Skip to content

Retention Policy

Retention policy is defined with creation of a dataset and specifies the time window subset of the data from which the model is trained. Let's say we create a dataset with retention policy of one day. The model will be trained only on a day old data and will disregard the data that is older then one day. The retention policy is not to be confused with deletion of the data - this policy only defines the training set and not when the data will be deleted.

If you use shared Autosuggest models, we recommend maintaining at most a 6-month retention window to ensure you comply with Visma’s customer owned data policy.


POST v1/<dataset-type>:create
<dataset-type> can be substituted with bank, scanned-invoice and electronic-invoice-line depending on the data structure needed.

Authorization - Bearer Token

Token: demo

Body - raw (json)

    "name": "dataset-name",
    "tags": [
    "retention_policy": {
        "max_days": 90
    "samples": [...]