horizondesk 1.1.0

SDK and CLI for building Horizon Desk AI Plugins

horizondesk SDK

This SDK provides the official interface for building plugins for Horizon Desk. It allows developers to extend OmniAgent's capabilities with custom tools and connect external AI workers.

Installation

pip install horizondesk

Quick Start: Create your first plugin

1. Initialize

Run the following command to scaffold a new plugin project:

horizondesk init MyAmazingPlugin

2. Define a Tool

Edit main.py to create your own tools:

from horizondesk import BaseTool, HorizonPlugin

class WeatherTool(BaseTool):
    def __init__(self):
        super().__init__("GetWeather", "Gets current weather for a city. Input: 'city'.")

    def execute(self, city=None, payload=None):
        return f"It is 22°C and sunny in {city}."

def register_tools(agent):
    plugin = HorizonPlugin("MyAmazingPlugin")
    plugin.add_tool(WeatherTool())
    plugin.register_all(agent)

3. Run the Workshop

Launch the high-fidelity GUI Workshop to test your plugin:

horizondesk run horizon_plugin.raf

Security & Privacy

• SecretStorage: Always use SecretStorage.get_secret("KEY") to access API keys. Do not hardcode them.
• Redaction: Use SecretStorage.redact_pii(text) before sending data to external workers.
• Auditing: All plugin tools are visible to the user and subject to manual intervention (Alt+F7).

Connecting External Agents

To link your own AI worker, create a tool that performs a secure HTTP request to your backend:

import requests
from horizondesk import BaseTool, SecretStorage

class ExternalBrainBridge(BaseTool):
    def __init__(self):
        super().__init__("AskMyBot", "Consults my specialized AI model.")

    def execute(self, **kwargs):
        api_key = SecretStorage.get_secret("MY_BOT_API_KEY")
        response = requests.post("https://my-api.com/v1/chat", json=kwargs, headers={"Auth": api_key})
        return response.json().get("reply")