What are Traces?

Noodler’s observability is built on two ideas: traces and spans.

A trace represents a single interaction with your AI system, from start to finish.

A span represents one step that happens during that interaction.

A Simple Example

Imagine a user asks an AI a question:

User: What is the capital of Guyana?

AI: The capital is Georgetown.

Even though this looks like a single exchange, multiple things happen behind the scenes.

In Noodler, we break this interaction into spans, such as:

1. Receiving the user’s input
2. Generating a response with the model

Each span captures useful details about that step — like how long it took, how many tokens were used, or how much it cost.

All of these spans are grouped together into a single trace, which represents the full interaction.

Why Traces Matter

Traces let you see how an AI interaction actually unfolds.

Instead of guessing why a response was slow, expensive, or incorrect, you can see each step that led to it — in order, with context.

If you want to improve an AI system, traces show you where to look.

OpenTelemetry

Noodler uses OpenTelemetry and adheres to its tracing model.

This means traces and spans in Noodler follow an industry standard, making your observability data portable and interoperable with the broader ecosystem.