The Extract Result object

Extract is asynchronous: POST /v1/extract returns a request, and this object is the result on that request once it completes (and the data returned directly on a ?wait call that finishes in time).

{
  "result": "extracted",
  "document": { "media_type": "application/pdf", "pages": 1 },
  "template": "sch_8f2c...",
  "fields": {
    "invoice_number": "INV-1042",
    "total": 1250.0
  },
  "missing": [],
  "extracted_at": "2026-08-18T12:00:00Z"
}

An unreadable or encrypted document:

{
  "result": "unreadable",
  "document": { "media_type": "application/pdf", "pages": 1 }
}

When you send template: "auto", we pick the schema type for you and tell you which one we picked:

{
  "result": "extracted",
  "document": { "media_type": "application/pdf", "pages": 2 },
  "template": "sch_8f2c...",
  "match": {
    "name": "Bill of lading",
    "confidence": "high",
    "alternatives": [
      { "template": "sch_41ab...", "name": "Arrival notice", "confidence": "low" }
    ]
  },
  "fields": { "container_number": "MSCU1234567" },
  "missing": [],
  "extracted_at": "2026-10-09T12:00:00Z"
}

If nothing in your workspace fits the document, you get unmatched instead. We do not guess, and we do not fill:

{
  "result": "unmatched",
  "document": { "media_type": "application/pdf", "pages": 1 },
  "match": {
    "confidence": "high",
    "alternatives": [
      { "template": "sch_41ab...", "name": "Arrival notice", "confidence": "low" }
    ]
  }
}

Attributes

enum

required

Outcome of the read. One of:

extracted — the fields are in fields.
unreadable — a legal file we will not stand behind. Not billed.
unmatched — template: "auto" only. The document was readable, but no schema type in your workspace fits it. Nothing was filled.

object

required

The document that was read.

Show properties

string

The declared media type — application/pdf, image/jpeg, image/png, or image/webp.

integer

Page count. Images are 1. This is the unit billing is charged on: 4 credits a page to extract, plus 1 credit a page when you use auto.

string

The schema type id (sch_…) the call pointed at, when one was used — or the one we matched, if you sent auto. Omitted for a fully inline schema, and when result is unmatched.

object

How we chose the schema type. Present only when you sent template: "auto".

Show properties

string

Name of the schema type we matched. Omitted when result is unmatched.

enum

How sure we are: high, medium, or low.What it is sure about depends on result. On extracted it is our certainty in the schema type we picked. On unmatched it is our certainty that nothing in your workspace fits — so unmatched with high confidence is the normal, clearest case, not a contradiction.Because of that, route on result, not on confidence. Filtering for confidence: "low" to build a review queue would miss every confidently unmatched document.This is a level, not a score. Treat high as safe to process, medium as worth a glance, and low or unmatched as work for a person. We publish a level rather than a number on purpose — a number invites a threshold we would then have to keep stable forever, and the level is the honest resolution of the judgement behind it.

object[]

The schema types we considered and rejected, best first, each with its own template, name, and confidence. When a match is wrong or unmatched, this is what you route a review queue on.

object

The extracted values, shaped by your schema. Present only when result is extracted. A field that wasn’t found is omitted here (and listed in missing), never guessed.

string[]

Declared fields the model could not fill from the document. Present only when result is extracted.

timestamp

When the read completed. Present only when result is extracted.

See the Extract guide for schemas and examples, or Requests for the wait/poll/webhook lifecycle this result is delivered through.

Core

Conversations

Configuration

Delivery

MCP

Workbench

The Extract Result object

Attributes

​Attributes

Attributes