fix: update decommissioned Groq model to llama-3.3-70b-versatile

feat: switch AI backend to Groq for ultra-fast medical responses

fix: explicit model ID for Gemini 1.5 Flash

fix: switch to Gemini 1.5 Flash for higher rate limits and add error handling

refactor: update to modern google-genai library and fix deprecation warnings

Robust deployment: lazy data loading and package structure

Switch to Gemini 3 Flash model

Initial commit: Medical AI Agent with Gemini and OpenMed datasets