How can Kagi be so fast and customizable?

Question

Each customization I do to Kagi means the next search query has to be totally different. That might mean the UI also has to change to add/remove things that certainly require an index/DB search.And yet, I always get sub 200ms first-byte responses.How is that even possible? Is that edge computing? How would you architect such a thing?

ecf · Accepted Answer

Well this is easy. Kagi is fast because of how strictly they limit the number of searches one can perform.Take the standard, $5 plan. It gives you 300 searches PER MONTH. That&rsquo;s 10 per day. I&rsquo;m sure most of us do 10 searches within minutes while working.

hanifak · Answer

https://help.kagi.com/kagi/search-details/search-speed.html

smcleod · Answer

I&rsquo;m sure part of it is secret source but Vlad (CEO) is a lovely, helpful chap - I&rsquo;m sure if you dropped an open question on the Kagi forums, he or someone else might be able to point you to some of the technologies or patterns used.

necovek · Answer

Edge computing would ensure that latency is the same in all regions: eg. if they could get it down to 30ms somewhere, they could get it to 30ms everywhere (kind of, ofc).
Your singular case does not really prove much: maybe you are just in the same area their DC is?
Still, a full, customized web search result in 200ms (though, first byte is not really indicative of that) would be impressive.

How can Kagi be so fast and customizable?

Well this is easy. Kagi is fast because of how strictly they limit the number of searches one can perform.
Take the standard, $5 plan. It gives you 300 searches PER MONTH. That’s 10 per day. I’m sure most of us do 10 searches within minutes while working.

https://help.kagi.com/kagi/search-details/search-speed.html

I’m sure part of it is secret source but Vlad (CEO) is a lovely, helpful chap - I’m sure if you dropped an open question on the Kagi forums, he or someone else might be able to point you to some of the technologies or patterns used.