Module execution 

Execution identity: what a run did, told apart from what it was asked to do.

A tool call is an intended operation: a tool name and its arguments. An execution is one attempt to carry that out. The two were the same thing here until now, because the journal used the provider’s own call id as identity, and a provider is under no obligation to make those unique: a retried or reissued call can arrive with the id an earlier one had. Two attempts sharing an id cannot be told apart afterwards, which is exactly the question a run debugger exists to answer.

So an execution gets an id this crate mints, and the provider’s id travels beside it as correlation. Every record about one attempt carries the execution id; anything a provider says about it is matched through the correlation id, which may repeat.

Enums

ToolOutcome

How one attempt to execute a tool call ended.

Functions

mint_attempt_id

Mint an id for one trip to a provider.

mint_execution_id

Mint an id for one execution attempt.

mint_visit_id

Mint an id for one stay in a stage.