Vendor templates - SnapIt Help Center

Vendor templates extract structured fields from known vendor layouts using OCR text and regular expressions. When a template matches enough required fields, SnapIt keeps the invoice on Tier 1. Tier 1 uses no LLM tokens, so extraction is free. Template confidence is calculated as:

matched_required_fields / total_required_fields × 100

SnapIt accepts the Tier 1 result when confidence reaches the threshold (default 85). Below the threshold, the invoice escalates to the LLM tiers.

Create and edit templates

Create the vendor

Add the supplier under Manage → Vendors if it does not exist yet.

Open the Templates page

Go to Manage → Templates. The vendor list shows which vendors already have a saved template and which are missing one.

Generate or edit the YAML

Select a vendor and edit the YAML in the editor panel. Click Generate Template to populate the editor with all supported fields and common regex patterns for that vendor.

Save

Click Save. New uploads for that vendor can take the Tier 1 path as soon as confidence clears the threshold.

Template structure

A template has two sections: fields for header values and tables.line_items for line-item rows.

vendor_name: "Acme Supplies"
currency: "MYR"
fields:
  invoice_number:
    regex: 'Invoice\s*No[:\s]+([A-Z0-9/-]+)'
    required: true
  invoice_date:
    regex: '(\d{1,2}[\/\-]\d{1,2}[\/\-]\d{2,4})'
    required: true
  total_amount:
    regex: 'TOTAL\s*:?\s*(?:MYR|RM)?\s*([\d,]+\.\d{2})'
    required: true
  tax_amount:
    regex: 'GST\s*/\s*SST\s*:?\s*([\d,]+\.\d{2})'
    required: false
tables:
  line_items:
    header_keywords: ["Description", "Item", "Product"]
    columns:
      - name: description
        index: 0
      - name: quantity
        index: 1
        type: number
      - name: unit_price
        index: 2
        type: number
      - name: line_total
        index: 3
        type: number

Header fields

Each entry under fields names a capture and provides a regex. The first capture group becomes the field value.

required: true fields count toward the confidence score. Mark the fields that reliably appear on every invoice from that vendor as required.
required: false fields are extracted when present but never lower confidence.

Line-item rules

tables.line_items maps OCR table columns to invoice line fields:

header_keywords identifies the line-item table. The extractor looks for a table whose header row contains one of these keywords.
columns maps each column by zero-based index to a field name. Set type: number for quantity and amount columns so values are parsed numerically.

Manage templates through the API

Workspace users can also manage templates with a session JWT. These endpoints use the Bearer JWT from login, not a programmatic API key.

curl -X PUT https://api.trysnapit.com/api/vendors/<vendor-id>/template \
  -H "Authorization: Bearer <jwt>" \
  -H "Content-Type: text/plain" \
  --data-binary @my-template.yml

curl https://api.trysnapit.com/api/vendors/<vendor-id>/template \
  -H "Authorization: Bearer <jwt>"

curl https://api.trysnapit.com/api/vendors/template-ids \
  -H "Authorization: Bearer <jwt>"

Improve accuracy and cost over time

Every invoice stores the tier and provider that produced its data, so you can audit where your extraction spend goes.

Watch the Tier Distribution chart on the Dashboard. A high share of Tier 2 or Tier 3 invoices means those vendors are missing templates or their templates match too few required fields.
After improving a template, reprocess a past invoice from that vendor (or re-upload a sample) and confirm it now resolves at Tier 1.
Anchor regexes to stable labels on the invoice, such as Invoice No or TOTAL, rather than to values that change between invoices.
Keep required: true limited to fields that appear on every invoice from the vendor. A required field that only appears sometimes drags confidence below the threshold and forces LLM escalation.

​Create and edit templates

​Template structure

​Header fields

​Line-item rules

​Manage templates through the API

​Improve accuracy and cost over time