Aug 17 — Sep 16, 2026
The past month saw major Databricks ecosystem momentum around production agent infrastructure, with the emergence of Lakebase for transactional workloads alongside rapid maturation of AgentOps tooling and IDE integration. Unity AI Gateway and Genie also saw significant feature expansions focused on governing and operationalizing multi-step agentic workflows.
1.Databricks Lakebase introduces serverless Postgres and LTAP architectures to the lakehouse
Databricks expanded its transactional footprint with Lakebase, a managed serverless PostgreSQL offering built to eliminate the gap between OLTP state and OLAP lakehouses. Features include compute-level autoscaling caching, branching workflows for test-driven development, and state coordination for durable AI agent systems running frameworks like Temporal.
Sources
- Managed Postgres: What Lakebase Actually Takes Off Your PlateNews · databricks-blog · Sep 14
- How to Build Production-Ready Apps in Replit with Lakebase Postgres | Databricks + Replit DemoVideo · Databricks · Sep 10
- Improving Lakebase Postgres compute cacheNews · databricks-blog · Sep 10
- Build durable agents with Temporal and LakebaseNews · databricks-blog · Sep 8
- 140. Databricks Lakebase Explained | Episode 1 | Fully Managed PostgreSQL for Lakehouse (2026)Video · Raja's Data Engineering · Sep 7
2.AgentOps tooling and Unity AI Gateway formalize LLM evaluation and cost control
Observability and governance for production LLM workloads converged across MLflow, Unity AI Gateway, and new AgentOps frameworks. MLflow 3.16.0 established trace exploration and conversation grouping as default workflows, while AI Gateway added policy guards and token-tracking to curb runaway costs and failed tool retries.
Sources
- Evaluation-First AI Agents: How Zepto Scales Customer Support on Databricks and MLflowNews · databricks-blog · Sep 9
- 3. What is Unity AI Gateway | How to Control AI SpendingVideo · Databricks · Sep 7
- MLflow 3.16.0Release · mlflow/mlflow · Sep 4
- Announcing the Databricks Big Book of AgentOpsNews · databricks-blog · Sep 2
- How we eliminated $1 million a year of wasted AI agent spend in one hourNews · databricks-blog · Sep 1
3.Genie Agents advance into multi-step reasoning across structured and unstructured data
Databricks expanded Genie from basic natural-language-to-SQL dashboards into autonomous Genie Agents capable of file-based reasoning over Unity Catalog volumes and multi-step investigation. Teams are operationalizing semantic ontologies and embedding Genie into operational applications to automate domain-specific analysis.
Sources
- How energy teams turn theft detection into governed action with Genie and AI business processesNews · databricks-blog · Sep 15
- How Databricks’ marketers use data 3x more with Genie, an AI analytics assistantNews · databricks-blog · Sep 15
- Diagnose Manufacturing OEE Issues with Genie AgentsVideo · Databricks · Sep 10
- Expanding Genie Agents: Deep analysis, file reasoning, and moreNews · databricks-blog · Sep 2
- Operationalizing Genie Ontology in Your Data StackNews · databricks-blog · Sep 1
4.IDE extensions and SSH tunneling solidify remote development for Databricks Apps and compute
Developer tooling received major stability upgrades centered on bidirectional IDE connectivity and Databricks Asset Bundles (DABs). Improvements across the Databricks CLI and VS Code extension delivered automatic SSH tunnel reattachment and session sync, enabling developers and local coding agents to execute remotely against Databricks compute with minimal friction.
Sources
- Release: v2.17.0 (#2187)Release · databricks/databricks-vscode · Sep 10
- Connect VS Code to Databricks: The SSH Tunnel GuideVideo · Databricks Skill Builder · Sep 9
- v1.16.0Release · databricks/cli · Sep 9
- How to Connect Your Coding Agent and IDE to Databricks Compute | Databricks SSH Tunnel DemoVideo · Databricks · Sep 3
- 137: Databricks Apps Explained | Part 1: The Problem Every Data Team FacesVideo · Raja's Data Engineering · Sep 2