TableFlow Templates: Define Once, Extract Thousands

Why Templates Are a Game-Changer

TableFlow transforms document extraction into adaptive workflows. Instead of hardcoding rules for each format, you define the data structure you need, and AI figures out how to extract it from any layout.

Think of templates as smart blueprints. They define what data matters to your business, while the AI finds it no matter where it appears on the page.

The Old Problem

Before templates, document extraction involved rigid processes:

• • Parse Document A with Rule Set A
• • Parse Document B with Rule Set B

Each new format required custom development. Layout changes broke rules, and scaling increased complexity.

The Template Solution

Templates flip this model:

• • Define the data you need once
• • Apply the template to any document variation
• • AI adapts extraction automatically

One template handles hundreds of layouts, and changes enhance future extractions.

Key Template Components

Template Structure

Each template consists of:

• • Basic Information: Name, description, and workspace association
• • File Types: Supported formats (document, image, spreadsheet)
• • Template Settings: AI extraction behavior and processing options
• • Fields: Individual data points to extract
• • Tables: Complex tabular data structures

Field Definitions: Your Data Blueprint

Fields represent specific data points to extract. Each field includes:

  1{
  2  "name": "Invoice Number",
  3  "key": "invoice_number",
  4  "data_type": "string",
  5  "description": "The unique identifier for this invoice",
  6  "default_value": "",
  7  "extraction_guide": {
  8    "rules": [
  9      "Look for text containing 'Invoice #', 'INV-', or 'Invoice Number'",
 10      "Usually found in the top portion of the document"
 11    ],
 12    "examples": ["INV-2024-001", "Invoice #12345"]
 13  },
 14  "validations": [
 15    {
 16      "validate": "not_blank",
 17      "message": "Invoice number is required",
 18      "severity": "error"
 19    },
 20    {
 21      "validate": "regex",
 22      "options": {"pattern": "^(INV-)?[0-9]{4,}"},
 23      "message": "Invalid invoice number format",
 24      "severity": "warn"
 25    }
 26  ],
 27  "transformers": ["trim", "uppercase"]
 28}

Supported data types:

• • string: Text data
• • number: Numeric values (integers or decimals)
• • boolean: True/false values
• • date: Date values with various formats

Table Structures: Organizing Complex Data

Templates handle tabular data like line items or transactions:

  1{
  2  "name": "Line Items",
  3  "key": "line_items",
  4  "description": "Individual items or services on the invoice",
  5  "columns": [
  6    {
  7      "name": "Description",
  8      "key": "description",
  9      "data_type": "string",
 10      "validations": [
 11        {
 12          "validate": "not_blank",
 13          "message": "Item description required",
 14          "severity": "error"
 15        }
 16      ]
 17    },
 18    {
 19      "name": "Quantity",
 20      "key": "quantity",
 21      "data_type": "number",
 22      "validations": [
 23        {
 24          "validate": "min",
 25          "options": {"value": 0.01},
 26          "message": "Quantity must be positive",
 27          "severity": "error"
 28        }
 29      ]
 30    },
 31    {
 32      "name": "Unit Price",
 33      "key": "unit_price",
 34      "data_type": "number",
 35      "transformers": ["currency"]
 36    },
 37    {
 38      "name": "Total",
 39      "key": "line_total",
 40      "data_type": "number",
 41      "transformers": ["currency"]
 42    }
 43  ]
 44}

Extraction Guides: Teaching AI Your Business Logic

Extraction guides provide natural language instructions to guide AI:

  1{
  2  "extraction_guide": {
  3    "rules": [
  4      "This is a purchase order document",
  5      "Vendor information appears in the header section",
  6      "Look for 'Bill To' or 'Vendor' labels",
  7      "Line items are in the main table with quantity and price columns"
  8    ],
  9    "examples": [
 10      "Vendor: Acme Corporation",
 11      "Bill To: John's Hardware Store"
 12    ]
 13  }
 14}

Available Validations

TableFlow supports powerful validation rules:

• • not_blank: Ensures field has a value
• • email: Validates email format
• • phone: Validates phone number format
• • number: Ensures numeric value
• • decimal: Validates decimal precision
• • min/max: Range validation for numbers
• • length: String length validation
• • regex: Pattern matching
• • list: Value must be from predefined list
• • boolean: True/false validation

Transformers: Data Cleanup Made Easy

Transformers automatically clean and format extracted data:

• • trim: Remove leading/trailing whitespace
• • uppercase/lowercase: Case conversion
• • currency: Format as currency
• • date: Parse and format dates
• • phone: Normalize phone numbers
• • remove_special: Remove special characters

Best Practices

1. 1. Start Simple: Begin with essential fields and add complexity gradually
2. 2. Use Extraction Guides: Provide clear rules and examples for better accuracy
3. 3. Choose the Right AI Model: Different providers excel at different document types
4. 4. Validate Critical Data: Use error-level validations for must-have fields
5. 5. Test with Real Documents: Upload sample documents to refine your template
6. 6. Enable Extraction Editing: For critical workflows, allow manual corrections