Browse with WebFetch
WebFetchtakes aurland aprompt. It fetches the page, converts it to markdown, and returns what a small model extracts from it for your prompt.- Ask for exactly the fact you need in the prompt. A narrow prompt comes back shorter than a broad one.
- To search Wikipedia, fetch a search URL such as
https://en.wikipedia.org/w/index.php?search=<query>&fulltext=1&ns0=1, read the result titles, then fetch the article itself. - To read an article directly, fetch
https://en.wikipedia.org/wiki/<Title>. - Other language wikis work the same way with their own subdomain, for example
https://de.wikipedia.org/wiki/Berlin. - Follow up with another fetch when the first answer is incomplete. Each call is a fresh fetch, so include everything the prompt needs.