A developer studies a large monitor displaying code, data pipelines, and analytics dashboards in a blue-lit office.
Microsoft's Azure Cosmos DB extension for Visual Studio Code now has an AI-assisted Data Modeler in preview. It lets you work through partition-key and container design inside the editor, before any code ships. The announcement came at Microsoft Ignite 2026 on November 17. In Azure Updates it carries the "In preview" label, which Microsoft defines as available to all Azure customers for non-production use and testing. Treat it as a design aid, not a production dependency.

A developer studies a large monitor displaying code, data pipelines, and analytics dashboards in a blue-lit office. What Microsoft announced​

Microsoft's Ignite 2026 Cosmos DB post says developers can design data models using AI-assisted recommendations. Those recommendations are based on document structures, query patterns, write distribution and expected growth. The tool is meant to help with the following:

  • Spotting potential hot partitions and cross-partition queries.
  • Comparing partition-key strategies.
  • Generating Bicep, Terraform or SDK configurations.

The same release adds an updated account dashboard meant to simplify navigation between resources in VS Code. Microsoft's Ignite Book of News describes the package as account insights, AI-powered recommendations and guided data modeling for partition keys and container configurations directly in VS Code.

Why partition keys are the right target​

In Cosmos DB for NoSQL, a poor partition key is expensive to fix later. A skewed key concentrates traffic on one partition. Queries that fan out across partitions cost more request units and add latency. This is general industry knowledge rather than something stated in the announcement. Moving that conversation into the editor, next to the code, is the practical pitch.

Microsoft offers no benchmarks or measured improvements for the modeler. Treat the recommendations as a starting point. Validate them against your real access patterns, data distribution and growth.

What the extension docs and release notes add​

Microsoft Learn's extension overview, updated November 17, now lists both new items as previews:

  • Account dashboard (preview) covers account health, monitoring metrics, container inventory, advisories and alerts in Account Overview.
  • Data Modeler (preview) lets you describe workloads, compare Copilot partition-key recommendations, and optionally deploy containers or export deployment code.

Other sources add detail on how the modeler works:

  • Copilot dependency. The Visual Studio Marketplace listing says Data Modeler (Preview) helps you choose partition keys for Azure Cosmos DB for NoSQL using GitHub Copilot. It also says Try Data Modeler opens the modeler for this account; it requires AI features enabled and available GitHub Copilot models. If your organization restricts Copilot, expect the modeler to be unavailable or limited.
  • How to open it. The Marketplace page says you can open it with Try Data Modeler in Account Overview, from an account in Explorer, or run Data Modeling Wizard from the Command Palette.
  • Local data. The Marketplace page describes a command, Cosmos DB: Delete All Data Modeler Data, that permanently clears local Data Modeler data for every account, closing all modeler tabs and removing saved sessions.
  • Inputs. An earlier extension release (v0.38.0) added an experimental modeler with schema imports, input validation, workload throughput warnings, and query inputs that accept a description, a query, or both. That release notes the experience is enabled through configuration.

Prerequisites and version caveats​

Microsoft Learn lists the baseline as VS Code on Windows, macOS or Linux and a Cosmos DB account with a database and container. It says to install the Azure Cosmos DB extension from the Extensions view and sign in through Microsoft Entra ID. It also advises using Entra ID role-based access control for Cosmos DB resources.

The minimum VS Code version is unclear. Learn's overview says 1.85 or later. However, the extension's v0.38.0 release notes say it raised the minimum supported VS Code version to 1.125.0. If the modeler doesn't appear, update VS Code first. This is my inference from the conflicting version information, not a documented fix.

The modeler applies to Cosmos DB for NoSQL. That scope comes from the extension documentation.

Open questions​

The sources I reviewed don't answer several things admins will want to know:

  • Whether prompts or schema details leave your environment beyond what GitHub Copilot already does.
  • What permissions the modeling workflow needs.
  • Whether the feature has regional or account-type limits.
  • Which SDK languages the generated configuration covers.

Check these in a non-production subscription before rolling the modeler out to a team.

Don't confuse it with the other Ignite announcements​

Microsoft's same post announces several separate Cosmos DB items. These include Cosmos DB Shell (generally available), the Rust SDK (generally available) and a managed MCP server (public preview). They are not part of this preview.

Bottom line​

For teams starting a new Cosmos DB for NoSQL project, the preview is worth trying on a throwaway account. It can surface hot-partition and cross-partition risks early and produce deployable Bicep or Terraform. Verify its suggestions against real workload data, and confirm your Copilot and AI-feature settings allow it to run.

 

References

  1. (In preview) Public Preview: AI-assisted data modeler and other updates in Azure Cosmos DB extension for VS Code Azure Updates 2026-11-17T20:00:32Z
  2. Azure Cosmos DB - Visual Studio Marketplace marketplace.visualstudio.com
  3. Visual Studio Code extension for Azure Cosmos DB - Azure Cosmos DB | Microsoft Learn learn.microsoft.com