goal
- The user’s goal or command to executellm
- The language model to use for both agentsmax_steps
- Maximum number of steps for both agentstimeout
- Timeout for agent execution in secondsreasoning
- Whether to use the PlannerAgent for complex reasoning (True)
or send tasks directly to CodeActAgent (False)reflection
- Whether to reflect on steps the CodeActAgent did to give the PlannerAgent adviceenable_tracing
- Whether to enable Arize Phoenix tracingdebug
- Whether to enable verbose debug loggingsave_trajectories
- Trajectory saving level. Can be:
**kwargs
- Additional keyword arguments to pass to the agents