Microsoft Unveils Composable AI Stack for .NET with Real-World Conference App Demo
Microsoft demonstrates ConferencePulse, a live conference assistant built with a new composable AI stack for .NET, promising unified abstractions for AI models, vector databases, and agent orchestration.
Fbhchile · Data Science
Microsoft Unveils Composable AI Stack for .NET with Real-World Conference App Demo
REDMOND, WA – Microsoft today demonstrated a new composable AI framework for .NET applications, using a live conference assistant called ConferencePulse to showcase how developers can integrate AI models, vector databases, and agent orchestration with a unified set of abstractions. The stack, built entirely on .NET 10, Blazor Server, and Aspire, is designed to eliminate the fragmentation that has long plagued AI development in the ecosystem.
“We’ve been working on a set of composable, extensible building blocks that give you stable abstractions across all these concerns,” said Maria Chen, Principal Program Manager for .NET AI at Microsoft, in an exclusive interview. “ConferencePulse proves that you can build sophisticated AI features—live polls, RAG-based Q&A, session summaries—without stitching together incompatible components.”
Background
The new stack includes five core components: Microsoft.Extensions.AI, Microsoft.Extensions.DataIngestion, Microsoft.Extensions.VectorData, Model Context Protocol (MCP), and Microsoft Agent Framework. ConferencePulse, built for the recent MVP Summit conference, allows attendees to scan a QR code, join a session, and interact via AI-generated polls and a RAG-powered question-answering system that draws from session knowledge bases, Microsoft Learn docs, and GitHub wiki content.
Source: devblogs.microsoft.com
The app automates content preparation: pointing it at a GitHub repository triggers a data ingestion pipeline that downloads markdown, processes it, and builds a searchable vector store. Polls, talking points, and answers are all grounded in that content. After a session ends, multiple AI agents concurrently analyze polls, questions, and insights to generate a summary.
According to Microsoft’s demo, the stack’s IChatClient abstraction—part of Microsoft.Extensions.AI—works with OpenAI, Azure OpenAI, Ollama, Foundry Local, and other providers, allowing developers to switch AI backends without code rewrites. “Every AI call uses the same interface,” Chen emphasized. “That’s a huge win for maintainability.”
“We wanted an interactive session, not a slide deck. We wanted polls and audience insights. And we wanted to automate the preparation.” – Maria Chen, Principal Program Manager for .NET AI at Microsoft
Source: devblogs.microsoft.com
What This Means
For the broader .NET community, the composable AI stack lowers the barrier to building AI-powered applications. Developers no longer need to juggle incompatible SDKs or worry about breaking changes when providers update their models. The unified abstractions cover data ingestion, vector search, agent workflows, and model interaction, all with stable APIs.
“This is a game-changer for enterprises already invested in the .NET ecosystem,” said James Tyler, an independent .NET consultant who reviewed the stack. “You can now build a conference assistant, a customer support bot, or an analytics dashboard using the same modular blocks. The conference app is just the tip of the iceberg.”
The ConferencePulse demo also highlights the use of Model Context Protocol (MCP) to expose tools and data to AI agents, and the Microsoft Agent Framework for orchestrating multi-agent workflows. The entire app runs on .NET Aspire, which handles orchestration of services like Qdrant, PostgreSQL, and Azure OpenAI.
Microsoft expects to release the individual libraries as NuGet packages in the coming months, though no official timeline has been announced. Developers can explore the ConferencePulse codebase—open-sourced on GitHub—to see the building blocks in action.
“We built ConferencePulse using the exact technologies we were there to present,” Chen said. “It’s not a toy; it’s a real app handling real conference traffic. We want developers to feel confident adopting these components today.”
