Rebuilds the paused agent from the checkpoint, applies the recorded decision (approve / reject / edit), and continues the tool loop to completion. The resumed work keeps the original accounting, so the run reads as one continuous exchange.
Usage
resume_run(checkpoint, ...)
# S3 method for class 'agent_resume_result'
as.character(x, ...)
# S3 method for class 'agent_resume_result'
print(x, ...)Arguments
- checkpoint
A
llmragent_checkpointwith a decision recorded byapprove_tool_call().- ...
Reserved.
- x
An
agent_resume_resultobject.
Value
An object of class agent_resume_result with text (the final
reply), agent (the rebuilt agent, or NULL), and checkpoint (the
checkpoint that was resumed).
