amazon.com

•

API Key

About Amazon Textract

Amazon Textract is a machine learning service from AWS that automatically extracts printed text, handwriting, tables, forms, key-value pairs, and structured data from scanned documents and images. It goes beyond simple OCR with specialized APIs for expense documents, identity documents, and mortgage lending packages. Pricing is pay-per-page with no upfront commitments, and a free tier is available for new customers.

AI Agent Use Cases

- Detect Document Text API for OCR of printed text and handwriting

- Analyze Document API with Forms, Tables, Queries, and Signatures detection

- Analyze Expense API for automated invoice and receipt parsing

- Analyze ID API for identity document (passport, driver's license) data extraction

- Analyze Lending API for mortgage document classification and extraction

- Asynchronous processing for multi-page PDF documents

Available Actions

These are the specific actions that AI agents can perform with this tool

Analyze Document

5 inputs

This function analyzes a document to detect text, key-value pairs, tables, and other elements using AWS Textract.

Inputs

Document

The input document must be an image in JPEG, PNG, PDF, or TIFF format, either as base64-encoded bytes or an S3 object reference.

FeatureTypes

A list of the types of analysis to perform on the document.

AdaptersConfig

Specifies the adapter to be used when analyzing a document.

HumanLoopConfig

Sets the configuration for the human-in-the-loop workflow for analyzing documents.

QueriesConfig

Contains queries and aliases for those queries as determined by the input.

Analyze Expenses

1 input

Analyze expenses on a document to extract financial information such as items purchased and their prices.

Inputs

Document

The input document for analysis, provided as either a byte object or an S3 object reference.

Analyze Identity Documents

1 input

Analyze identity documents and extract relevant information using Amazon Textract.

Inputs

DocumentPages

The document pages being passed to AnalyzeID for analysis.

Document Text Detection

1 input

Detect the text present within a document using Amazon Textract.

Inputs

Document

The input document for text detection. It must be provided as either a base64-encoded byte array for JPEG or PNG, or an S3 object indicating the location of the document.

Get Document Analysis

3 inputs

Retrieve the results of an analysis job processed by Amazon Textract for text in a document.

Inputs

JobId

A unique identifier for the text-detection job. Returned from StartDocumentAnalysis.

MaxResults

The maximum number of results to return per paginated call. Defaults to 1000.

NextToken

Pagination token to retrieve the next set of blocks if the previous response was incomplete.

Get Document Text Detection

3 inputs

Retrieve the results of an asynchronous text detection operation for a document

Inputs

JobId

A unique identifier for the text detection job.

MaxResults

The maximum number of results to return per paginated call.

NextToken

Pagination token to retrieve the next set of results for an incomplete response.

Start Document Analysis

9 inputs

Start the analysis of a document for detecting relationships such as key-value pairs, tables, and selection elements.

Inputs

AdaptersConfig

Specifies the adapter to be used when analyzing a document.

ClientRequestToken

An idempotent token for identifying the start request to prevent duplicate jobs.

DocumentLocation

The location of the document in an Amazon S3 bucket.

FeatureTypes

The types of analysis to perform. Valid values include: TABLES, FORMS, QUERIES, SIGNATURES, LAYOUT.

JobTag

Identifier included in completion notification published to Amazon SNS topic.

KMSKeyId

KMS key used for encrypting inference results.

NotificationChannel

The Amazon SNS topic ARN for completion status notification.

OutputConfig

Configuration for saving the output to a customer-defined bucket.

QueriesConfig

Configuration for analyzing specific queries in the document.

Start Document Text Detection

6 inputs

Start asynchronous text detection in a document stored in an Amazon S3 bucket using AWS Textract.

Inputs

ClientRequestToken

Idempotent token to identify start request to prevent starting the same job accidentally more than once.

DocumentLocation

Location of the document in Amazon S3 to be processed.

JobTag

Identifier included in completion notification published to Amazon SNS topic.

KMSKeyId

KMS key for encrypting inference results. Can be Key ID or Key Alias format.

NotificationChannel

Amazon SNS topic ARN where Amazon Textract will publish completion status.

OutputConfig

Defines if the output will be saved to a customer-defined bucket. Defaults to internal storage.