Change Details

**Context** All liftwing endpoints support a [[ https://wikitech.wikimedia.org/wiki/Machine_Learning/LiftWing/Usage#Request_a_bearer_token | wme ]] tier rate limit which is 200 K requests per hour . We call ML APIs with a deadline (configurable). **To do** [] Test the wikidata revertrisk for scale. Achieve at least 150K requests per hour. (Tip: you can reuse existing liftwing scale test under experiments repos). [] Capture response time for 200 returned calls (passing calls) **Acceptance criteria** [] Share a report on scale achieved and latency distribution. [] Connect with Francisco and ML team if for most (90%) 200 returned calls, the latency is not <= 500 ms.

**Context** All liftwing endpoints support a [[ https://wikitech.wikimedia.org/wiki/Machine_Learning/LiftWing/Usage#Request_a_bearer_token | wme ]] tier rate limit which is 200 K requests per hour . We call ML APIs with a deadline (configurable). Run the test from an aws instance. **To do** [] Test the wikidata revertrisk for scale. Achieve at least 150K requests per hour. (Tip: you can reuse existing liftwing scale test under experiments repos). [] Capture response time for 200 returned calls (passing calls) **Acceptance criteria** [] Share a report on scale achieved and latency distribution. [] Connect with Francisco and ML team if for most (90%) 200 returned calls, the latency is not <= 500 ms.

**Context** All liftwing endpoints support a [[ https://wikitech.wikimedia.org/wiki/Machine_Learning/LiftWing/Usage#Request_a_bearer_token | wme ]] tier rate limit which is 200 K requests per hour . We call ML APIs with a deadline (configurable). Run the test from an aws instance. **To do** [] Test the wikidata revertrisk for scale. Achieve at least 150K requests per hour. (Tip: you can reuse existing liftwing scale test under experiments repos). [] Capture response time for 200 returned calls (passing calls) **Acceptance criteria** [] Share a report on scale achieved and latency distribution. [] Connect with Francisco and ML team if for most (90%) 200 returned calls, the latency is not <= 500 ms.