Closed 6DiegoDiego9 closed 1 year ago
For now it only returns the most semantically relevant excerpt from each page
is "the most semantically relevant" the part that I see in the prompt? or the real prompt is invisible and longer than the visible one?
Yes, it's only the parts that are returned from the backend.
Ah, including the partial (truncated by "...") final sentences?
Based on my own experiments and what I learned so far from OpenAIs documentation, basically, what the prompt is showing you is the entire thing that the model is receiving, no more, no less.
So the truncated
parts and the data inside the links are all ignored when it is returning the results.
For now it only returns the most semantically relevant excerpt from each page Im hoping to grab the H1 - H4 Titles from web pages to build out similar outlines. Is there a reason for limiting what content is captured from each result? If this is due to server resources can this plugin be self hosted? Im hoping chatbpt can use live data to better answer questions answered within websites it has never crawled before. Thanks
Does it read the entire content of the web pages or only the first lines (those displayed in the prompt ending with "...") for each page?