Databricks is an analytics system and isn't optimized to perform as an OLTP system.  Additionally, Databricks compute can scale to zero if you set it to do so.  This means if you want to use gold/silver data in a real-time way you need to keep a cluster running continuously (I'd set to scale to 1 and then up as needed).  Databricks has a nifty SQL execution API, Statement Execution API: Run SQL on warehouses - Azure Databricks | Microsoft Learn.  But, that is read-only.  There are also ODBC and JDBC drivers you can use to write a traditional API.  If you want OLTP performance, maybe look at Lakebase | Databricks.