Add WebSocket reconnection and increased timeouts for HF Space

Fix outer exception [END] log format to match spec

Pass task_level to reset() for correct difficulty when using HF Space

Fix: fallback to HF Space when LOCAL_IMAGE_NAME not set (judges don't set it)

Reorder endpoints to match OpenEnv spec, add Gemma-4-26B to scores

Restore HF README with YAML header

Clean README for GitHub (no HF YAML header)

Add Gemma-4-26B baseline scores and trajectory proof

Add OpenAI sync import for checklist compliance

Fix: restore LOCAL_IMAGE_NAME default, wrap main in try/except

Fix: match submission checklist — defaults only for API_BASE_URL and MODEL_NAME

Fix: open README link in new tab (iframe blocked)

Clean up baseline table formatting, inline hackathon note

Update landing page: scientific description + README link

Elevate README: scientific abstract, real-world applications, hackathon acknowledgment

Fix: deterministic language in baseline scores

Update baseline scores: Gemma-4-31B, Qwen 3.6+, MiniMax M2.7

Add baseline trajectory logs as proof (Gemma-4-31B, Qwen 3.6+, MiniMax M2.7)

Tighten DIC grader: require critical flagging + reference range queries

Fix: accept task_level in reset() for proper level selection via API/WS

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload Dockerfile with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload server/app.py with huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

initial commit