06.1-observability

Chapter 6.1 -- Observability with OpenTelemetry

Companion code for Generative AI in .NET, Chapter 6 sections 6.1.2 -- 6.1.3 ("Distributed Tracing with OpenTelemetry" and "Token Usage Tracking").

Wraps an IChatClient with UseOpenTelemetry(...) and configures an OTel pipeline that prints spans to the console and (optionally) exports OTLP to a collector. Each chat call shows up as a span with gen_ai.* attributes -- model, system, input/output token counts.

Run it

ollama pull phi4-mini
dotnet run --project samples/ch06-production/06.1-observability

To send spans to a real collector (Jaeger / Application Insights / Grafana Tempo):

export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
dotnet run --project samples/ch06-production/06.1-observability

Manuscript reference

• Manuscript/Chapter-06.md, sections 6.1.2 -- 6.1.4.
• Figure 6.1 ("AI Pipeline Observability Architecture").

Prerequisites

• .NET 9 SDK, Ollama with phi4-mini.