agentos-framework 1.0.5

Agent Operating System — Production-ready multi-model agent framework with Tool-Using Agent, LLM Provider abstraction (OpenAI/DeepSeek/Anthropic), Function Calling, streaming, retry, checkpoint/resume, A2A protocol, swarm coordination, and comprehensive observability.

---

AIGC: Label: "1" ContentProducer: 001191440300708461136T1XGW3 ProduceID: 3e9b297b16411e0e0848fc0302358070_9f6cc124744311f1897e5254002afed2 ReservedCode1: IMz08bNnClQ30HefCeW7qyiOyWeXzDuNrXHE5WRO5QkzLJqdjephTgTAmi0TrkvcT0NZ+gKPRuzSPQM4VBEYtDZfVOw9lVTsYJcGnxbUAGFnrM+V13g78mBsOnYa9BL56dEcqwICuI6mRlFWgvODbNtl8KRQ8OsZ13fBwaHtSlr9GXQUZ/plPV+ctJQ= ContentPropagator: 001191440300708461136T1XGW3 PropagateID: 3e9b297b16411e0e0848fc0302358070_9f6cc124744311f1897e5254002afed2 ReservedCode2: IMz08bNnClQ30HefCeW7qyiOyWeXzDuNrXHE5WRO5QkzLJqdjephTgTAmi0TrkvcT0NZ+gKPRuzSPQM4VBEYtDZfVOw9lVTsYJcGnxbUAGFnrM+V13g78mBsOnYa9BL56dEcqwICuI6mRlFWgvODbNtl8KRQ8OsZ13fBwaHtSlr9GXQUZ/plPV+ctJQ=

AgentOS v1.0

Production-ready multi-model agent framework.

Quick Start

pip install agentos

With an API key:

export OPENAI_API_KEY=sk-...
agentos "What's the weather in Beijing?"

Without an API key (mock demo mode):

agentos demo

Features

• LLM Provider: OpenAI / DeepSeek / Anthropic, unified interface
• Tool-Using Agent: Autonomous multi-step reasoning with Function Calling
• Streaming: run_stream() for real-time step yielding
• Retry & Checkpoint: Configurable retry + JSON checkpoint/resume
• Mock Mode: Test without API keys
• A2A Protocol: Agent-to-agent communication
• Swarm Coordination: Multi-agent topologies

CLI

agentos <task>         Run a task
agentos demo           Weather agent demo
agentos serve          Start API server
agentos version        Show version

Provider Auto-Detection

Env Var	Provider	Default Model
OPENAI_API_KEY	OpenAI	gpt-4o-mini
DEEPSEEK_API_KEY	DeepSeek	deepseek-chat
ANTHROPIC_API_KEY	Anthropic	claude-sonnet-4

Python API

from agentos.llm import create_provider
from agentos.llm.base import Tool, ToolParameter
from agentos.agent import ToolAgent, ToolExecutor, AgentConfig

provider = create_provider("openai")
executor = ToolExecutor()
executor.register(
    Tool.from_function("get_weather", "Get weather", {
        "city": ToolParameter(type="string", description="City name"),
    }),
    lambda city: f"{city}: 22°C sunny",
)

agent = ToolAgent(provider, executor, config=AgentConfig())
result = agent.run("What's the weather in Tokyo?")
print(result.final_answer)

（内容由AI生成，仅供参考）