From Zero Docker Experience to Tracing Every LLM Call: Building an Observable AI Assistant with SigNoz

Short summary

A second-year student documents building an observable AI assistant using SigNoz, Docker, and OpenTelemetry to trace every LLM API call. They instrument Groq API requests with spans capturing duration, token counts, and per-call cost, then visualize failures and build a dashboard for broader monitoring. The article is a hands-on walkthrough from zero Docker experience to a working observability setup for AI agents.

• •Set up SigNoz via Docker and instrument Groq LLM calls with OpenTelemetry spans
• •Track per-request token counts, latency, and dollar cost as span attributes
• •Debug a silent 401 error in 510ms using the trace explorer, proving the value of AI agent observability

Generated with AI, which can make mistakes.