OCR-Doc-Parser

Extract structured data from receipts, tax invoices, and bank statements into clean JSON with field-level confidence flags and arithmetic reconciliation.

Open @OCR-Doc-Parser on Poe →

What You Provide

Upload or attach a photo, smartphone capture, or PDF scan of a physical receipt, commercial tax invoice (including Indian GSTIN bills), or bank statement table.

Retail, grocery, and restaurant bills
B2B tax invoices with CGST/SGST/IGST
Tabular bank statement transactions

What You Receive

Clean structured JSON containing normalized vendor name, purchase date, time, invoice number, itemized line items, subtotal, taxes, and grand total.

Per-field confidence ratings (high, medium, low)
Arithmetic check: subtotal + tax = total
Copyable JSON block ready for accounting sheets

Copyable Starter Prompts

Attach your document image and paste any of these prompt patterns:

PROMPT 01 · GENERAL EXPENSE

Extract the vendor, date, line items, tax, and grand total.

Standard full extraction for restaurant, taxi, and retail receipts.

PROMPT 02 · TAX INVOICE RECONCILE

/receipt Reconcile line items, GST tax breakdown, and verify arithmetic total.

Triggers explicit GSTIN parsing and line-item summation checks.

PROMPT 03 · BANK STATEMENT TABLE

/statement Extract transactions into a markdown table with Date, Description, and Amount.

Reconstructs tabular banking and credit ledger rows.

Verified Synthetic Output Example

Sample extraction produced from a synthetic cafe receipt fixture:

{
  "document_type": "receipt",
  "vendor": {
    "value": "ACME CAFE & ROASTERY",
    "confidence": "high"
  },
  "date": {
    "value": "2026-08-14",
    "confidence": "high"
  },
  "line_items": [
    { "description": "Flat White", "quantity": 1, "amount": 4.50 },
    { "description": "Almond Croissant", "quantity": 2, "amount": 9.00 }
  ],
  "subtotal": { "value": 13.50, "confidence": "high" },
  "tax": { "value": 1.15, "confidence": "high" },
  "total": { "value": 14.65, "confidence": "high" },
  "arithmetic_reconciled": true
}

Limitations & Responsible Use

Apex Forge OCR is designed for bookkeeping convenience and developer data workflows. Severely blurred photos, reflective thermal paper glare, or crumpled receipts may cause character misrecognition. The tool provides confidence scores and arithmetic warnings; users must review figures before submitting official tax returns or regulatory filings. It does not replace official KYC or certified accounting validation.

Frequently Asked Questions

What image formats are supported by Apex Forge OCR?

The engine processes standard JPEG, PNG, WEBP, and single-page scans up to 10MB.

Does Apex Forge OCR store uploaded receipts or customer documents?

No. Images are processed ephemerally in RAM during the active query turn and discarded immediately upon stream completion.

How does the tool verify invoice arithmetic?

The parser sums extracted line items and taxes, comparing the sum against the detected grand total within standard rounding tolerance.