cleaned data and changed default

Add trained models (int8 cpu + fp32 gpu) to model dropdown

Fix audio players accumulating across runs (fixed pool + explicit clear)

Show ALL detected turns (dynamic gr.render, remove 12-turn audio cap)

Rename module to turn_sim.py; cpu-only model; unified params box

Add end-of-recording trigger (run Smart Turn on full audio even without a VAD stop)

Deploy pipecat VAD + Smart Turn v3.2 simulator

initial commit