agentsociety.cityagent.blocks.plan_block¶
Module Contents¶
Classes¶
A block for generating and managing execution plans through LLM-guided decision making. |
Data¶
API¶
- agentsociety.cityagent.blocks.plan_block.GUIDANCE_SELECTION_PROMPT = <Multiline-String>¶
- agentsociety.cityagent.blocks.plan_block.DETAILED_PLAN_PROMPT = <Multiline-String>¶
- class agentsociety.cityagent.blocks.plan_block.PlanBlock(agent: agentsociety.agent.Agent, toolbox: agentsociety.agent.AgentToolbox, agent_memory: agentsociety.memory.Memory, agent_context: agentsociety.agent.DotDict, max_plan_steps: int = 6, detailed_plan_prompt: str = DETAILED_PLAN_PROMPT)¶
Bases:
agentsociety.agent.BlockA block for generating and managing execution plans through LLM-guided decision making.
Attributes: configurable_fields: List of configurable parameter names default_values: Default values for configurable parameters fields_description: Human-readable descriptions for configurable parameters guidance_options: Predefined options mapped to specific needs max_plan_steps: Maximum allowed steps in generated plans (configurable)
Initialization
Initialize PlanBlock with required components.
Args: llm: Language Model interface for decision making environment: Environment for contextual data memory: Agent’s memory storage for status tracking
- async select_guidance(current_need: str) Optional[Tuple[dict, str]]¶
Select optimal guidance option using Theory of Planned Behavior evaluation.
Args: current_need: The agent’s current need to fulfill
Returns: Optional[tuple[dict, str]]: Selected option with TPB evaluation scores and reasoning. None if no guidance option is selected by bad response from LLM.
- async generate_detailed_plan() Optional[dict]¶
Generate executable steps for selected guidance option.
Args: plan_target: The target of the plan
Returns: dict: Structured plan with target and typed execution steps. None if no plan is generated by bad response from LLM.
- async forward()¶
Main workflow: Guidance selection -> Plan generation -> Memory update