I my previous post I should have mentioned, for completeness, that this is not my first RAG application - I developed RAG as part of the desktop Word Add-In, using direct OpenAI API and Pinecone vector DB. I was quite disappointed with this solution: I found responses from OpenAI API unpredictable timewise and Pinecone DB features and API immature. I could not see how corporate users would choose this solution from the point of view of security, reliability, flexibility, and availability. RAG is special - the user must entrust the corporate private corpus of data to the service provider. I would not let OpenAI and Pinecone have it. So, by all means I trust Microsoft Azure way more, as all data resides in private account and baked up in variety of geo-locations.
Associate VP AI & Automation. Problem Finder, LLM Teacher. Department of Unreasonable Asks
1moStill fighting the assimilation since Win95 - more power to you Neal Krawetz