Hi @Kaspar –
QuickSight does perform schema and type inference on JSON files during import, including flattening nested JSON structures. However, there are some limitations on the types of JSON structures that are supported. QuickSight supports importing flat JSON files and semi-structured JSON data, and it automatically infers the schema, data types, and flattens the JSON data during import. It also supports JSON records with structures and JSON records with root elements as arrays. However, list attributes and list objects within a JSON record are not supported and will be skipped during import. This could be the reason why some columns are not appearing in your QuickSight dataset, if those columns are nested within list attributes or list objects in the JSON file. To address this issue, it is recommended to use tools like AWS Glue to unnest the JSON data before importing it into QuickSight. These tools can help flatten the nested JSON structures, making all columns accessible in the QuickSight dataset.
This post talks along similar direction.
How many columns do you have? There is a limit of 2K columns on SPICE dataset.