New limits for public HTTP API

Hi, we now have several servers running in the cloud. We should now have enough computing to answer all requests sent to our API, no matter whether they come from our homepage, our add-ons, or third-party add-ons or websites. Still, we regularly experience slowdowns, as shown in this graph. It shows the access time for a very simple request, and still it’s sometimes slower than 2 seconds (the graph stops at 2 seconds, so it doesn’t show how much slower):

To make sure everyone has a good, fast, and reliable experience, there will be a new limit for using the public API: you can only send requests with up to 75KB text data per minute. The old limits of requests per minute and maximum request size are not changed. All limits are documented at Public HTTP Proofreading API - LanguageTool Wiki.

If you’re a API user and the new limits are a problem for you, let me know. Keep in mind that you can easily set up your own server, as it has always been the case: HTTP Server - LanguageTool Wiki

I had to add another check: if more than 40% of the words cause a rule match, the checking on languagetool.org will now stop with an error. That’s because these are basically all cases where the wrong language is selected, causing the spell checker to complain about almost every word. This used to cause some wasted CPU power, as generating a suggestion for every “typo” is quite costly.