Concept
Execution traces
An execution trace is the recorded sequence of what an agent actually did: the calls it made, the arguments it passed, the outputs or errors that came back, and the intermediate state at each step. Think of it like a program's stack trace or a structured application log,…
The rest of “Execution traces” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→