AI Task

This page gives you an overview of the AI Task in Agent Builder

Overview

AI Tasks are designed for intelligent processing operations using artificial intelligence capabilities. These tasks can handle complex operations like text analysis, content generation, and data processing.

Visual Example

Configuration Structure

{
  "type": "TASK",
  "block": {
    "name": "AI Task Name",
    "type": "TASK",
    "instructions": "Detailed instructions for the AI",
    "input_parameters": [],
    "expected_output": [],
    "dependencies": [],
    "error_policy": "RAISE"
  }
}

Required Fields

Field	Type	Description	Required
name	string	Unique identifier for the task	Yes
instructions	string	Clear instructions for processing	Yes
input_parameters	array	Input configuration	Yes
expected_output	array	Output configuration	Yes
dependencies	array	List of dependent tasks	No
error_policy	string	Error handling strategy	No

Input Parameters

Text Input

{
  "name": "input_text",
  "type": "STRING",
  "description": "Text content to process",
  "required": true,
  "source": "task_config"
}

Structured Data Input

{
  "name": "data_object",
  "type": "OBJECT",
  "description": "Structured data for analysis",
  "properties": [
    {
      "name": "field1",
      "type": "STRING",
      "description": "First field"
    },
    {
      "name": "field2",
      "type": "INTEGER",
      "description": "Second field"
    }
  ]
}

Array Input

{
  "name": "items",
  "type": "ARRAY",
  "description": "List of items to process",
  "items": {
    "type": "STRING"
  }
}

Output Parameters

Basic Output

{
  "name": "result",
  "type": "STRING",
  "description": "Processing result"
}

Analysis Output

{
  "name": "analysis",
  "type": "OBJECT",
  "properties": {
    "sentiment": {
      "type": "STRING",
      "description": "Detected sentiment"
    },
    "confidence": {
      "type": "FLOAT",
      "description": "Confidence score"
    },
    "categories": {
      "type": "ARRAY",
      "items": {
        "type": "STRING"
      },
      "description": "Detected categories"
    }
  }
}

Common Use Cases

1. Text Analysis

{
  "name": "Analyze Text",
  "instructions": "Analyze the input text for sentiment and key topics",
  "input_parameters": [
    {
      "name": "text",
      "type": "STRING",
      "description": "Text to analyze",
      "required": true
    }
  ],
  "expected_output": [
    {
      "name": "analysis",
      "type": "OBJECT",
      "properties": {
        "sentiment": "STRING",
        "topics": "ARRAY",
        "summary": "STRING"
      }
    }
  ]
}

2. Content Generation

{
  "name": "Generate Content",
  "instructions": "Generate content based on provided parameters",
  "input_parameters": [
    {
      "name": "topic",
      "type": "STRING",
      "description": "Content topic",
      "required": true
    },
    {
      "name": "style",
      "type": "STRING",
      "description": "Writing style",
      "required": true
    },
    {
      "name": "length",
      "type": "INTEGER",
      "description": "Target word count",
      "required": true
    }
  ],
  "expected_output": [
    {
      "name": "content",
      "type": "STRING",
      "description": "Generated content"
    }
  ]
}

Best Practices

1. Writing Instructions

✅ Do:

• Be specific and clear
• Include examples when helpful
• Specify format requirements
• Define expected behavior

❌ Don’t:

• Use vague descriptions
• Omit important details
• Assume context

2. Input Configuration

✅ Do:

• Validate input types
• Set required fields
• Provide clear descriptions
• Use appropriate types

❌ Don’t:

• Use overly complex structures
• Skip parameter descriptions
• Ignore validation

3. Output Configuration

✅ Do:

• Define clear structure
• Include all necessary fields
• Document format requirements
• Handle error cases

❌ Don’t:

• Use ambiguous types
• Omit error handling
• Ignore edge cases

Error Handling

Error Policies

{
  "error_policy": "RAISE", // Stop on error
  "error_policy": "IGNORE", // Continue execution
  "error_policy": "RETRY" // Retry on failure
}

Validation

1. Input validation
2. Output validation
3. Error reporting

Examples

Text Classification

{
  "name": "Classify Text",
  "instructions": "Classify the input text into predefined categories",
  "input_parameters": [
    {
      "name": "text",
      "type": "STRING",
      "description": "Text to classify",
      "required": true
    },
    {
      "name": "categories",
      "type": "ARRAY",
      "items": {
        "type": "STRING"
      },
      "description": "Available categories",
      "required": true
    }
  ],
  "expected_output": [
    {
      "name": "classification",
      "type": "OBJECT",
      "properties": {
        "category": "STRING",
        "confidence": "FLOAT",
        "alternatives": "ARRAY"
      }
    }
  ]
}

Common Issues and Solutions

Issue	Solution
Unclear Instructions	Provide specific, detailed instructions
Input Validation Fails	Check input types and requirements
Output Mismatch	Verify output structure matches requirements
Performance Issues	Optimize input/output size

Integration with Other Tasks

Passing Results