API Documentation Sample: Policy Extraction Endpoint

Demonstration writing sample. This documents the internal extraction interface of my AI Policy Extraction System in public-API style to demonstrate developer documentation. It is not a hosted public API.

Extract policy fields

Extracts structured policy information from a captured document.

POST /v1/extract

Request body

Field	Type	Required	Description
document_text	string	yes	Plain text of the captured policy document.
institution	string	yes	Institution identifier from the source register.
capture_date	string	yes	ISO 8601 date the document was captured.
schema_version	string	no	Extraction schema version. Defaults to the latest.

Example request

{
  "document_text": "Students may use generative AI tools for...",
  "institution": "example-university",
  "capture_date": "2026-03-14"
}

Response

Returns an extraction record. Every record is created with "verification": "unverified" — records must pass human verification before research use.

{
  "institution": "example-university",
  "capture_date": "2026-03-14",
  "schema_version": "1.2",
  "fields": {
    "permission_model": "permitted-with-conditions",
    "disclosure_required": true,
    "verification_expected": true,
    "defined_terms": ["generative AI", "unauthorised assistance"]
  },
  "confidence_flags": [],
  "verification": "unverified"
}

Errors

Status	Meaning	Resolution
400	Missing required field or empty document_text.	Check the request body against the table above.
422	Model output failed schema validation.	Retry; if persistent, file the document for manual coding.
429	Rate limit exceeded.	Respect the Retry-After header.

Notes for integrators

• Treat all extracted values as claims requiring verification, not as data.
• confidence_flags lists fields the extractor marked as uncertain; route flagged records to verification first.