Add Ollama fallback support to ai_fixer

Add Hugging Face Inference API fallback for AI fixer

Fix builtin syntax fixer for inline one-line headers and validate TGI fixes

Fix TGI endpoint plumbing and use full TGI runtime image in Dockerfile

Add TGI integration for cloud LLM serving

Serve React frontend from FastAPI backend for one-click access

Finalizing CodeArena RL Benchmark: frontend improvements, GRPO training scripts, and cleaned environment

Complete all tasks: Adaptive curriculum, GRPO, React frontend, LLM-as-a-judge

Upload meta

fix: reset task_id parsing, grader tuple crash fallback, and inference score output

fix: clamp scores to strict (0.001, 0.999) range

fix: openenv.yaml tasks graders + strict score bounds

fix: add safe_reward() clamp at every reward return point

Add normalize_reward(); enforce strict (0,1) on all paths

Update grading logic to strict (0,1) ratio fallback

fix: OpenEnv multi-mode compliance — add main() entrypoint and uv.lock

Production-ready: add server/app.py with fallback-safe /reset, fix Dockerfile, add HF metadata, add task JSON files

Initial commit