wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[WIP] AWS Nova Sonic service - add send_transcription_frames option

Browse Source
This commit is contained in:
Paul Kompfner
2025-04-30 14:53:22 -04:00
parent 3960c604a4
commit 5e0803479e
2 changed files with 9 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/pipecat/services/aws_nova_sonic/aws.py
View File

@@ -118,6 +118,7 @@ class AWSNovaSonicLLMService(LLMService):
voice_id: str = "matthew", # matthew, tiffany, amy voice_id: str = "matthew", # matthew, tiffany, amy
instruction: Optional[str] = None, instruction: Optional[str] = None,
tools: Optional[ToolsSchema] = None, tools: Optional[ToolsSchema] = None,
send_transcription_frames: bool = True,
**kwargs, **kwargs,
): ):
super().__init__(**kwargs) super().__init__(**kwargs)
@@ -129,6 +130,7 @@ class AWSNovaSonicLLMService(LLMService):
self._voice_id = voice_id self._voice_id = voice_id
self._instruction = instruction self._instruction = instruction
self._tools = tools self._tools = tools
self._send_transcription_frames = send_transcription_frames
self._context: AWSNovaSonicLLMContext = None self._context: AWSNovaSonicLLMContext = None
self._stream: DuplexEventStream[ self._stream: DuplexEventStream[
InvokeModelWithBidirectionalStreamInput, InvokeModelWithBidirectionalStreamInput,
@@ -802,10 +804,12 @@ class AWSNovaSonicLLMService(LLMService):
# Manually add new user transcription text to context. # Manually add new user transcription text to context.
# We can't rely on the user context aggregator to do this since it's upstream from the LLM. # We can't rely on the user context aggregator to do this since it's upstream from the LLM.
self._context.add_user_transcription_text_as_message(text) self._context.add_user_transcription_text_as_message(text)
# Report that some new user transcription text is available. # Report that some new user transcription text is available.
await self.push_frame( if self._send_transcription_frames:
TranscriptionFrame(text=text, user_id="", timestamp=time_now_iso8601()) await self.push_frame(
) TranscriptionFrame(text=text, user_id="", timestamp=time_now_iso8601())
)
# #
# Context # Context

4
src/pipecat/services/aws_nova_sonic/context.py
View File

@@ -94,11 +94,11 @@ class AWSNovaSonicLLMContext(OpenAILLMContext):
f"Unhandled content type in context message: {c.get('type')} - {message}" f"Unhandled content type in context message: {c.get('type')} - {message}"
) )
# There won't be content if this is an assistant tool call entry. # There won't be content if this is an assistant tool call entry.
# We're ignoring those since they can't be loaded into AWS Nova Sonic conversation # We're ignoring those since they can't be loaded into AWS Nova Sonic conversation
# history # history
if content: if content:
return AWSNovaSonicConversationHistoryMessage(role=Role[role.upper()], text=content) return AWSNovaSonicConversationHistoryMessage(role=Role[role.upper()], text=content)
# We're ignoring messages with role "tool" since they can't be loaded into AWS Nova Sonic # We're ignoring messages with role "tool" since they can't be loaded into AWS Nova Sonic
# conversation history # conversation history
logger.error(f"Unhandled message type in from_standard_message: {message}") logger.error(f"Unhandled message type in from_standard_message: {message}")