API Reference

# Authentication

Every request requires an API key. Generate one from your dashboard. Each key is shown in full only once — store it securely.

Pass the key in one of two ways:

# Required Headers

For multipart uploads, the Content-Type is set automatically by your HTTP client.

# Endpoint

# Supported Formats

The 40 MB limit is the raw upload size (multer config). PDFs are decoded internally — the rendered image sent to the vision model is smaller.

# Request

Send your document as a multipart file upload or as JSON with base64-encoded bytes.

Multipart (file upload)

curl -X POST https://parseflow.cc/v1/document/analyze \
  -H "Authorization: Bearer ocrg_live_key" \
  -F "document=@invoice.pdf" \
  -F 'featureTypes=["FORMS","TABLES"]'

JSON (base64 bytes)

curl -X POST https://parseflow.cc/v1/document/analyze \
  -H "Authorization: Bearer ocrg_live_key" \
  -H "Content-Type: application/json" \
  -d '{
  "document": { "bytes": "iVBORw0KGgo...", "mimeType": "image/png" },
  "featureTypes": ["FORMS", "TABLES"]
}'

Request Parameters

# Response (Native Format)

The native response includes both an aggregate view and per-page detail. Only successful pages are billed — pages that fail to render are reported with an Error field and not charged.

Switch to Textract or DocAI formats from dashboard settings.

{
  "DocumentMetadata": { "Pages": 1 },
  "Text": "Invoice #1042\nBilled to: Acme Corp...",
  "Forms": [
    { "key": "Invoice Number", "value": "1042", "confidence": 0.98, "page": 1 }
  ],
  "Tables": [{ "rows": [["Item", "Qty", "Price"]], "page": 1 }],
  "Confidence": 0.97,
  "NeedsReview": false,
  "Pages": [{ /* per-page detail */ }],
  "Billing": { "totalCostCents": 1, "pages": 1, "failedPages": 0 }
}

# Response Formats

Switch between three response formats from Dashboard → Settings → Response format. No API changes needed.

# Feature Types

The featureTypes array controls extraction features. Billing is at the rate of the most expensive tier requested.

# Multi-Page PDFs

PDFs are auto-detected (by MIME type or file header) and split into one image per page. Each page runs through the same extraction pipeline independently.

• Up to 50 pages per request (safety cap)
• Processed in parallel with configurable concurrency (default 3)
• Per-page confidence check and escalation
• Only successful pages are billed — failed pages (render errors) return "Error" and are not charged
• Scanned/image-based PDFs and embedded-font PDFs work best

# Billing & Wallet

Prepaid wallet model. Credits are deducted after each successful extraction.

• Free trial — $5.00 credits on first signup
• Top up from dashboard via Razorpay Checkout
• 402 insufficient_balance if wallet can't cover estimated cost
• Failed pages are not billed
• Full transaction history in dashboard

# Error Codes

// 401 Unauthorized
{ "error": "unauthorized", "message": "Missing API key." }

// 402 Insufficient Balance
{ "error": "insufficient_balance",
  "message": "Insufficient wallet balance. You have $0.00 but this request needs at least $0.01.",
  "balanceCents": 0, "requiredCents": 1 }

// 400 Bad Request (missing document)
{ "error": "bad_request",
  "message": "Provide a file under the "document" field, or JSON body." }

// 413 File Too Large
{ "error": "entity_too_large",
  "message": "File exceeds the 40 MB size limit." }

// 429 Rate Limited
{ "error": "too_many_requests",
  "message": "Rate limit exceeded. Try again later." }

// 500 Internal Error
{ "error": "internal_error",
  "message": "Something went wrong. Please try again." }

// 502 Extraction Failed
{ "error": "extraction_failed",
  "message": "Document extraction failed. Check the document format or try again later." }

# Rate Limits

The extraction and usage endpoints are rate-limited to 60 requests per minute per IP address. Exceeded requests return HTTP 429 with a Retry-After header.

# Migration from Competitors

ParseFlow is designed as a drop-in replacement for AWS Textract and Google Document AI. Change your response format in dashboard settings — your existing integration code continues to work.

From AWS Textract

1. Generate a ParseFlow API key from the dashboard
2. Set response format to "AWS Textract-compatible" in settings
3. Replace your Textract endpoint URL with https://parseflow.cc/v1/document/analyze
4. Replace your AWS credentials with the Bearer token
5. That's it — your existing code that parses Blocks[] works unchanged

Note: ParseFlow does not require S3 buckets, IAM roles, Lambda functions, or complex infrastructure setup.

From Google Document AI

1. Generate a ParseFlow API key from the dashboard
2. Set response format to "Google Document AI-compatible" in settings
3. Replace your Document AI endpoint with https://parseflow.cc/v1/document/analyze
4. Replace your GCP credentials with the Bearer token
5. Your existing code that parses document.text and entities[] works unchanged

# FAQ

# Need Help?

Have questions or ran into an issue? Reach out to us:

# Next Steps