feat: implement chat API routes, note generation logic, and model loading utilities

feat: initialize FastAPI application structure with Pydantic configuration and model loader service

feat: add singleton AI model loader and note generation service

feat: implement AI-driven recommendation service and integrate Qwen2.5/mDeBERTa model management into FastAPI lifecycle

feat: implement local model inference with Qwen2.5 and mDeBERTa, and add recommendation system service.