Support Simple/Naive RAG

[abhi1092]

Add support doing RAG directly on retreived context without any agentic pipelines.

This can be done with Response Model compact (default option)

Here is documentation
https://docs.llamaindex.ai/en/stable/module_guides/deploying/query_engine/response_modes/

And here is usage
https://github.com/DS4SD/docling/blob/main/examples/rag_llamaindex.ipynb