Summary

In a fast-evolving world, gathering information rapidly for decision-making provides a competitive advantage. Dynamic RAG is one way to bring AI into meeting rooms with rapid and cost-effective AI. We built a system that simulated the need to obtain answers to hard science questions in a daily meeting. After installing and analyzing the environment, we downloaded and prepared the SciQ dataset, a science question-and-answer dataset, to simulate a daily meeting during which hard science questions would be asked. The attendees don’t want to spend their time searching the web and wasting their time when decisions must be made. This could be for a marketing campaign, fact-checking an article, or any other situation in which hard science knowledge is required.

We created a Chroma collection vector store. We then embedded 10,000+ documents and inserted data and vectors into the Chroma vector store on our machine with all-MiniLM-L6-v2. The process proved cost-effective...