coreouto 0.3.3

A minimal, extensible agent library for Python. Five philosophies: minimalism, extensibility, explicitness, fragmentation, conciseness.

coreouto

A minimal, extensible agent library for Python.

Built on five philosophies: minimalism, extensibility, explicitness, fragmentation, conciseness.

The whole library reduces to one idea: an agent is called with a message, runs an internal loop, and returns its response when the model wraps its final answer in <finish>...</finish> tags. Everything else — providers, tools, presets, hooks, multi-agent — is an opt-in extension.

import os

import coreouto as co
from coreouto.providers.openai import OpenAIProvider

@co.register_tool("search")
def search(query: str) -> str:
    """Search the web for `query`."""
    return f"<results for {query}>"

co.register_provider("minimax", OpenAIProvider(
    api_key=os.environ["MINIMAX_API_KEY"],
    base_url="https://api.minimax.io/v1",
))

preset = co.register_agent_preset(
    "researcher", model="MiniMax-M3", provider="minimax",
    system_prompt="You are a research assistant.",
    tools=["search"],
)
response = co.Agent(preset.to_config()).call_sync("Find me recent news about fusion energy.")
print(response.content)

Install

pip install coreouto
# with providers
pip install coreouto[openai]
pip install coreouto[anthropic]
pip install coreouto[google]
pip install coreouto[all]

Documentation

See docs/ for the full documentation set:

• Philosophy — the five principles
• Quickstart
• Agent
• Providers
• Tools
• Presets
• Hooks
• Multi-agent

License

Apache 2.0 — see LICENSE.