nat.experimental.test_time_compute.search.multi_llm_planner#

Attributes#

Classes#

MultiLLMPlanner

A planner that uses multiple LLMs to generate plans. Each LLM can generate

Functions#

register_multi_llm_planner(config, builder)

Register the MultiLLMPlanner strategy with the provided configuration.

Module Contents#

logger#

class MultiLLMPlanner( config: nat.experimental.test_time_compute.models.search_config.MultiLLMPlanConfig, )#

Bases: nat.experimental.test_time_compute.models.strategy_base.StrategyBase

A planner that uses multiple LLMs to generate plans. Each LLM can generate a specified number of plans, and all plans are combined.

config#

llms_bound = []#

async build_components(builder: nat.builder.builder.Builder) → None#: Build the components required for this multi-LLM planner. Binds each LLMRef from the config with the selected framework wrapper (LANGCHAIN).

supported_pipeline_types() → [nat.experimental.test_time_compute.models.stage_enums.PipelineTypeEnum]#: Return the stage types supported by this selector.

stage_type() → nat.experimental.test_time_compute.models.stage_enums.StageTypeEnum#: Return the stage type of this strategy.

async _generate_plan_for_temperature( llm, base_prompt: str, temperature: float, ) → nat.experimental.test_time_compute.models.ttc_item.TTCItem#

async _generate_plans_for_llm( llm, base_prompt: str, ) → list[nat.experimental.test_time_compute.models.ttc_item.TTCItem]#

async ainvoke(

items: list[nat.experimental.test_time_compute.models.ttc_item.TTCItem],

original_prompt: str | None = None,

agent_context: str | None = None,

**kwargs,

) → list[nat.experimental.test_time_compute.models.ttc_item.TTCItem]#: Generate a list of PlanningItems by querying each LLM in self.llms_bound. Each LLM produces ‘plans_per_llm’ plans.

async register_multi_llm_planner( config: nat.experimental.test_time_compute.models.search_config.MultiLLMPlanConfig, builder: nat.builder.builder.Builder, )#: Register the MultiLLMPlanner strategy with the provided configuration.