RAG Assistant Agent
One of the difficulties of adopting RAG to a mass audience is lack of understanding of the underline NLP techniques required to produce good queries. With this tool, there is an AI agent that looks at the query and the results to help the user make better queries in the future. For example, If the user never used RAG before, they may ask a vague question. The agent will pick up on this and inform the user. In addition, it will provide suggestion of how to query for better results. This tool is general enough to be easy to adapt with already established RAG pipelines, in addition it is agnostic to data meaning it could be adopted to many fields.
Category tags:
"This is an interesting idea and one that has some merit on its own. Looking at the code, the project does not use Vectara, but instead poses to handle some uncertain/ambiguous questions outside of Vectara. I couldn't get the demo to work, even after plugging an OpenAI key in, so I suspect a bug. I'd have liked to see this take a slightly different direction to be competitive for this particular hackathon where Vectara was required. Rather than assuming the query is ambiguous, perhaps waiting for Vectara to return that the query was ambiguous and then try to guide the user, for example. Still, a neat idea!"
Shane Connelly
Head of Product
For Innovators & Creators