Agentic RAG Socratic Tutor versus Plain RAG

Is agentic RAG really better than plain RAG? This project builds a science tutor two ways with the same hybrid retriever (Chroma + BM25, reciprocal-rank fusion) and the same answer prompt: plain RAG as a LangChain LCEL chain (retrieve once, answer once) and an agentic tutor as a LlamaIndex Workflow (rewrite the query, grade the passages, corrective or multi-hop re-retrieval, answer with a quoted evidence sentence). A Socratic hint tool answers a wrong student answer with a guiding question. qwen2.5:3b and llama3.2:3b were served by Ollama on a second laptop (RTX 2050 4 GB, 32 GB RAM) over Wi-Fi; the code ran on a laptop with RTX 2050 4 GB (Ryzen 7, 8 GB RAM).

NOTE: Learning aid only. Answers and hints come from a small language model over a small science corpus and can be wrong; check them with your teacher or textbook.

Category: Department of Computer Science and Engineering / Agent Development

Agent Development Project AG07 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.