[BUG-python/deployment] subsequent calls to `/api/v1/metadata-properties/<id>/metrics` are slow

[davidberenstein1957]

Describe the bug When loading the dataset there are subsequent call to the /api/v1/metadata-properties/<id>/metrics-endpoint where the server takes up to 6 seconds to respond, which is too slow to wait with a blank and unresponsive screen.

Stacktrace and Code to create the bug N.A.

Expected behavior I would expect this to be faster

Environment:

• Argilla Version [e.g. 1.0.0]: 1.18
• ElasticSearch Version [e.g. 7.10.2]: N.A.
• Docker Image (optional) [e.g. argilla:v1.0.0]: 1.18

Additional context N.A.

[damianpumar]

I would like to talk about this, because i detected this little thing some days ago, and it affects the frontend performance.

[jfcalvo]

Not sure if this will improve the slowness but we can reduce the number of requests to only one creating a new GET /api/v1/datasets/:dataset_id/metadata-properties/metrics returning all the metrics at once.

What do you think @frascuchon ?

[frascuchon]

This is also related to the nature of the data. @davidberenstein1957, Which kind of properties are taking slower? how can we reproduce it?

[davidberenstein1957]

Hi @frascuchon , I was using the https://huggingface.co/datasets/argilla/text-descriptives-metadata/discussions dataset.

[damianpumar]

• Data fetching too slow, not only for metadata metrics.
• Spike new way to load the metrics.
• Understand why the metrics are too slow.
• Why the metadata-properties does not have the "metrics" why we need to consume other endpoint.
• Query params to load all metrics in same endpoint?
• Set metadata properties by default using "metrics" endpoint on initialization step from SDK ?

@davidberenstein1957 @frascuchon @jfcalvo

[github-actions[bot]]

This issue is stale because it has been open for 90 days with no activity.

[github-actions[bot]]

This issue was closed because it has been inactive for 30 days since being marked as stale.