Tutorials

Walkthroughs: instrument a RAG chatbot, trace a multi-step agent with parallel tool calls, and run nightly LLM-as-judge evals on production traffic.

Add observability to a RAG chatbot

End-to-end tutorial. Take a basic RAG chatbot, add Spanlens, and see retrieval + generation as one trace with token cost and per-step latency.

Multi-step agent tracing

Tutorial: trace an agent that classifies intent, fans out to parallel tools, and composes an answer. Per-step cost and critical path in one waterfall.

Nightly evals on production traffic

Tutorial: set up an LLM-as-judge evaluator on a nightly sample of production traffic and catch prompt quality regressions before users complain.

Back to the docs overview.