Skip to main content
These recipes adapt established streaming and lakehouse patterns to the Verglas component model. Instead of provisioning a cluster, assign each responsibility to a Worker, Durable Object, Stream, Pipeline, or built-in state binding.

Webhook to lakehouse

Acknowledge quickly, ingest durably, normalize, and publish to Iceberg.

Deduplicate events

Use a keyed Durable Object before an event enters the Stream.

Bronze, silver, gold

Turn the medallion pattern into serverless resource boundaries.

Semantic search

Store embeddings in Vectorize and serve ranked results from a Worker.

​
Pattern translation

​
Sources behind the patterns

The recipes are original Verglas implementations informed by the official documentation for Databricks medallion architecture, Apache Flink deduplication, Apache Iceberg table maintenance, and Cloudflare Vectorize. They copy the useful architectural ideas, not vendor-specific cluster or job machinery.