View a markdown version of this page

Limits and quotas - Amazon Textract

Limits and quotas

Amazon Textract Insights has two kinds of limits:

  • Quotas are per-account, per-Region defaults. You can request an increase through Service Quotas.

  • Limits are fixed by the service and cannot be raised. A request that exceeds a request limit, such as a name length or the number of blueprints, is rejected. A document that exceeds a page limit is accepted, and the job then fails (see Troubleshooting).

Quotas (adjustable)

The following quotas are per-account, per-Region defaults. You can request an increase for a quota through Service Quotas.

Adjustable quotas

Quota

Default

StartDocumentInsights requests per second

10

GetDocumentInsights requests per second

10

Concurrent extraction jobs

20

Concurrent generative jobs

20

Concurrent custom jobs

20

Figures captioned per page

5

Projects per account

100

Blueprints per account

350

Versions per blueprint

10

A job counts against its concurrent-job quota until it reaches a terminal status (SUCCEEDED or FAILED).

Limits (fixed)

The following limits are fixed by the service and cannot be raised.

Documents

The following limits apply to each input document.

Document limits

Limit

Value

Supported file types

PDF, PNG, JPEG, TIFF. (DOC and DOCX are not supported.)

Maximum file size

500 MB

Pages per document insights job

3,000

Pages per document for custom extraction without splitting (SplitterConfiguration not enabled)

20

Pages per document with figure captioning enabled (GenerativeConfiguration.FigureCaption.State is ENABLED)

1,500

When a document exceeds a page limit

Page limits are checked after the job starts, so StartDocumentInsights still returns a job ARN. The job then ends with Status FAILED, and GetDocumentInsights reports the cause in ErrorType:

  • More pages than the Pages per document limit: PageLimitExceeded, with ErrorMessage "The number of pages in the input document is too high."

  • Custom extraction without splitting on a file over its page limit: DocumentTooLargeException, with ErrorMessage "The input document size is too large to process."

  • Figure captioning enabled on a file over its page limit: DocumentTooLargeException, with the same ErrorMessage.

The two DocumentTooLargeException cases return the same message. To tell them apart, compare the file's page count with each limit and check which features the job had enabled.

Document summarization does not have its own page limit. Only the overall Pages per document limit applies.

To run custom extraction on a file longer than 20 pages, enable splitting so the file is divided into individual documents before blueprints are applied. For more information, see Splitting multi-document files.

Projects and blueprints

The following limits apply to projects and blueprints.

Project and blueprint limits

Limit

Value

Blueprints per project

40

Blueprints per job when passed inline

10

FallbackBlueprints entries per project

1

Blueprint schema size

100,000 characters

Leaf fields per blueprint

100

List leaf fields per blueprint

30

Field name and custom type name

1–60 characters; can't contain /

Field description

Up to 600 characters

ProjectName, BlueprintName

1–128 characters, [a-zA-Z0-9-_]+

ProjectDescription

Up to 300 characters

For blueprint structure rules (nesting, lists, enum), see Blueprint schema best practice.

Requests and tags

The following limits apply to request tokens and resource tags.

Request and tag limits

Limit

Value

ClientRequestToken

1–64 characters, [a-zA-Z0-9-_]+

Tags per resource

50

Tag key

1–128 characters; can't begin with aws:

Tag value

0–256 characters

Requesting a quota increase

Request an increase in Service Quotas for the specific quota and AWS Region you need. Limits in the "Limits (fixed)" section cannot be raised.