agents.planning.llm_compiler_v3.agent

LLM Compiler V3 Agent using Enhanced MultiAgent V3 Architecture.

This implementation modernizes the LLM Compiler pattern by using Enhanced MultiAgent V3 for simplified architecture, better maintainability, and consistent patterns.

Classes

LLMCompilerV3Agent

LLM Compiler V3 Agent using Enhanced MultiAgent V3.

Module Contents

class agents.planning.llm_compiler_v3.agent.LLMCompilerV3Agent(name='llm_compiler_v3', config=None, tools=None, **kwargs)

LLM Compiler V3 Agent using Enhanced MultiAgent V3.

This agent implements the LLM Compiler pattern with three specialized sub-agents: 1. Planner - Decomposes tasks into parallelizable DAG 2. Task Fetcher - Manages task coordination and dependency resolution 3. Parallel Executor - Executes individual tasks with tools 4. Joiner - Synthesizes results into final answer

Initialize LLM Compiler V3 Agent.

Parameters:

  • name (str) – Agent name

  • config (haive.agents.planning.llm_compiler_v3.config.LLMCompilerV3Config | None) – Configuration for the compiler

  • tools (list | None) – List of tools available for task execution

  • **kwargs – Additional arguments for Enhanced MultiAgent V3

async arun(query, context=None, **kwargs)

Execute LLM Compiler pattern asynchronously.

Parameters:

  • query (str) – User query to process

  • context (dict[str, Any] | None) – Additional context

  • **kwargs – Additional execution parameters

Returns:

CompilerOutput with final results and execution details

Return type:

haive.agents.planning.llm_compiler_v3.models.CompilerOutput

run(query, context=None, **kwargs)

Execute LLM Compiler pattern synchronously.

Parameters:
Return type:

haive.agents.planning.llm_compiler_v3.models.CompilerOutput