Fix: move score clamping to finally block so it always runs

Clamp task scores to strictly (0,1) range

Fix openai/httpx proxies conflict: upgrade openai, pin httpx<0.28

Extreme robustness fix: match checklist and fix asyncio conflict

Restore server wait and add global exception handler

Robust OpenAI client initialization and proxy fix

Fix proxy API compliance

fix(inference): restore try-except error handling for evaluation resilience

fix(inference): remove LLM error fallback and update default model to gpt-4.1-mini

fix(inference): initialize OpenAI client directly with os.environ

Deep debug: Fix LLM Proxy Authentication by loading 'API_KEY' instead of 'HF_TOKEN'

Fix Phase 2 evaluation unhandled exception for missing OpenAI API Key by adding dummy fallback and try/except

Deep debug: Runtime patch HfFolder compatibility for forced huggingface_hub upgrade

Pin huggingface-hub<1.9.0 to fix Gradio crash on HF deploy

Fix HF Docker dependencies adding gradio and fixed multipart

Fix Phase 1 validation, add multi-mode server/app.py, update dependencies

Add professional root landing page to API

Allow env/ package and fix ModuleNotFoundError

Final cleanup: Remove temporary logs and error files

Resolve merge conflict in README.md

initial commit

Initial commit