Back to updates

2026-06 - 22

2026 (12)

June (12)

azure: [In preview] Public Preview: Incremental SharePoint permissions sync for Azure AI Search and Foundry IQ

Azure Updates

2026-06-02

Azure AI Search and Foundry IQ Knowledge Sources now support incremental SharePoint permissions sync in public preview. This keeps document-level, site, and library ACLs up to date as SharePoint permissions change.

azure: [In preview] Public Preview: Image serving in Foundry IQ knowledge bases

Azure Updates

2026-06-02

Foundry IQ knowledge bases in public preview can now return document-extracted images at query time. During ingestion, images such as figures, screenshots, and diagrams are extracted from PDFs, slides, and Office files and stored alongside the source text.

azure: [In preview] Public Preview: Go language support on Azure Functions

Azure Updates

2026-06-02

Azure Functions now supports Go as a first-class language option in public preview. Developers can build event-driven serverless apps in Go using the Azure Functions Go programming model and SDK.

azure: [In preview] Public Preview: Foundry Memory preview refresh

Azure Updates

2026-06-02

Microsoft refreshed the Foundry Memory public preview with a redesigned storage model, new retrieval controls, and tighter integration with the Foundry Agent Service tool surface. Agents can now use Memory to write, search, and update user-scoped data.

azure: [In preview] Public Preview: Deployment Templates support packaging deployment configurations in registries for curated deployment

Azure Updates

2026-06-02

Azure Machine Learning deployment templates now support public preview deployments from registries. This expands the registries that can be used as the source for a templated deployment.

azure: [In preview] Public Preview: Copilot-assisted coding & agent skills to build, deploy, and run Azure Functions faster in VS Code, CLI, and azd

Azure Updates

2026-06-02

Azure Functions public preview adds Copilot-assisted coding and agent skills for building, deploying, and running Functions apps in VS Code, the CLI, and azd. The preview aims to improve AI help quality by providing Azure Functions-specific guidance.

azure: [In preview] Public Preview: Content Understanding NextGen in Microsoft Foundry

Azure Updates

2026-06-02

Microsoft Foundry now includes Content Understanding in the portal as a public preview, so users no longer need to switch to the standalone Content Understanding Studio for common tasks. The preview also adds Read and Layout playgrounds in Foundry for OCR text extraction.

azure: [In preview] Public preview: Connector Namespace

Azure Updates

2026-06-02

Azure has introduced Connector Namespace in public preview. It is a managed service intended to simplify app integration with SaaS and line-of-business systems by providing pre-built connectors instead of custom API client code.

azure: [In preview] Public Preview: CLI and SDK support for Deployment Templates

Azure Updates

2026-06-02

Azure Machine Learning deployment templates now support management through the CLI and SDK in public preview. Platform teams can create, update, delete, and manage templates from infrastructure-as-code pipelines instead of using only the Azure portal.

azure: [In preview] Public Preview: Change streams (multi-shard) in Azure DocumentDB

Azure Updates

2026-06-02

Azure DocumentDB change streams are now in public preview with multi-node support. The feature lets you capture live data changes using controlled polling for real-time, event-driven applications.

azure: [In preview] Public Preview: BM25 full-text search in Azure HorizonDB

Azure Updates

2026-06-02

BM25 full-text search is now in public preview for Azure HorizonDB. The feature provides production-style keyword search with BM25 relevance ranking and runs entirely inside the database.

azure: [In preview] Public Preview: Benchmark evaluations for fine-tuned models in Microsoft Foundry

Azure Updates

2026-06-02

Microsoft Foundry now offers public preview benchmark evaluations for fine-tuned models. The feature is aimed at comparing model performance on tasks such as reasoning, math, and domain knowledge, where leaderboard results may not reflect real-world behavior.