Three New Data Sources Supported Without Conversion

Databricks AI Search expands its ecosystem by now natively supporting three types of tables: Streaming Tables, Materialized Views, and Managed tables using Iceberg v3+, in addition to the existing Delta tables. According to the official Databricks blog, this evolution avoids the creation of intermediate Delta copies and the management of separate synchronization pipelines.

Databricks AI Search now eliminates intermediate synchronization pipelines to directly connect real-time data, custom views, and Iceberg tables to its applications.

Streaming Tables: Real-Time Artificial Intelligence Without Latency

Streaming Tables enable AI Search to be fed with real-time events, such as financial transactions, security alerts, or IoT data. Before this update, organizations had to build custom pipelines to synchronize these streams with a dedicated Delta table. Now, AI Search connects directly to these tables, reducing operational complexity and accelerating critical applications like fraud detection or continuous monitoring, as explained by Databricks.

Materialized Views: Customization and Aggregation Made Effortless

Materialized Views provide increased flexibility by allowing aggregation, filtering, or joining data from multiple sources to create a customized state. For example, a Materialized View can combine recent customer data with historical preferences to power real-time recommendations. AI Search can index these views directly, without requiring prior transformation, as highlighted by ReleaseBytes. This simplifies the creation of dynamic user experiences, such as searches tailored to a customer’s recent activity.

Iceberg v3+: Opening data without forced migration

The Managed tables using Iceberg v3+ can now serve as a direct source for AI Search, without conversion to another format like Delta. This native compatibility with an open format reduces migration costs and technological lock-ins. According to Databricks, organizations can thus leverage multi-format data lakes while benefiting from AI Search’s advanced search capabilities, without technical overhead.

What this changes here: fewer pipelines, more agility for African businesses

For businesses and government agencies in Benin and West Africa, this update could significantly reduce costs and complexity related to integrating real-time or multi-source data into AI applications. For example, a Cotonou-based bank using real-time transaction streams to detect fraud could now eliminate manual synchronization steps between its systems and AI Search, speeding up its analyses. Similarly, public administrations managing heterogeneous data (such as voter registries or health data) could simplify indexing of this information without resorting to costly migrations to a single format.

The Materialized Views would also benefit e-commerce platforms or financial services, where real-time personalization (recommendations, alerts) is critical. Finally, support for Iceberg v3+ would allow organizations already using this open-source format to maintain their existing infrastructure while adopting AI Search, an advantage for limited budgets or varied technical environments.