I have a ESP32-S3 AI Smart Speaker set up with HA (docker), piper, faster-whisper, openwakeword and ollama 3b-instruct-q8_0. All latest versions. Also running AI locally on a gpu. I am trying to figure out whether it is waiting for the full response from the AI before starting speaking or whether it is starting to speak whilst the rest is being generated. If I use the Assist feature, then the response is almost immediate. Using the voice assistant, it waits 7 seconds before answering. I tried Debubg for the pipeline, but it only appears to show old queries and not a live view? Any help appreciated
#Testing VA streaming?
1 messages · Page 1 of 1 (latest)
I am interested in this https://github.com/BigBobbas/ESP32-S3-Box3-Custom-ESPHome
The Box is ordered and will arrive in a few days