GPT-4 surprisingly excels at Googling (Binging?) to retrieve up-to-date information about current issues. Tools like Perplexity.ai are impressive. Now that we have a highly capable smaller-scale model, I feel like not enough open-source research is being directed towards enabling local models to perform internet searches and retrieve online information.
Did you manage to add that functionality to your local setup, or know some good repo/resources to do so?
I have. You simply parse the prompt for a url and then write a handler to retrieve the page content using whatever language or framework you use. Then you clean it up and send the content along with the prompt to the LLM and do QA over it.