Schema

Screenshots

Schema Worker Interface - Configure data extraction schema

Overview

The Schema worker generates structured data by extracting information from input text based on a predefined schema. It uses TypeChat and Zod validation to ensure the output conforms to the specified structure.

Key Features

• Dynamic Schema Definition: Define custom data structures using connected handlers
• Type Safety: Built-in validation using Zod schemas
• Flexible Field Types: Support for strings, numbers, booleans, arrays, and enums
• AI-Powered Extraction: Uses OpenAI models for intelligent data parsing

Configuration

Parameters

• Model: OpenAI model to use (default: gpt-4o)

Input/Output

• Input: Raw text content to extract data from
• JSON Output: Structured JSON object matching the defined schema
• Dynamic Fields: Additional output fields based on connected handlers

Use Cases

1. Contact Information Extraction

Extract structured contact details from unformatted text:

• Name, email, phone number
• Address components
• Company information

2. Product Data Parsing

Structure product information from descriptions:

• Price, features, specifications
• Categories and tags
• Availability status

3. Event Information Processing

Extract event details from text:

• Date, time, location
• Attendee information
• Requirements and restrictions

4. Document Analysis

Parse structured data from documents:

• Form field extraction
• Data normalization
• Content categorization

How It Works

1. Schema Definition: Connected handlers define the output structure
2. Type Mapping: Field types are converted to Zod validation schemas
3. AI Processing: Input text is processed using the specified model
4. Validation: Output is validated against the defined schema
5. Field Population: Extracted data populates the connected output fields

Best Practices

Schema Design

• Use descriptive field names that match expected content
• Provide clear prompts for each field to guide extraction
• Choose appropriate data types for validation

Input Preparation

• Ensure input text contains the information you want to extract
• Use consistent formatting when possible
• Provide sufficient context for accurate extraction

Model Selection

• Use gpt-4o for complex schema extraction
• Consider gpt-3.5-turbo for simpler, faster processing
• Test different models for optimal accuracy

Validation Handling

• Design schemas to handle optional fields gracefully
• Use enum types for fields with limited valid values
• Consider array types for lists or multiple values

Example Schema

// Example: Contact extraction schema
{
  name: string,           // Person's full name
  email: string,          // Email address
  phone: string,          // Phone number
  company: string,        // Company name
  role: string,           // Job title
  tags: string[]          // Category tags
}

Integration Tips

• Chain with Text Workers: Process raw content before schema extraction
• Combine with Display: Show extracted data in formatted views
• Use with State: Store extracted data for later use
• Connect to API: Send structured data to external systems

Error Handling

The worker includes built-in validation and will:

• Skip extraction if input is empty
• Handle partial matches gracefully
• Provide structured error information
• Maintain type safety throughout the process

Performance Considerations

• Larger schemas may require more processing time
• Complex nested structures need more capable models
• Consider breaking large schemas into smaller, focused ones
• Cache results when processing similar content repeatedly