Skip to content

Latest commit

 

History

History
167 lines (114 loc) · 3.12 KB

api-reference.md

File metadata and controls

167 lines (114 loc) · 3.12 KB

API Reference

Core Classes

AgentOSKernel

主内核类，管理所有 Agent 生命周期。

from agent_os_kernel import AgentOSKernel

kernel = AgentOSKernel(config=None)

Methods

Method Description
spawn_agent(name, task, priority) 创建 Agent
run(max_iterations) 运行内核
stop() 停止内核
get_agent(pid) 获取 Agent
list_agents() 列出所有 Agent
print_status() 打印状态

ContextManager

虚拟内存式上下文管理。

from agent_os_kernel import ContextManager

cm = ContextManager(max_context_tokens=128000)

Methods

Method Description
allocate_page(agent_pid, content, importance) 分配页面
access_page(page_id) 访问页面
get_agent_context(agent_pid) 获取上下文
swap_out(agent_pid) 换出
swap_in(agent_pid) 换入

StorageManager

PostgreSQL 五重角色存储。

from agent_os_kernel import StorageManager

storage = StorageManager.from_postgresql(url)

Methods

Method Description
save_conversation(agent_pid, messages) 保存对话
create_checkpoint(agent_pid) 创建检查点
semantic_search(query, limit) 向量搜索
acquire_lock(name) 获取锁
log_action(...) 记录操作

AgentScheduler

进程调度器。

from agent_os_kernel import AgentScheduler

scheduler = AgentScheduler()

Methods

Method Description
add_process(agent) 添加进程
remove_process(pid) 移除进程
schedule() 调度
get_stats() 获取统计

LLM Module

LLMProviderFactory

LLM Provider 工厂。

from agent_os_kernel.llm import LLMProviderFactory, LLMConfig

factory = LLMProviderFactory()
provider = factory.create(LLMConfig(
    provider="deepseek",
    model="deepseek-chat"
))

Supported Providers

Provider Models
openai gpt-4o, gpt-4-turbo
deepseek deepseek-chat, deepseek-reasoner
kimi moonshot-v1-8k, moonshot-v1-32k
qwen qwen-turbo, qwen-plus, qwen-max
anthropic claude-3-opus, claude-3-sonnet

Communication Module

AgentMessenger

Agent 消息系统。

from agent_os_kernel.agents.communication import (
    AgentMessenger, Message, MessageType
)

messenger = AgentMessenger()
await messenger.register_agent("agent-1", "Alice")

KnowledgeSharing

知识共享系统。

from agent_os_kernel.agents.communication import KnowledgeSharing

knowledge = KnowledgeSharing()

GroupChatManager

群聊管理。

from agent_os_kernel.agents.communication import GroupChatManager

chat = GroupChatManager()
chat_id = chat.create_chat("讨论主题")

Exceptions

Exception Description
AgentError Agent 基础异常
ContextError 上下文异常
StorageError 存储异常
SecurityError 安全异常
SchedulerError 调度异常