pollen-robotics/marionette

UI: recover to a local dataset on page load and My Moves tab

Never restore a downloaded dataset as active on startup

Add banner pointing to JS version

Remove manual token login/logout — single auth path via desktop app

Refine auto-detection: skip network calls once username known

Use stored HF token for dataset operations (no manual token required)

relieve strict dep version on huggingface-hub (#7)

Revert test_beep_collision_sync.py: collision test is timing sensitive

Revert test_antenna_collision.py: collision test is timing sensitive

Use new antenna neutral position [-0.1745, 0.1745] to reduce shaking

Adding scipy and requests dependencies

Removing yt-dlp dep

Pre-resample audio assets to 16kHz to avoid runtime delay

Replace second intro sound with open_and_record.wav

Improve HF upload discoverability and sharing CTA

Merge branch 'main' of hf.co:spaces/RemiFabre/marionette

Consolidate Marionette app runtime into main module

Add E2E tests for audio indicator, auto-fill name, moves refresh, playback stop

Thumbnail update (#6)

Add images via Git LFS

Fix recording sync, auto-fill name from WAV, refresh moves list

Refactor marionette into modules, fix audio sync, improve tests

Simplify replay compensation: always-on lead model, default lead tuned, and GUI tuning UX

Sync fix

fix(playback): sync audio start with first motion command

fix(playback): warm GStreamer pipeline before starting audio thread

fix(playback): preload move audio before goto so it starts instantly

fix(audio): restore resampling in _play_wav_chunked for move playback

Resample wavs offline, remove runtime scipy resampling from startup

Split lead compensation into separate antenna and head/body values

Speed up intro: shorter pause (0.5s) and audio starts with head rise

fix(playback): sync audio and motion start, enable lead compensation

fix(startup): preload second wav and smooth final torque-off

fix(recording): prevent stuck stop-recording on Mac and slow networks

Replace pleaaaase.wav with please.wav for second welcome message

Rewrite startup optimization doc with detailed explanations

Add startup optimization analysis document

Optimize startup time: pre-resample WAVs, lazy-import scipy and requests

Use pleaaaase.wav as default second welcome message

Fix HF login lost on restart and improve token hint text

Preserve uploaded audio file across recordings

Fix audio playback cutting off at ~80% of total duration

Fix recording blocked after stopping audio playback

Replace play_sound() with push_audio_sample() for stoppable audio

Harden recording against SDK hangs: retries, hard deadline, error guard

Fix recording stuck at 0.0s: timeout SDK calls, add safety net

Add HF manual token login/logout in settings drawer

Merge branch 'frontend-redesign'

Revert "Revert "Redesign frontend: new Studio theme with settings drawer, phase overlay, and tabbed layout""

Add configurable welcome messages (0/1/2) at startup