Improved
Scoring API now scores a typical request 2.2× faster
10 days ago by ReadMe API
GET /profiles/scoring and GET /jobs/scoring now score a typical request 2.2× faster, and their slowest requests take 27% less time. Each change was checked to return the same results as before.
Action required
- Now: none.
GET /profiles/scoring,GET /jobs/scoring,GET /profiles/searchingandGET /jobs/searchingkeep their parameters and their response. - Later: none. Nothing in this update is deprecated.
What changed
GET /profiles/scoringandGET /jobs/scoringtake less time to score, with the same scores: a typical request (the median) is scored 2.2× faster. Not every request gains as much: the slowest 1 in 100 take 27% less time.GET /profiles/searchingandGET /jobs/searchinganswer as fast as before.
Performance
| Score Profiles and Score Jobs | Before | Now |
|---|---|---|
| Time to score, typical request (median) | 1.13 s | 0.50 s, 2.2× faster |
| Time to score, slowest 1 in 100 (99th percentile) | 4.16 s | 3.00 s, 27% less time |
Search Profiles and Search Jobs answer as fast as before.
Measured on the same production requests, replayed on the previous and the current version: time to score on our side, without the network, one request at a time.
Example
curl -G https://api.hrflow.ai/v1/profiles/scoring \
-H "X-API-KEY: <your_api_key>" \
-H "X-USER-EMAIL: <your_email>" \
--data-urlencode "algorithm_key=<your_algorithm_key>" \
--data-urlencode "board_key=<your_board_key>" \
--data-urlencode "job_key=<your_job_key>" \
--data-urlencode 'source_keys=["<your_source_key>"]'The response keeps its shape and its scores: data.profiles and data.predictions, one [1-score, score] pair per profile.
{
"code": 200,
"message": "Profile scoring results",
"meta": {"page": 1, "maxPage": 42, "count": 1, "total": 42},
"data": {
"predictions": [[0.03323773294687271, 0.9667623043060303]],
"profiles": [{"key": "<profile_key>"}]
}
}