@CosmicEggEarth@thinkymachines The rerun-with-hot-spares pattern is interesting. We've been trying to get it right first time on a single stream - feels like the wrong frame. Did you find a clean way to pick the winner across the 3 reruns, or is it manual?
Yeah that makes total sense, we tried building something like this and it felt a little sluggish for our use (perhaps we built it wrong), but definitely seemed the right direction
Was your latency (time to first audio) good? how did you deal with ASR errors and therefore llm response error in the first part being corrected by the second thread?
@kwindla Tried very impressive, single word responses don't seem to work though e.g. "yep". Is this a prompting issue or is there another way of solving this?