Skip to main content

A lightweight and composable graph-structured workflow engine for LLM automation

Project description

AgentOS2

An LLM-driven intelligent workflow system featuring graph-based orchestration, nested flow execution, and dual-layer concurrency architecture. LLM驱动的智能工作流系统,支持图结构编排、流程嵌套、双层并发架构

📌 项目结构

agent_os2/
├── agent_os/              # 核心框架
│   ├── base_agent.py      # Agent抽象
│   ├── flow.py            # DAG调度器  
│   ├── base_model/        # 模型接口
│   ├── utility.py         # 工具函数
│   └── visualize.py       # 可视化
├── agents/examples/       # 示例
├── agent_settings.json    # Agent配置
├── model_settings.json    # 模型配置
└── DEVELOPING_GUIDE.md    # 开发指南

🚀 核心特性

一、BaseAgent

1.1 生命周期

已完成: 标准流程(激活→批处理→执行→下发) | 三阶段(setup/pipeline/post_process) | 纯函数特性

1.2 上下文系统

已完成: Settings(构建时继承) | Args/SharedContext/ExtraContexts(运行时) | 模板系统({src.*}/{ctx.*}) | 深度合并
未完成: 上下文压缩 | RAG/KAG集成

1.3 批处理

已完成: batch_field | 批处理索引 | 双层并发架构

1.4 模型交互

已完成: 统一接口 | 智能重试 | 超时控制 | 严格模式 | JSON校验
未完成: 模型Tag精选 | 动态切换

1.5 日志系统

已完成: user_info | debug_info | 批处理标识 | 即时输出

1.6 command_to_flow

已完成: memory(更新context) | actions(修改图结构) | 条件执行 | 可扩展

二、Flow

2.1 图结构调度

已完成: 有向图依赖 | 拓扑并发 | 动态修改 | 循环/条件分支
未完成: 分布式调度 | 远程执行

2.2 Flow功能

已完成: Agent/Flow一体化 | YAML DSL | 入口管理 | 别名机制 | 热重载 | Mock注入

2.3 上下文隔离

已完成: 父子隔离 | 按需继承(expected_shared_context_keys) | 动态更新 | 优化传递

2.4 执行统计

已完成: 资源限制(max_*_limit) | 性能指标 | 执行可视化 | flow_results

三、辅助系统

3.1 模型系统

已完成: Processor注册 | LLM/ImageConfig | 多模型配置 | 流式输出 | 多模态
未完成: 模型路由 | 自适应

3.2 工具函数

已完成: DSL解析 | Agent发现 | JSON解析 | 字典合并 | 模板解析
未完成: Token压缩 | RAG/KAG工具

3.3 执行接口

已完成: execute() | 可视化执行 | 单Agent测试 | 双模式 | 并发限制 | observer
未完成: 分布式执行 | MCP插件

📖 最新更新

  • 上下文传递优化: 只传递存在的键,改进get()行为

🔗 相关文档

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

agent_os2-1.2.0.tar.gz (40.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

agent_os2-1.2.0-py3-none-any.whl (44.5 kB view details)

Uploaded Python 3

File details

Details for the file agent_os2-1.2.0.tar.gz.

File metadata

  • Download URL: agent_os2-1.2.0.tar.gz
  • Upload date:
  • Size: 40.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.10

File hashes

Hashes for agent_os2-1.2.0.tar.gz
Algorithm Hash digest
SHA256 79705de3391ac73587288c413cd6bbe8f83d46972268a6aff8a861189a6f77a2
MD5 de03a44304c8fb49279bd0e19c1abb9c
BLAKE2b-256 ab5aa36f5960db1a546397473bb8a6f97fb4c5edf59a39c88553e8415753db44

See more details on using hashes here.

File details

Details for the file agent_os2-1.2.0-py3-none-any.whl.

File metadata

  • Download URL: agent_os2-1.2.0-py3-none-any.whl
  • Upload date:
  • Size: 44.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.10

File hashes

Hashes for agent_os2-1.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 b391e91e4d3452eba548baf3e84eacab0af68e1ff8b918744a5230bbdec8f198
MD5 17ff82a49584f95a98b52e3d7d36e4c8
BLAKE2b-256 965914dbb24a1de93117a764e83a11bcc6e663c8cbe4a67bc415cf13e7b49e7a

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page