As a part of considering migration to liftwing, we want to test out whether the MVP liftwing language-agnostic revertrisk API can operate at the following scale and latency:
- able to hold 200 K requests/hr
- the response time is in the ball park of ~200 ms
Steps:
- Authenticate using OAuth as mentioned in the doc
- Switch to wme tier as mentioned in the doc above.
- Get random article using Random API and 10 latest revisions of the article as follows:
https://en.wikipedia.org/w/api.php?action=query&prop=revisions&titles=Green_Island_(Rideau_River)&rvlimit=10&formatversion=2&format=json&rvprop=ids
- Call revertrisk API for each revision as follows:
POST https://api.wikimedia.org/service/lw/inference/v1/models/revertrisk-language-agnostic:predict with body { "rev_id": 1149319610, "lang": "en" }
- Gather revision id : project identifier : response time data (in csv or db). Calculate what percentage of response time are: a) 200 ms and under b) 300 ms and under c) 500 ms and under d) above 500 ms