processors(realtime-ai): add assistant before output transport
Cartesia can do word-to-word output instead of full sentences. This means that for properly adding things into the context we need to add it before the transport, otherwise some words might be lost.
This commit is contained in:
@@ -166,8 +166,8 @@ class RealtimeAIProcessor(FrameProcessor):
|
||||
self._tma_in,
|
||||
self._llm,
|
||||
self._tts,
|
||||
self._tma_out,
|
||||
self._transport.output(),
|
||||
self._tma_out
|
||||
])
|
||||
self._pipeline = pipeline
|
||||
|
||||
|
||||
Reference in New Issue
Block a user