Within the same conversation, we're interested in changing the endpointing ms. Right now have some messy asyncio code to do it, but was curious if there was a recommended / built-in way to do this (could send to two different deepgram transcription instances but would rather not).
#Dynamic endpointing?
vestal totem
mild forge
I've asked about this. You can toggle numerals (swapping numeric words for numbers) while the stream is open, but I am unsure about other features.
vestal totem
yeah looks like there's def configure(self, config: ToggleConfigOptions) -> None:
"""Sends messages to configure transcription parameters mid-stream."""
self._queue.put_nowait((False, json.dumps({
"type": "Configure",
"processors": config
})))
but the only ToggleConfigOptions in the python sdk at least is numerals
mild forge
I can confirm that it only supports numerals right now.
You could buffer your audio, and drop the client and reconnect with your settings toggled before clearing the buffer. Although, the time you buffer for will off-set your word level timestamp by that amount.