Chapter 1: Tool Registration Metadata

Welcome to the first chapter of building SleepTool!

Before we can teach our tool how to work, we first need to tell the system who the tool is.

The Motivation: The ID Badge Problem

Imagine walking into a large office building. You might have amazing skills—maybe you are a master chef or a brilliant mathematician. However, if you don't have an ID Badge with your name and job title, security won't let you in, and no one will know to call you when they need a meal or a calculation.

In our software, the "System" is the security guard, and the "AI" is the manager looking for workers.

Tool Registration Metadata is simply creating that ID Badge. It solves the problem of discoverability. It provides the essential labels—Name and Role—that allow the larger system to list the tool in a menu so the AI knows it is available to use.

Key Concept: The Exports

We define this metadata in a file called prompt.ts. We use specific constants that act as the interface identifiers.

1. The Tool Name

The name is the unique identifier. It's how the AI refers to this specific tool in its internal logic. It needs to be short, unique, and capitalized.

// In prompt.ts

// This is the unique ID for our tool
export const SLEEP_TOOL_NAME = 'Sleep'

Explanation: By writing export, we are telling the code: "Make this value public so other files can see it." We name our tool 'Sleep'.

2. The Description

The name tells us who it is, but the description tells us what it does. This is a one-sentence summary. It appears in the "menu" of tools provided to the AI.

// In prompt.ts

// A short summary for the system's tool list
export const DESCRIPTION = 'Wait for a specified duration'

Explanation: This string isn't the full set of instructions (we cover those in Tool Behavior Definition). This is just the "elevator pitch" so the system can quickly categorize the tool.

How It Works: The Registration Flow

When you start the application, the system needs to "onboard" all available tools. Here is what happens under the hood when the metadata is read.

1. System Startup: The main application scans the folder.
2. Importing: It opens prompt.ts.
3. Reading Metadata: It looks specifically for SLEEP_TOOL_NAME and DESCRIPTION.
4. Registration: It creates an entry in its memory. Now, when the AI asks "What can I do?", the system can reply: "You can use the Sleep tool to Wait for a specified duration."

Internal Implementation Details

While our code in prompt.ts is simple, it connects to a larger architecture.

The file prompt.ts serves as the Single Source of Truth. Instead of typing the string "Sleep" in multiple places (which leads to typos and bugs), we define it once here.

You might notice other imports in the file, such as TICK_TAG:

import { TICK_TAG } from '../../constants/xml.js'

// ... exports ...

This import is a helper we will use later for Periodic Heartbeat Handling. For now, just know that prompt.ts is the central hub where we define the tool's identity.

Example Input/Output: If we were to run a script that checks this file:

• Input: Load module prompt.ts.
• Output: An object containing { SLEEP_TOOL_NAME: 'Sleep', DESCRIPTION: 'Wait...' }.

The system takes this output and builds the UI or API definition for the AI.

Conclusion

You have successfully created the "ID Badge" for your tool!

• We defined SLEEP_TOOL_NAME so the system knows the tool's name.
• We defined DESCRIPTION so the system knows the tool's role.

However, an ID badge doesn't tell the AI how to use the tool or the nuances of its personality. For that, we need to write the detailed prompt instructions.

Next Chapter: Tool Behavior Definition

Generated by Code IQ