BhargavaKandala/LangGraph-Multi-Agent-RAG-System ? reverse-engineered prompt

Build me a Python app that uses LangGraph to run a simple multi agent research workflow.

I want one agent that acts like a researcher and looks up fresh information on the web, and another agent that acts like a writer and turns that research into a clean, well formatted answer. The writer should only use what the researcher found, so it feels like a proper RAG style flow instead of making things up.

Please make it run locally with Ollama and Llama 3, and use DuckDuckGo search for the web lookup step. I want a shared state passed between the agents so the research gets handed off smoothly, and a single main script I can run to start the whole thing.

If you need to check current docs online while building it, go for it. Keep the setup simple so I can install dependencies, start Ollama, and run the app from the terminal.