A high-performance RAG application for chatting with documents, built on Llama 3.3 served through SambaNova's lightning-fast inference engine. It uses LlamaIndex for orchestration and Qdrant as the vector store, with a Streamlit front end for uploading documents and querying them.