About explainability

A trustworthy agent is one whose reasoning you can inspect. ConfPlanner is built so that every run leaves two readable artefacts behind: the planning log and the trace. This page is about why those artefacts matter and how to read them as a model of the agent's mind.

The planning log is the plan made visible

Because the agent uses goal-oriented planning, the order of actions is derived, not written. The planning log is where that derivation becomes legible:

UserInput → extractAttendeeProfile → loadCatalog → shortlistSessions → researchSessions
          → assembleSchedule → confirmSchedule

Reading it top to bottom tells you exactly which capabilities the planner chose and in what order — and, when a run stalls, which required type was never produced. The log is the first place to look when behaviour surprises you.

The trace is the plan executed

Where the planning log shows intent, the trace shows what actually happened: each action's inputs and outputs, the LLM calls made, the time and tokens spent. With Zipkin tracing enabled, the same information is rendered as a span tree you can click through.

Why two artefacts, not one

The gap between what was planned and what executed is where most agent bugs live. Keeping the two views separate — the derived plan and the recorded execution — lets you ask precise questions: did the planner choose the wrong path, or did a chosen action behave wrongly? That distinction is the heart of debugging an agent you can trust.

Where it connects

Explainability is the observable face of the dual harness: the trace is to run-time what the planning log is to design-time. To turn on the optional distributed trace, see how to enable Zipkin tracing. For the configuration keys that control trace output, see the configuration reference.