Module output 

Available on crate features agents and codeact only.

The typed value a script returns to the host.

A CodeAct script communicates its result by returning a tagged value (the runtime’s completion value, decoded to JSON). The host classifies it into one of three outcomes:

• ScriptOutput::Observation — fed back to the model for the next turn,
• ScriptOutput::Error — fed back to the model (an opaque message string),
• ScriptOutput::FinalResult — returned to the caller; ends the loop,
• ScriptOutput::TransferToAgent — hands control to another agent; ends the loop. Only offered to the model when transfer targets exist.

Errors are just strings. The framework is language-agnostic: it does not model exception types, tracebacks, or “kinds”. Whatever the runtime engine produces — a Python traceback, a JS stack, a shell error — is passed through verbatim, because that is what the LLM expects for that engine.

Enums

ScriptOutput

The typed value a CodeAct script must return.