Stop Letting Provider SDKs Define Your .NET AI Architecture
How Microsoft.Extensions.AI keeps provider SDKs at the edge of .NET applications through shared abstractions, dependency injection, and middleware.
Dotnet
Treat uploaded files as untrusted
Keep uploads quarantined until required metadata, content, and resource checks succeed, while treating client filenames as untrusted.
Use streaming for large or frequent uploads
Use MultipartReader when buffered ASP.NET Core uploads put too much pressure on memory, temporary-disk capacity, or disk I/O.
Use async for I/O, not as a substitute for CPU parallelism
Use async to avoid blocking while I/O is in flight, and treat CPU parallelism as a separate measured design decision.
Calling a Model Is Easy. Running an AI System Is Not
Why a successful model call says little about whether an AI feature is reliable, secure, observable, affordable, and ready to operate in production.
Measure retrieval misses separately from model failures
Give retrieval its own outcome and telemetry so missing or irrelevant context is not diagnosed as a model problem.
Project EF Core queries before materializing
Use Select to project only the required values before ToListAsync, reducing transferred columns and unnecessary entity materialization on EF Core read paths.
Describe tool parameters like API contracts
Give AI tools precise parameter names, units, formats, and side-effect expectations so the model receives an unambiguous contract.
Keep public API DTOs separate from domain entities
Use explicit request and response contracts so persistence fields and domain state do not become part of the public API accidentally.
Microsoft Foundry: When Agent Framework Agents Become Managed Resources
How to host a Microsoft Agent Framework agent in Foundry while keeping authorization, conversation ownership, identity, and governance boundaries explicit.