Currently, the Assistant will respect your Kagi-wide rewrite rules as long as Personalization is enabled, which is usually what you want.
However, in various cases, the rewrites I setup are to privacy-focused frontends like redlib instances, which usually put up Anubis to combat scrapers, which means Assistant hits a wall waiting to solve Anubis, when accessing the original reddit.com link would have worked.
Similarly, some sites like x.com are fine to access directly as a user, but effectively blind for scrapers, and in those cases I'd like to set separate rewrite rules to allow programmatic access.
One of:
- A separate list of URL rewrites which only apply to Assistant
- An additional toggle in the current URL rewrites which determine if they should apply to Assistant as well
- Some other UI I haven't thought of that effectively allows this behavior
As a workaround you can currently set custom instructions like this:
When you see URLs:
From redlib.catsarch.com -> rewrite host to www.reddit.com
From x.com/:username/status/:tweetid OR twitter.com/:username/status/:tweetid -> rewrite to api.fxtwitter.com/2/status/:tweetid
However, it feels a bit silly to be relying on the model to do this to perform what should be a deterministic pattern match, and also undoing the work that the URL rewrite logic Kagi already has.
