AgentOS SDK

Official Python SDK for integrating your custom AI agents with AgentOS.

Installation

pip install nl-agentos-sdk

(For private or internal usage, you can install directly from the repository using pip install git+<repo-url> or by building and sharing a wheel .whl file.)

Usage

import os
from agentos_sdk import AgentOS

def my_handler(action, payload):
    print(f"Executing {action} with {payload}")
    return {"success": True, "data": "Results from agent"}

client = AgentOS(url=os.getenv("AGENTOS_URL"), key=os.getenv("AGENTOS_KEY"))

manifest = {
    "agent_id": "my_agent",
    "version": "1.0.0",
    "team": "My Team",
    "capabilities": ["do_work"],
    "actions": [{"label": "Do work", "action": "DO_WORK"}],
    "ui_url": "http://localhost:3000"
}

client.run(manifest=manifest, handler=my_handler)

Release Checklist

This package uses PyPI Trusted Publishing via GitHub Actions. To release a new version:

1. Update Version: Bump version = "X.Y.Z" in pyproject.toml.
2. Commit: Commit your changes.
3. Tag & Push:

   git tag vX.Y.Z
   git push origin master --tags
   

4. Publish: The GitHub Action .github/workflows/pypi-publish.yml will automatically build the wheel and publish to PyPI using OIDC!

Initial PyPI Setup (Trusted Publishing) If you are deploying this for the first time, you must configure PyPI to trust this repository:

1. Go to pypi.org > Manage > Publishing.
2. Add a new "GitHub" publisher.
3. Set the repository name (Sachin-NLtechnologies/agentos-sdk), workflow filename (pypi-publish.yml), and environment name (pypi).
4. (No API tokens required to be stored in GitHub Secrets!)