surajamit/GenAI-RAG-LLM ? reverse-engineered prompt

Build me an enterprise AI platform in Python that can answer questions over multiple PDFs, evaluate resumes for ATS fit, help with research, and turn natural language into SQL. I want it to feel like a real GenAI product, with hybrid retrieval using both embeddings and a knowledge graph, plus an agent layer that can decide how to answer each request.

Please set it up as a microservices based app with document ingestion, chunking, embedding generation, graph construction, retrieval, response synthesis, and basic evaluation tools. Use PostgreSQL for metadata, Redis for caching, and a vector store for semantic search. Include a simple way to run benchmarks and experiments for things like retrieval quality, ATS scoring, and SQL accuracy, and make sure the code is organized so it is easy to extend later.

If anything is unclear, look up current docs online and keep the implementation practical and runnable.