Reviewer Agent¶
osa_tool.osa_agent.agents.reviewer.agent
¶
ReviewerAgent
¶
Bases: BaseAgent
ReviewerAgent validates the execution results and user feedback, determines whether: - the task is completed, - user needs changes, - user wants a completely new intent or task scope.
run(state)
¶
Ask the user to approve or reject the result, then optionally analyze feedback.
If the user approves, state.approval is set and the graph routes to finalizer. If not, collects feedback, runs LLM to extract intent/scope changes, and routes back to planner.