feat: add privacy filtering and execution modes

fix: keep synthesis progress active until result

fix: parse qwen thinking json output

fix: enable qwen thinking with larger token budget

fix: improve qwen json generation

fix: support batch-encoded model inputs

fix: compact dense trail timeline

fix: compact dense trail timeline

fix: harden codebook and model fallback UX

fix: render trace report timestamps as readable clock times

feat: stream real pipeline progress to drive the analyze checklist

fix: read uploaded trace file fields whether dict or FileData

feat: serve designer React frontend via gradio.Server on ZeroGPU

feat: improve qwen zerogpu ux

feat: improve qwen zerogpu ux

feat: improve qwen zerogpu ux

fix: add zerogpu entrypoint

feat: enable oauth-backed model assist

feat: add small model assisted analysis

chore: remove local cache artifacts

feat: implement trace field notes mvp

initial commit