autourgos-starter

The fastest way to get a working Autourgos agent running.

Newcomers to the framework normally have to find and install a core agent package, an LLM backend, and a memory backend separately, each with its own README, before they get anything working. autourgos-starter bundles the recommended default stack as real pip dependencies (nothing vendored or copied) and gives you one function that wires them together.

Install

pip install autourgos-starter

Set your API key:

export OPENAI_API_KEY="sk-..."

Quick Start

This whole block is copy-pasteable — no placeholder variables to fill in.

from autourgos_starter import create_starter_agent, tool

@tool
def add(a: float, b: float) -> float:
    """Add two numbers together."""
    return a + b

agent = create_starter_agent()
agent.add_tools(add)

result = agent.invoke("What is 12 + 30?")
print(result)

That's it — two lines to build the agent (create_starter_agent() and agent.add_tools(add)), one line to run it (agent.invoke(...)).

What's actually happening

autourgos-starter is just a convenience wrapper. create_starter_agent() does this:

from autourgos_react_agent import ReactAgent
from autourgos_openaichat import OpenAIChatModel
from autourgos_buffer_memory import ConversationBufferMemory

llm = OpenAIChatModel(model="gpt-4o-mini", api_key=None, system_instruction=None)
memory = ConversationBufferMemory()
agent = ReactAgent(llm=llm, memory=memory)

Three real, independently-installable packages, each maintained on its own:

• autourgos-react-agent — the ReAct agent loop itself (ReactAgent, tool).
• autourgos-openaichat — the LLM backend (OpenAIChatModel), talks to the OpenAI Chat Completions API or any OpenAI-compatible endpoint (set base_url to point at a local server such as Ollama, LM Studio, or vLLM).
• autourgos-buffer-memory — the memory backend (ConversationBufferMemory), an unbounded in-memory conversation buffer.

This package is optional scaffolding, not a requirement to use the Autourgos framework. If you want a different memory backend (e.g. autourgos-summary-memory, autourgos-token-memory) or a different LLM backend (e.g. autourgos-responses), skip autourgos-starter and wire ReactAgent up to those packages directly — that's exactly what this package does under the hood, just with sensible defaults picked for you.

create_starter_agent() reference

def create_starter_agent(
    api_key: str | None = None,
    model: str = "gpt-4o-mini",
    system_prompt: str | None = None,
    **kwargs,
) -> ReactAgent

Returns a ReactAgent instance. Call agent.add_tools(...) before agent.invoke(...) / agent.ainvoke(...).

Also re-exported

So you don't need to know which sub-package a class lives in:

from autourgos_starter import ReactAgent, tool, OpenAIChatModel, ConversationBufferMemory

License

MIT