refactor: system/user prompt split + robust JSON extraction

fix: localize condition names + add edge-tts dependency

feat: TTS listen button + follow-up Q&A chat

feat: 3-step agentic pipeline — A/B comparison, red-flag triage, SOAP note

feat: Anatomical Region Selector — body map + dropdown inject region into prompt

feat: multilingual support (ZH/ES/FR/JA) + remove all mock mode

refactor: switch to AMD Dev Cloud vLLM backend with Qwen2.5-VL

Production update: full Gradio app with AMD branding, mock+real mode, bilingual support