status: operational latency: 42ms uptime: 99.98%
deploy: 9a3d9d5 docs ↗
┬ ┬┌┬┐┬  
│││ │ │  
└┴┘ ┴ ┴  
watch-the-ink
~/watch-the-ink > ⌘K
~/watch-the-ink / blog / designing-your-first-form-template

Designing Your First Form Template

From Blank Template to Extraction Schema

The watchtheink Form Designer lets you model any paper form as a structured hierarchy before a single scan is submitted. This post walks through creating your first template end to end.

Understanding the Hierarchy

Every template follows the same shape:

FormTemplate
  └─ FormGroup         (logical section group, e.g. "Personal Details")
       └─ FormSection  (a visual block on the page, e.g. "Contact Information")
            └─ FormField  (an individual field, e.g. "Full Name")

Groups and sections are organisational — they don't appear on the paper form. They tell the extraction engine how to group confidence scores and structure the JSON output.

Step 1: Create a Template

Open the Designer at /designer and fill in the template metadata at the top:

  • Name — a stable identifier (e.g. quiz-answer-sheet-v2)
  • Description — human-readable context for reviewers
  • Version — increment this whenever the field layout changes

Step 2: Add Groups and Sections

Click Add Group in the left panel to create your first group. Give it a descriptive label. Then click Add Section inside that group to create the first visual block.

Step 3: Define Fields

Inside a section, click Add Field. The right panel opens the field editor. For each field:

Property What it controls
FieldType How the AI interprets the region (Text, Checkbox, Signature, ...)
Label The key used in extracted JSON output
isRequired Whether a missing value should flag the scan for review
maxOccurrences 1 = single value; >1 = repeating rows (e.g. answer grids)

Step 4: Save and Submit a Scan

Hit Save in the top bar. Your template is now available in the Template Library and in the scan submission form. Navigate to /scan to submit your first scan against it.

Next Steps

Once you have a few scans processed, the confidence scores will tell you where the AI is uncertain. Use those signals to refine your template — add validation rules, split ambiguous sections, or adjust field types.

← back to blog