Fix OOM: single-stage Dockerfile (no Node build), pre-built frontend

fix: robust response parsing, /no_think for Qwen

fix: argparse defaults from env vars, health route ordering

feat: add lightweight server-side analytics tracking

Update metadata with huggingface_hub

fix: default art_model to 14B in rl_bridge.py

fix: default art_model to 14B in server.py

fix: llm_agent.py

fix: server.py

fix: update RewardPanel.tsx

fix: update server.py

feat: lock to trained Qwen 2.5 14B model, disable model/mode switching

fix: add HF Space metadata to README.md

feat: multi-stage Dockerfile for HF Space (React + FastAPI on port 7860)

feat: multi-stage Dockerfile for HF Space (React + FastAPI on port 7860)

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

initial commit