Hey, I did some tests api.deathwing.me as the endpoint. I think you lost the speed of hivelytics while optimizing.
Acidyo's full build takes 8 minutes in your 2.0, in hivelytics it takes 22 seconds.
That doesn't necessarily mean that I am not going to implement ideas from this.
Browser caching and top voters are solid ideas.
Also, if you end up implementing browser cache 23 hours is probably too long, I am thinking one to two hours should be plenty as most people will stop interacting with this after a couple hours... That would reduce the API load even further and will be even faster on the second load+
I've already implemented browser cache, it is a bit smarter on how it handles the cache.
Every time you load a profile it does a liveliness check, and if for example you have voted on someone, or multiple someones it extrapolates how much it needs to update etc.
I guess it depends on which API endpoint is used. I tested with the default API endpoint: api.openhive.network.
Here are my results for 2.0 for initial load (full build):
Yes, the initial load (full build) is slower 5.5 minutes because I implemented a gentler rate to hit the API server so I don't get any 429 errors. Original Hivelytics gets a lot of 429 errors and is 3.7 minutes in the full build mode:
Note the 0 errors in 2.0 and 2219 errors in Hivelytics trying to pull Acidyo...