Currently to retrieve a list of text fields from the source data to test a model output we rely on Kibana Data Views. For users who mostly use the Elasticsearch API to ingest data and are not familiar with the concept of data views, it might be an unnecessary friction point. We can improve the UX by providing ad-hoc data views either for

1. all indices containing text fields

or

2. narrowing down the list to indices that utilize ingest pipelines with inference processors of the testing model