nodes

Node Development Rules (/src/nodes)

This file provides guidelines for creating and contributing custom tools in the /src/nodes directory.

Core Concepts

The nodes in AgentDock are specialized tool nodes that extend the system's capabilities:

• Each node is self-contained in its own directory
• Nodes follow the Vercel AI SDK pattern
• They provide component-based output formatting

Directory Structure

The actual /src/nodes directory contains:

• [README.md](mdc:src/nodes/README.md): Overview and architecture documentation
• [custom-tool-contributions.md](mdc:src/nodes/custom-tool-contributions.md): Detailed guide for contributors
• [registry.ts](mdc:src/nodes/registry.ts): Node registration system
• [init.ts](mdc:src/nodes/init.ts): Node initialization logic
• [types.ts](mdc:src/nodes/types.ts): Shared types for nodes
• Individual tool directories (e.g., search/, weather/, cognitive-tools/)

Implementation Guidelines

Tool Implementation Pattern

Each tool follows this pattern:

// index.ts
import { z } from 'zod';
import { Tool } from '../types';
import { MyComponent } from './components';

// 1. Define parameters schema
const myToolSchema = z.object({
  input: z.string().describe('What this input does')
});

// 2. Create and export your tool
export const myTool: Tool = {
  name: 'my_tool',
  description: 'What this tool does',
  parameters: myToolSchema,
  async execute({ input }) {
    // 3. Get your data
    const data = await fetchData(input);
    
    // 4. Use your component to format output
    return MyComponent(data);
  }
};

// 5. Export for auto-registration
export const tools = {
  my_tool: myTool
};

Best Practices

• Use proper error handling with try/catch blocks
• Create components to format tool output
• Store API keys in environment variables, never hardcode them
• Implement server-side API calls for security
• Follow TypeScript best practices with proper typing
• Include comprehensive JSDoc comments
• Create fallbacks for when external services are unavailable