Drag in a step, pick the columns, see the result. Every step compiles to SQL and runs as pushdown in your own Snowflake, so the canvas is a way of writing SQL rather than a layer on top of it – and the SQL is always there to read.
Steps compile to a single SQL statement and execute in your own Snowflake. Refyner never holds your data – only the dataflow definition and metadata.
Every step shows the resulting rows and the SQL behind them, so you check the output while you build rather than after it ships.
Pre-flight checks confirm every column exists, the grain is what you expect, and no join will fan out unexpectedly. AI assistance will be able to draft and explain steps – coming soon.
Yes. Any step can be a SQL step, and the SQL generated for visual steps is always visible. The canvas and the SQL are the same dataflow, not two modes.
No. Transformations execute in your own warehouse as pushdown SQL – see the security model.
Our production models could run to 100+ steps. Because execution is pushdown, throughput is a function of the Snowflake warehouse you point it at, not of Refyner.
Yes. A dataflow's output is a Snowflake table, so another dataflow can read it – and production models give you a working starting point to adapt.
Connect Snowflake, drag in a few steps, and check the generated SQL against what you would have written.
Refyner