dev10
+Version: 0.0.85.dev12
 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.dev10 → dv_pipecat_ai-0.0.85.dev12}/src/dv_pipecat_ai.egg-info/SOURCES.txt

@@ -370,6 +370,8 @@ src/pipecat/services/rime/tts.py
 src/pipecat/services/riva/__init__.py
 src/pipecat/services/riva/stt.py
 src/pipecat/services/riva/tts.py
+src/pipecat/services/salesforce/__init__.py
+src/pipecat/services/salesforce/llm.py
 src/pipecat/services/sambanova/__init__.py
 src/pipecat/services/sambanova/llm.py
 src/pipecat/services/sambanova/stt.py
@@ -442,6 +444,7 @@ src/pipecat/utils/__init__.py
 src/pipecat/utils/base_object.py
 src/pipecat/utils/logger_config.py
 src/pipecat/utils/network.py
+src/pipecat/utils/redis.py
 src/pipecat/utils/string.py
 src/pipecat/utils/time.py
 src/pipecat/utils/utils.py

{dv_pipecat_ai-0.0.85.dev10 → dv_pipecat_ai-0.0.85.dev12}/src/pipecat/frames/frames.py

@@ -451,12 +451,14 @@ 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"]
     content: str
     user_id: Optional[str] = None
     timestamp: Optional[str] = None
+    message_id: Optional[int] = None
 
 
 @dataclass
@@ -510,6 +512,17 @@ class TranscriptionUpdateFrame(DataFrame):
         return f"{self.name}(pts: {pts}, messages: {len(self.messages)})"
 
 
+@dataclass
+class TranscriptDropFrame(DataFrame):
+    """Frame indicating previously emitted transcript chunks should be discarded.
+
+    Parameters:
+        transcript_ids: List of frame/message identifiers to drop.
+    """
+
+    transcript_ids: List[int]
+
+
 @dataclass
 class LLMContextFrame(Frame):
     """Frame containing a universal LLM context.

{dv_pipecat_ai-0.0.85.dev10 → dv_pipecat_ai-0.0.85.dev12}/src/pipecat/processors/aggregators/llm_response.py

@@ -51,6 +51,7 @@ from pipecat.frames.frames import (
     StartFrame,
     StartInterruptionFrame,
     TextFrame,
+    TranscriptDropFrame,
     TranscriptionFrame,
     UserImageRawFrame,
     UserStartedSpeakingFrame,
@@ -446,6 +447,7 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
         self._latest_final_transcript = ""
         self._last_user_speaking_time = 0
         self._last_aggregation_push_time = 0
+        self._pending_transcription_ids: List[int] = []
 
     async def reset(self):
         """Reset the aggregation state and interruption strategies."""
@@ -453,6 +455,7 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
         self._was_bot_speaking = False
         self._seen_interim_results = False
         self._waiting_for_aggregation = False
+        self._pending_transcription_ids.clear()
         [await s.reset() for s in self._interruption_strategies]
 
     async def handle_aggregation(self, aggregation: str):
@@ -548,7 +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:
@@ -614,10 +618,18 @@ 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()
         self._user_speaking = True
@@ -662,6 +674,7 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
             return
 
         self._aggregation += f" {text}" if self._aggregation else text
+        self._pending_transcription_ids.append(frame.id)
         # We just got a final result, so let's reset interim results.
         self._seen_interim_results = False
 
@@ -791,6 +804,7 @@ 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

{dv_pipecat_ai-0.0.85.dev10 → dv_pipecat_ai-0.0.85.dev12}/src/pipecat/processors/transcript_processor.py

@@ -20,6 +20,7 @@ from pipecat.frames.frames import (
     EndFrame,
     Frame,
     StartInterruptionFrame,
+    TranscriptDropFrame,
     TranscriptionFrame,
     TranscriptionMessage,
     TranscriptionUpdateFrame,
@@ -44,6 +45,7 @@ class BaseTranscriptProcessor(FrameProcessor):
         super().__init__(**kwargs)
         self._processed_messages: List[TranscriptionMessage] = []
         self._register_event_handler("on_transcript_update")
+        self._register_event_handler("on_transcript_drop")
 
     async def _emit_update(self, messages: List[TranscriptionMessage]):
         """Emit transcript updates for new messages.
@@ -57,6 +59,19 @@ class BaseTranscriptProcessor(FrameProcessor):
             await self._call_event_handler("on_transcript_update", update_frame)
             await self.push_frame(update_frame)
 
+    async def _handle_transcript_drop(self, frame: TranscriptDropFrame):
+        """Handle transcript drop notifications by removing stored messages."""
+        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
+            ]
+
 
 class UserTranscriptProcessor(BaseTranscriptProcessor):
     """Processes user transcription frames into timestamped conversation messages."""
@@ -72,9 +87,15 @@ class UserTranscriptProcessor(BaseTranscriptProcessor):
 
         if isinstance(frame, TranscriptionFrame):
             message = TranscriptionMessage(
-                role="user", user_id=frame.user_id, content=frame.text, timestamp=frame.timestamp
+                role="user",
+                user_id=frame.user_id,
+                content=frame.text,
+                timestamp=frame.timestamp,
+                message_id=frame.id,
             )
             await self._emit_update([message])
+        elif isinstance(frame, TranscriptDropFrame):
+            await self._handle_transcript_drop(frame)
 
         await self.push_frame(frame, direction)
 

dv_pipecat_ai-0.0.85.dev12/src/pipecat/services/salesforce/__init__.py

@@ -0,0 +1,9 @@
+#
+# Copyright (c) 2024–2025, Daily
+#
+# SPDX-License-Identifier: BSD 2-Clause License
+#
+
+from .llm import SalesforceAgentLLMService
+
+__all__ = ["SalesforceAgentLLMService"]