fix: preload models on startup and increase LLM timeout

fix: dynamically bind to PORT and specify port 8000 in render.yaml

“…I”

hoàn thiện Multimodal RAG chatbot với FastAPI Backend & Web UI