ViewsAlpha

Incremental Views

Define a view once in dbt-style SQL. Melt maintains it incrementally off the same Snowflake CDC stream that keeps your lake fresh — sub-second freshness, queryable from any engine, no warehouse credits to keep it warm.

What it’ll do

The capabilities, at a glance.

dbt-compatible model files

Drop a .

Strategy auto-selection

Melt classifies the view body and picks an incremental plan: append-only for monotonic event streams, key-based upsert for dimensions and aggregates, delete+insert for windowed views with retraction, full recompute on schema change.

CDC-fed off the existing pipeline

The same CREATE STREAM we already create per synced base table feeds the operator graph.

Built for

Where this fits.

Teams who want streaming freshness without a streaming database

You already run Snowflake.

dbt users with incremental models that need to be more frequent than 15 min

The model is already correct.

Embedded analytics product teams

You ship live tiles to customers.

Want incremental views in your stack early?

We’re shipping this with a small group of design partners. Tell us about your workload and we’ll set you up.