First up: Next-Level RAG on Azure: Building Knowledge Bases with Azure AI Search and Foundry – A guide to designing production-ready RAG using Azure AI Search with hybrid retrieval combining keyword search, vector search, and Reciprocal Rank Fusion. Microsoft Agent Framework: Implementing Human-in-the-Loop AI Agents – How to implement human approval checkpoints in AI agents using the Microsoft Agent Framework for scenarios requiring oversight like payment processing.
Microsoft Fabric + Snowflake: Zero-copy interoperability and practical details – Microsoft and Snowflake now offer bi-directional data access using Apache Iceberg and Delta Lake formats, enabling a single copy of data across both platforms. Lakehouse Schemas (Generally Available) – Schema-enabled lakehouses are now GA and the default choice, allowing users to organize tables in folder-like structures with support for cross-workspace queries. End-to-End Observability for Azure Databricks: From Infrastructure to Internal Application Logging – A comprehensive guide covering ten approaches to monitoring Azure Databricks, from diagnostic settings and VM insights to system tables and data quality monitoring.
Finally, we've published our yearly performance benchmarks of our OSS library against the latest version of .NET (10). This year we see a 7% performance boost, which takes the cumulative improvements to 134% since .NET Core 2.0. On Surface Laptop Studio 2, we can now process 10 million messages per second.