Hi,

I'm no expert in DLT, and I've only been using it for a few months. My reading of docs so far suggests streaming into bronze all your raw data using an autoloader and then using apply_changes into (SCD1/2) in silver.

Finally, in the gold layer, use materialized views; as you correctly pointed out, you can't use appy_changes target as a source for another SCD1/2 target. DLT is quite opinionated in its approach, which is probably not a bad thing.

For everything else, there is always structured streaming, CDC foreachBatch() etc.

See: https://docs.databricks.com/en/delta-live-tables/cdc.html#

Regards

Toby.
https://thedatacrew.com