Hey @SalmanJafferCFA ! Can you clarify what you are asking? This example did use fully open source/public data (you can find sources in the repo).
Are you asking about the difference between using an approach like this and just asking an LLM model and relying on its native search? There are a few reasons you might want to do so:
In some cases, it's going to be cheaper to use specialist models that are highly optimized for retrieval and answering and reasoning over data.
You might be in a situation where you need to be absolutely certain that the answers are backed up by a particular canonical source, and not just what happens to be retrieved from the Internet with citations.
Relatedly, different AI models and agents have different approaches to searching and to using information from the internet and can return different answers from each other. Again, if you have canonical, trusted sources and you need to know that the answers are rooted in particular data and particular sources, then you want an approach that is retrieving from your preferred data.
Though these are not comparing retrieval versus model-native open web search, they are comparing how different models and agents perform on retrieval tasks.
