You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, the API /spend/logs accepts query parameters start_date and end_date which may, on the surface, appear to provide this functionality already. In fact they provide a different feature altogether:
If both parameters are specified, the response is presented as a daily spend aggregate rather than a full list of log events.
If the parameters contain a time component as well as the date, the request fails.
If either parameter is omitted, the other is ignored.
Motivation, pitch
I'd like to process fine grained usage data based on the full content of the spend logs, which requires datetime-level filtering of those logs. Currently the only way I can do this is by querying the DB directly.
Are you a ML Ops Team?
Yes
Twitter / LinkedIn details
No response
The text was updated successfully, but these errors were encountered:
Oh wow thanks, I hadn't noticed there's a whole separate (and much more intuitive) endpoint for the UI to access the logs. The issue with this one aside from the fact that it's not part of the documented API, is that it appears to return results only if prompt and response bodies are logged, which is undesirable for me (I'm only interested in the metadata).
OK I see. That'll be useful when it's merged then. Is there a rationale for implementing this in a UI-specific API rather than improving the "official" API? How stable will the UI-specific API be if I decide to integrate with it?
The Feature
I'd like to augment the
/spend/logs
API with a time range filter that allows you to retrieve individual log entries between two datetimes.PR here: #7956
Sidenote
Currently, the API
/spend/logs
accepts query parametersstart_date
andend_date
which may, on the surface, appear to provide this functionality already. In fact they provide a different feature altogether:Motivation, pitch
I'd like to process fine grained usage data based on the full content of the spend logs, which requires datetime-level filtering of those logs. Currently the only way I can do this is by querying the DB directly.
Are you a ML Ops Team?
Yes
Twitter / LinkedIn details
No response
The text was updated successfully, but these errors were encountered: