does the flux model work as expected for any one h...
# agents
b
does the flux model work as expected for any one here? for me even with incomplete sentences its not giving any end of turn delay, if we say incomplete sentence with high eot_threshold value like 0.7 and give eot_timeout_ms as 3000, then we should be getting is_final: true after 3000 milliseconds right?
l
If the sentences are incomplete, I would imagine the model does not consider the turn finished, and the 3000ms timer would not start during talking, at least that's how I would interpret https://developers.deepgram.com/docs/flux/quickstart#end-of-turn-detection-parameters
b
@lively-winter-88668 i think its like -> if the sentences are incomplete (meaning the end turn probability is less than eot_threshold value given ) it should wait for 3000 seconds maximum before returning is_final: True. because if the sentence is complete there is no point of waiting more right, but regardless of that i dont get the 3000 ms of wait in any case