In the last seven days: seven shipped, five in preview, four with a date on it and six new concepts on Lakenaut.

16 changes for you: four needs action, eight worth evaluating and four nice to know.

New features

• Document Intelligence REST APIs for parse, extract and classify reach GA Synchronous REST APIs for ai_parse_document, ai_extract and ai_classify are generally available, so an app or agent can process a file as it arrives.

• On-demand state repartitioning is generally available for stateful streaming You can now change the number of shuffle partitions of a stateful Structured Streaming query or a Lakeflow pipeline without losing its checkpoint state.

• External engines can query Unity Catalog views via the Unity REST API (GA) Teams that expose governed data to non-Databricks engines can include views in that access path instead of only tables.

• Feature Views reach general availability for batch, streaming and request-time data For workspaces with the compliance security profile, they will be switched on automatically in mid-November 2026.

• Lakeflow pipeline parameters are generally available for SQL sources One pipeline's SQL source can run against different catalogs, schemas or tables: defaults live in the pipeline settings, an update can override them.

• Databricks Apps: workspace admins can set default Python and npm package repositories Admins set default Python and npm repositories that every app inherits, and an app can still override them.

• Gemini 4 Argon is now available on Unity Gateway through Foundation Model APIs The release note says customers remain responsible for complying with the applicable model terms.

Previews and beta

• Predictive AI Functions enter Beta: forecasting, anomalies and predictions in SQL Predictive AI Functions are in Beta: an upgraded ai_forecast plus new ai_top_drivers, ai_detect_anomalies, ai_predict_value and ai_predict_class.

• Governed tags can now propagate through lineage to derived tables (Beta) With tag propagation enabled on a governed tag, Unity Catalog uses lineage to apply that tag to derived tables and views after supported operations such as CREATE TABLE AS SELECT.

• Agent memory pipelines distil conversations into agent memory (Beta) A memory pipeline links a session store to a memory store.

• Private Network Gateway for serverless: Private Preview on AWS, Public Preview on Azure Private Network Gateway lets serverless workloads reach resources in your own network through one delegated subnet, configured in a Network Connectivity Configuration.

• Unity Gateway Skills REST API enters Beta with CLI and SDK support A REST API for managing skills on Unity Gateway is now in Beta, covering create, read, update, list, delete and finalize.

Deprecations and end of support

• DeepSeek V4 Flash (0731) will be retired on 5 November 2026 DeepSeek V4.1 Flash is the recommended replacement.

• Databricks Runtime 10.4 LTS ends on 2026-11-01 Anything still relying on it needs a plan before then.

• Genie One MCP Beta endpoint (/api/2.0/mcp/genie) ends on 2026-10-31 Anything still relying on it needs a plan before then.

• DeepSeek V4 Pro (0813) ends on 2026-10-30 Anything still relying on it needs a plan before then.

New and rewritten concepts

• Paying for a foundation model: tokens, units and reservations The four capacity modes for a Databricks-hosted foundation model: pay-per-token, priority, on-demand provisioned throughput and reserved provisioned throughput.

• Structured outputs response_format constrains a chat model's answer to valid JSON or to a JSON schema, with a 64-key ceiling and a deliberately reduced subset of JSON Schema.

• Production monitoring for GenAI apps Registered scorers run continuously against a sampled fraction of live traces and attach their verdicts to each trace as feedback.

• SQL warehouse types and channels The three generally available warehouse types and the Beta fourth one, which acceleration features each has, how fast each starts, and what the Preview channel is for.

• Ingestion patterns: batch, streaming, incremental Batch, streaming and incremental are the three ways into the lakehouse, served by UI uploads, standard connectors and Lakeflow Connect. Choosing between them is an exam question.

• Lakeflow Connect: managed connectors Lakeflow Connect managed connectors ingest SaaS applications and databases into Unity Catalog streaming tables, with authentication.