Document Processing

Pull text out of images and PDFs with OCR, run it through the same moderation checks as your text content, and turn long documents into short, readable summaries. Two dedicated endpoints, each metered on its own quota.

Text Extraction

OCR images and PDFs into machine-readable text.

Built-in Moderation

Extracted text is checked for spam, toxicity, and language.

AI Summaries

Condense long documents into a concise summary.

Text Extraction (OCR)

Send one or more file URLs and receive the recognized text. Unless you opt out, the extracted text is also moderated using your project's text settings, so a single call both reads and screens the document.

Endpoint

POST https://api.discuse.com/api/v2/ocr

Request

{
  "file_urls": [
    "https://example.com/scanned-id.jpg",
    "https://example.com/contract.pdf"
  ]
}

Skip Moderation

Set moderate to false to extract text only, without running content checks.

{
  "file_urls": ["https://example.com/receipt.png"],
  "moderate": false
}

Request Parameters

Parameter	Type	Required	Description
`file_urls`	string[]	Yes	Image or PDF URLs to OCR (max 5 per request)
`moderate`	boolean	No	Run text checks on the extracted text. Defaults to true.

Response

{
  "text": "ACME Terms of Service\nEffective 1 January 2026 ...",
  "has_text": true,
  "num_files": 2,
  "has_violations": false,
  "results": {
    "hits": false,
    "sentiment": {
      "hit": false,
      "toxic": 0.02,
      "profanity": 0.01,
      "threat": 0.00,
      "insult": 0.01
    },
    "spam_finder": {
      "is_spam": false,
      "confidence": 0.04
    },
    "language": {
      "detected": "en",
      "confidence": 0.99
    }
  }
}

Spam Detected in Document

{
  "text": "buy now!!! free crypto giveaway, dm me ...",
  "has_text": true,
  "num_files": 1,
  "has_violations": true,
  "results": {
    "hits": true,
    "spam_finder": {
      "is_spam": true,
      "confidence": 0.93
    }
  }
}

Response Fields

Field	Type	Description
`text`	string	Concatenated text recognized across all files
`has_text`	boolean	True if any non-empty text was recognized
`num_files`	number	Number of files successfully OCR'd (each counts as one extraction against quota)
`has_violations`	boolean	True if moderation flagged the extracted text
`results`	object	Moderation result (sentiment, spam, language). Omitted when `moderate` is false or no text was found.

Summarization

Generate a concise summary of a document. Provide raw text directly, or supply file_urls and the documents are OCR'd before summarizing.

Endpoint

POST https://api.discuse.com/api/v2/summarize

Summarize a File

{
  "file_urls": ["https://example.com/quarterly-report.pdf"]
}

Summarize Raw Text

{
  "text": "The board convened on 14 March to review ... (long document text)"
}

Request Parameters

Parameter	Type	Required	Description
`text`	string	Conditional	Raw document text. Required if no `file_urls` are given.
`file_urls`	string[]	Conditional	Document URLs to OCR then summarize (max 5). Required if no `text` is given.

Response

{
  "summary": "The quarterly report covers revenue growth of 18% YoY, driven by enterprise renewals. Operating costs rose 6% on headcount, leaving margin broadly flat. The board approved a buyback and flagged FX exposure as the key downside risk."
}

Try It Free (No API Key)

Both capabilities have public demo endpoints you can call without an API key. They are rate limited per IP, and PDF uploads are capped at 20 pages.

Demo OCR

curl -X POST https://api.discuse.com/api/v2/demo/ocr \
  -H "Content-Type: application/json" \
  -d '{ "file_url": "https://example.com/sample.pdf" }'

Demo Summarization

curl -X POST https://api.discuse.com/api/v2/demo/summarize \
  -H "Content-Type: application/json" \
  -d '{ "file_url": "https://example.com/sample.pdf" }'

Demo Limits

Limit	Value
PDF pages	20 max
Files per request	2 max
Rate limit	10 requests / minute per IP

For higher limits, multi-file batches, and usage tracking, use the authenticated /api/v2/ocr and /api/v2/summarize endpoints with an API key.

Authentication: both endpoints require your API key in the X-API-Key header. File URLs must be publicly accessible; the API fetches each one directly.

Limitations

Limit	Value
Max files per request	5
Supported file types	JPG, PNG, WebP, PDF
OCR metering	One extraction per file
Summary metering	One summary per request

Moderate Plain Text Too

Already have the text? Send it straight to the text analysis endpoint for sentiment, spam, and language checks.

Text Analysis Docs