Attach build identity to operational telemetry
Stamp telemetry with the immutable build that produced it so incidents can be compared across releases and rolling deployments.
Observability
Choose histogram boundaries from the service objective
Put the latency objective into the histogram boundaries so compliance can be calculated without interpolating between buckets.
Separate ingestion health from retrieval readiness
A failed ingestion run should not stop retrieval while an accepted index remains queryable and eligible to serve.
The Model Is Only One Dependency. Map the Rest.
How to map the execution, prerequisite, and operational dependencies around an AI feature and give each interaction a usable failure contract.
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.
Make logs structured before adding dashboards
Capture queryable fields first so dashboards are built on data, not text search.
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.
Prefer health checks that test dependencies deliberately
Health checks should say something useful about readiness without turning every probe into production load.
Use OpenTelemetry before you need production debugging
Add traces, metrics, and logs while the system is still easy to reason about.