Building a near real-time application with Zerobus Ingest and Lakebase

This post explains how to build a near real-time application using Databricks' Zerobus Ingest and Lakebase, eliminating traditional multi-hop ingestion architectures.

• •Zerobus Ingest (part of Lakeflow Connect) lets applications push event data directly into Delta tables, bypassing Kafka or staging layers, with ≤5 second ingestion latency and up to 100 MB/sec throughput per client.
• •Lakebase is a fully managed, serverless Postgres database built into Databricks, designed for low-latency operational workloads with automatic data sync from the Lakehouse.
• •A continuous sync pipeline replicates Delta table data to Lakebase with 10–15 second lag; a ForeachWriter mode is available for sub-second latency.
• •A practical example builds a food delivery driver tracking app: Zerobus SDK ingests GPS telemetry → continuous sync to Lakebase → FastAPI backend via WebSockets → frontend on Databricks Apps.