why is it that you would want to trigger a web search if one of the nodes is irrelevant? Let's say the retriever extracts 3 nodes (pages in this case) and one one of them is on a different topic than the query's topic. But the other 2 are spot on. What is the purpose of searching the web? Finding other information than the one in the relevant nodes? Couldn't we search the web anyways?
why is it that you would want to trigger a web search if one of the nodes is irrelevant? Let's say the retriever extracts 3 nodes (pages in this case) and one one of them is on a different topic than the query's topic. But the other 2 are spot on. What is the purpose of searching the web? Finding other information than the one in the relevant nodes? Couldn't we search the web anyways?
Was about to comment this. I would say having a threshold value for relevancy check would work way better in terms of efficiency