This page provides a high-level overview of Scoopika’s Multi-agent Boxes (also known as Smart Boxes). For in-depth documentation, refer to the dedicated Smart boxes documentation.

What are Multi-agent Boxes?

Imagine a team of AI assistants working together! Multi-agent boxes are virtual containers that house multiple agents. The box intelligently selects the most suitable agent for the task in hand.

Important Note: Multi-agent boxes are currently a demo feature. Please contact us if you face any issues working with them.

Creating Multi-agent Boxes

You can visit the Scoopika platform to create a new box, The process is simple:

  1. Name your box: Give it a clear and descriptive name.
  2. Select the driving LLM: Choose the Large Language Model that will coordinate the agents within the box.
  3. Assemble your team: Select the agents you want to include from your existing agents.

Initializing a Box Instance

Before you can utilize a box in your application, ensure you have initialized a Scoopika instance. Then, proceed with initializing the box instance:

import { Box } from "@scoopika/scoopika";

const box = new Box("BOX_ID", scoopika);

Running the Box

Running a multi-agent box follows a similar approach to running individual agents. However, box instances introduce additional functionalities through streaming hooks.

const response = await box.run({
    inputs: {
        message: "Hello" // Required field for box inputs
    }
})

console.log(response);

Tools

Multi-agent boxes offer the flexibility to equip agents with tools. You can add tools globally, accessible by all agents within the box, or assign them to specific agents.

// Global tool (all agents)
box.addGlobalTool(toolFunc, toolSchema);

// Tool for a specific agent
box.addTool("AGENT_NAME", toolFunc, toolSchema);

For detailed instructions on creating tool schemas, refer to the tools documentation.

Was this page helpful?