API Reference

This section provides complete API documentation for all OmniEmbodied components.

• OmniEmbodied Framework API
  • Evaluation Framework
  • Agent Framework
  • LLM Integration
  • Configuration System
  • Utilities
  • Data Generation Framework
  • Usage Examples
• OmniSimulator API Reference
  • Core Engine
  • Environment Management
  • Action System
  • Agent System
  • Utilities
  • Core Types and Enums
  • Usage Examples
• Data Structures Reference
  • Evaluation Results
  • Agent Configuration Structures
  • Error Handling
  • Trajectory Data Structures
  • Usage Examples
  • See Also

Core Modules

OmniSimulator	Embodied Simulator - 文本具身任务模拟器
evaluation
core
llm
modes
config
data_generation

Quick API Overview

Core Simulation:

OmniSimulator.core.engine.SimulationEngine([...])	模拟引擎类 - 整个模拟器的核心控制器
OmniSimulator.agent.agent.Agent(agent_id, ...)	智能体类 - 表示模拟环境中的智能体

Environment and Actions:

OmniSimulator.environment.environment_manager.EnvironmentManager(...)	环境管理器 - 负责管理模拟环境中的所有实体（房间、物体、家具）
OmniSimulator.action.action_manager.ActionManager(...)	动作管理器 - 负责动作的创建、验证和执行
OmniSimulator.environment.room.Room(room_id, ...)	房间类 - 表示模拟环境中的房间

Evaluation Framework:

evaluation.evaluation_manager.EvaluationManager(...)	评测管理器 - 统一评测管理和场景级并行执行
evaluation.task_executor.TaskExecutor(...)	任务执行器 - 执行单个任务的详细步骤

Agent Modes:

modes.single_agent.llm_agent.LLMAgent
modes.centralized.centralized_agent.CentralizedAgent

Configuration Management:

config.config_manager.ConfigManager([...])

统一配置管理器 支持配置继承、环境变量替换、命令行参数覆盖等功能

LLM Integration:

Usage Examples

Evaluation

from evaluation.evaluation_interface import EvaluationInterface

# Run evaluation
result = EvaluationInterface.run_evaluation(
    config_file="single_agent_config",
    agent_type="single",
    task_type="independent",
    scenario_selection={
        "mode": "range",
        "range": {"start": "00001", "end": "00010"}
    }
)

print(f"Success rate: {result['overall_summary']['success_rate']:.2%}")

Basic Simulation

from OmniSimulator.core.engine import SimulationEngine

# Initialize simulation
engine = SimulationEngine(config_file="simulator_config")
engine.initialize_with_task(task_file="example_task.json")

# Process commands
result = engine.process_command("go to living room")
print(result)

Custom Agent

from modes.single_agent.llm_agent import LLMAgent
from OmniSimulator.core.engine import SimulationEngine

# Create simulation
simulator = SimulationEngine()

# Create agent
agent = LLMAgent(
    simulator=simulator,
    agent_id="agent_1",
    config={"model": "gpt-4"}
)

# Agent reasoning step
step_result = agent.decide_action()
print(step_result)