Zerodoc Documentation

Quickstart

Three steps to your first structured response.

1. 1
   Create an API key

   Create a free account and generate a key from your dashboard.

2. 2
   Send a document

   POST a file (or URL) to /v1/extract with your key.

3. 3
   Read the response

   Get structured fields, full text and a retention proof block.

curl -X POST "https://api.zerodoc.io/v1/extract" \
  -H "X-API-Key: zk_your_api_key" \
  -F "file=@invoice.pdf;type=application/pdf"

Authentication

Authenticate every request with your API key in the X-API-Key header. Keys are scoped to your account, created and revoked from your dashboard.

X-API-Key: zk_your_api_key

API reference

Extract text and structured fields from a document. Send a file or a url to fetch (if both are sent, the file wins). Supported types: PDF, PNG, JPEG, TIFF, WebP.

Parameters

Request example

curl -X POST "https://api.zerodoc.io/v1/extract" \
  -H "X-API-Key: zk_your_api_key" \
  -F "file=@invoice.pdf;type=application/pdf"

{
  "request_id": "req_3abc...",
  "status": "ok",
  "processing_time_ms": 412,
  "doc_type": "invoice",
  "document": { "mime_type": "application/pdf", "page_count": 1, "size_bytes": 24872 },
  "text": "ACME Trading Ltd\nInvoice Number: INV-2026-0042\n...",
  "blocks": [
    { "text": "Total Due: 1200.00", "confidence": 0.99, "page": 0,
      "bbox": { "x": 0.05, "y": 0.82, "width": 0.30, "height": 0.03 } }
  ],
  "fields": {
    "invoice_number": { "value": "INV-2026-0042", "confidence": 0.9 },
    "total_amount": { "value": 1200.0, "confidence": 0.85 },
    "taxes": [ { "amount": 200.0, "rate": 20.0, "base": 1000.0 } ],
    "locale": { "country": "GB", "currency": "GBP" }
  },
  "extraction_warnings": [],
  "mean_confidence": 0.97,
  "extraction_confidence": 0.88,
  "retention": { "stored": false, "policy": "zero-retention" }
}

/v1/classify

Detect a document’s type so you can route it automatically. Returns the best match from invoice, credit_note, receipt, purchase_order, bank_statement, delivery_note — or an honest other when the evidence is too weak, rather than a guess.

Takes the same file / url input as /v1/extract, and shares the response envelope (request_id, status, document, retention, …). The response is deliberately lean — no document text or blocks — so it stays cheap to route on.

Request example

curl -X POST "https://api.zerodoc.io/v1/classify" \
  -H "X-API-Key: zk_your_api_key" \
  -F "file=@document.pdf;type=application/pdf"

Response

{
  "request_id": "req_9fe2...",
  "status": "ok",
  "processing_time_ms": 241,
  "document": { "mime_type": "application/pdf", "page_count": 1, "size_bytes": 18240 },
  "doc_type": "invoice",
  "confidence": 0.7,
  "candidates": [
    { "doc_type": "invoice", "score": 0.7 },
    { "doc_type": "receipt", "score": 0.3 }
  ],
  "retention": { "stored": false, "policy": "zero-retention" }
}

/v1/split

Detect document boundaries in a multi-page upload — one PDF containing several documents comes back as segments, each with its page range and type. Boundaries are found from page markers (“Page 1 of 3”), document numbers changing between pages, and per-page type changes; ties keep pages together rather than cutting a document in half.

Takes the same file / url input as /v1/extract, and shares the response envelope (request_id, status, document, retention, …). Feed each segment’s pages to /v1/extract to get its fields.

Request example

curl -X POST "https://api.zerodoc.io/v1/split" \
  -H "X-API-Key: zk_your_api_key" \
  -F "file=@batch.pdf;type=application/pdf"

Response

{
  "request_id": "req_41bd...",
  "status": "ok",
  "processing_time_ms": 690,
  "document": { "mime_type": "application/pdf", "page_count": 5, "size_bytes": 88410 },
  "segments": [
    { "segment": 0, "pages": [0, 1], "page_count": 2, "doc_type": "invoice", "confidence": 0.82 },
    { "segment": 1, "pages": [2], "page_count": 1, "doc_type": "receipt", "confidence": 1.0 },
    { "segment": 2, "pages": [3, 4], "page_count": 2, "doc_type": "invoice", "confidence": 0.9 }
  ],
  "retention": { "stored": false, "policy": "zero-retention" }
}

/v1/crop

Digitize several documents scanned on a single page — e.g. three receipts on one flatbed A4. Each detected item is returned as its own image, positioned and classified, ready to process individually.

Takes the same file / url input as /v1/extract, and shares the response envelope (request_id, status, document, retention, …). Accepts images and PDFs only (text/plain is rejected — there are no pixels to crop). Items need readable text and whitespace between them to be detected.

Request example

curl -X POST "https://api.zerodoc.io/v1/crop" \
  -H "X-API-Key: zk_your_api_key" \
  -F "file=@flatbed-scan.png;type=image/png"

Response

{
  "request_id": "req_efb3...",
  "status": "ok",
  "processing_time_ms": 14356,
  "document": { "mime_type": "image/png", "page_count": 1, "size_bytes": 402133 },
  "items": [
    { "item": 0, "page": 0,
      "bbox": { "x": 0.04, "y": 0.03, "width": 0.38, "height": 0.21 },
      "doc_type": "receipt", "confidence": 1.0,
      "image_mime_type": "image/png", "image_base64": "iVBORw0KGgo..." },
    { "item": 1, "page": 0,
      "bbox": { "x": 0.04, "y": 0.67, "width": 0.33, "height": 0.17 },
      "doc_type": "invoice", "confidence": 1.0,
      "image_mime_type": "image/png", "image_base64": "iVBORw0KGgo..." }
  ],
  "retention": { "stored": false, "policy": "zero-retention" }
}

Errors

Errors use standard HTTP status codes and return a JSON body with a detail field explaining what went wrong. Processing failures (500) include a request_id you can quote to support.

Supported documents

Field extraction currently supports UK & EU invoices; more extraction types are on the roadmap. Classification (/v1/classify) already recognises receipts, purchase orders, bank statements, credit notes and delivery notes for routing.