blaiz2k/Qwen3.8-27B-Uncensored-Demo

Show the resolved MODEL_ID in the log and the UI header

Correct reasoning-effort measurements (were taken against an overridden MODEL_ID)

Remove prompt diagnostic; keep the reasoning_effort probe

Add DEBUG_PROMPT diagnostic for chat-template kwarg handling

Verify reasoning_effort reaches the template at startup, with system-prompt fallback

Default reasoning to off; document measured effort trade-off

Add reasoning-effort control, fix truncated replies, enlarge chat window

Fix multi-turn history parsing and text attachments

Handle unsupported/corrupt image uploads with a clear error; add HEIC support

Add measured capability benchmarks vs base model

ZeroGPU: xlarge slice for 54.7GB model, duration 120, theme on launch()

Measured behaviour: refusal rate and KL from the release run

Fix ZeroGPU startup: theme kwarg, bf16 default, pinned deps

Delete __pycache__/app.cpython-312.pyc with huggingface_hub

Read MODEL_ID from env, default to the release repo

Upload folder using huggingface_hub

Upload app.py with huggingface_hub

Upload folder using huggingface_hub

initial commit