Keep RAG citations tied to retrieved chunks
Citations should point back to the retrieved evidence, not to sources the model guessed after answering.
Dotnet
Observability for Agents with OpenTelemetry, Aspire, and Application Insights
How to make Microsoft Agent Framework behavior visible in production with OpenTelemetry traces, token usage, model calls, tool calls, Aspire local debugging, and Application Insights monitoring.
Use CSnakes when Python is a function call, not a service
Embed a small typed Python capability in .NET when an HTTP service would be more boundary than value.
Use IHttpClientFactory
Create configured HttpClient instances through the factory instead of scattering clients and handlers.
Cache GET responses only when user context is clear
Do not cache a GET response unless the cache key matches the user, tenant, and permission boundary that shaped it.
Assign least-privilege Azure roles, not broad contributor rights
Give workloads and people the narrow Azure role they need instead of defaulting to Contributor.
Avoid hidden allocations in hot paths
Watch for small per-item allocations inside loops, parsers, serializers, and request paths.
Do not hide exceptions inside fire-and-forget tasks
If work can fail, give it ownership, logging, cancellation, and a lifecycle.
Expose a Python FastAPI AI tool as an Aspire service
Keep Python AI code behind an HTTP boundary and let Aspire wire it into the local system.
Keep model name and deployment name in config
Do not hard-code model routing decisions into application logic.