Survey of Interactive Document Query Systems Using RAG and LangChain

This research is to build a chatbot based on Retrieval-Augmented Generation (RAG) to produce the right answers. Rag-mini-bioasq Dataset from HuggingFace is used for ground truth answers, generated answers (LLM), retrieved contexts (ChromaDB, Nomic-embed-text) . For the retrieval orchestration, the system uses LangChain, ChromaDB for storage of vector, and Deepseek r1 1.5B for linearity between long context queries up to 128,000 tokens. Chain of Thought (CoT) and Chain of Notes are techniques that increase multi-hop query resolution with reduced hallucinations and improved accuracy.