> ## Documentation Index > Fetch the complete documentation index at: https://docs.idmetagroup.com/llms.txt > Use this file to discover all available pages before exploring further. # Introduction to Document OCR > Extract structured data from identity documents without creating a Trust Validation. ## Overview Document OCR endpoints extract structured fields from a supported identity document — a KTP, Kartu Keluarga, NPWP, driving licence, passport, or Philippine ID card — and return them in the same request/response cycle. Like the rest of the [Stateless Verification API](/stateless-verification-api/introduction), there is no `trustValidationId`, no session, and no separate finalize step. Extracted fields are returned to the caller and **never stored**. The transaction record keeps only the outcome and quality signals (outcome, status, confidence, missing-field count, time to verify, and — for the Philippines plan — the classified document type). Cropped image regions (`photo`, `signature`) are never returned or persisted, and are always `null`. ## Endpoints ### Indonesia | Verification type | Method | Path | | --------------------------------------------------------------------------------------------------------------------- | ------ | -------------------------------------------------------------------------------- | | [KTP Details Extraction](/stateless-verification-api/document-ocr/ktp-details-extraction) | POST | `/api/v3/verifications/indonesia/ktp-details-extraction-passthrough` | | [Kartu Keluarga Details Extraction](/stateless-verification-api/document-ocr/kartu-keluarga-details-extraction) | POST | `/api/v3/verifications/indonesia/kartu-keluarga-details-extraction-passthrough` | | [NPWP Details Extraction](/stateless-verification-api/document-ocr/npwp-details-extraction) | POST | `/api/v3/verifications/indonesia/npwp-details-extraction-passthrough` | | [Driving Licence Details Extraction](/stateless-verification-api/document-ocr/idn-driving-license-details-extraction) | POST | `/api/v3/verifications/indonesia/driving-license-details-extraction-passthrough` | | [Passport Details Extraction](/stateless-verification-api/document-ocr/idn-passport-details-extraction) | POST | `/api/v3/verifications/indonesia/passport-details-extraction-passthrough` | ### Philippines | Verification type | Method | Path | | --------------------------------------------------------------------------------------------------------------- | ------ | --------------------------------------------------------------------------- | | [Document Details Extraction](/stateless-verification-api/document-ocr/philippines-document-details-extraction) | POST | `/api/v3/verifications/philippines/document-details-extraction-passthrough` | ## Request All six endpoints take the same request shape: | Field | Type | Required | Notes | | --------------- | -------------- | ------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `documentImage` | file or string | Yes | Multipart file upload, or a base64 string (raw or `data:` URI). Accepts JPG, JPEG, PNG, and PDF, up to 2 MB. | | `metadata` | object | No | Free-form correlation data. Echoed back unchanged in the response. | | `documentTypes` | — | Not accepted | Rejected with a 400 if supplied. There is no template to bound the accepted list against on a stateless call — the response reports whichever document type it found. | ## Reading the response Every response — success or failure — includes: ``` success, message, verificationStatus, verificationStatusCode, transactionId, verificationType, metadata, createdAt ``` A successful response also includes `data`. A coded failure includes `code` instead — the two never appear together. ### Status codes | `verificationStatus` | `verificationStatusCode` | Meaning | | -------------------- | ------------------------ | ------------------------------------------------------------------- | | `VERIFIED` | 3 | Every required field was read from the document. | | `REVIEW_NEEDED` | 2 | Some required fields could not be read; check `data.missingFields`. | | `REJECTED` | 1 | The document was read, but is not the expected type. | | `FAILED` | 6 | The request could not be completed — see [Errors](#errors) below. | `REVIEW_NEEDED` and `REJECTED` are still returned as HTTP `200` — the call succeeded and the engine produced an answer, it's just not a clean extraction. Only `FAILED` corresponds to a non-2xx response. ### Per-field confidence Extracted values are returned as `{"value": ..., "confidence": "99.35%"}` whenever the value can be traced back to scored OCR text. Values that cannot be scored — `null` results, empty strings, and classification fields such as `documentType` — are returned bare, without a confidence wrapper. Confidence reflects transcription certainty, not correctness of the underlying document. ## Errors | Code | HTTP | Meaning | | --------------------- | ---- | -------------------------------------------------------------------------------------------------------- | | `VALIDATION_FAILED` | 400 | The request body or the submitted image failed validation. Fix the field named in `errors` and resubmit. | | `SOURCE_UNAVAILABLE` | 500 | The OCR provider could not be reached, timed out, or returned no result. Safe to retry. | | `INTERNAL_ERROR` | 500 | A fault on IDmeta's side. Safe to retry; include the `transactionId` if you escalate. | | `INSUFFICIENT_CREDIT` | 402 | Your account holds a plan for this verification type but has no credit remaining. | A `401` is returned for a missing or invalid bearer token, and a `403` if your account has no plan configured for the verification type at all. ## Billing | Result | Billed | | --------------------- | ------------------------------------------- | | `VERIFIED` (200) | Yes | | `REVIEW_NEEDED` (200) | Yes | | `REJECTED` (200) | Yes — the engine returned a complete answer | | Any 400 | No | | Any 500 | No | | 401 / 402 / 403 | No |