Amazon Textract Block

Use Amazon Textract to detect text or analyze tables and forms in documents stored in Amazon S3. Authenticate with AWS access key credentials (SigV4 signed).

Overview

When to Use

• Extract raw lines and words of text from scanned PDFs or images stored in S3.
• Parse structured tables out of documents (invoices, reports, spreadsheets) to get key-value data.
• Extract form fields and their values from documents such as contracts, applications, or surveys.
• Feed extracted document content into downstream blocks for classification, summarization, or storage.
• Process multiple document types (PDF, PNG, JPEG, TIFF) from a centralized S3 bucket.
• Automate document ingestion pipelines that require structured data from unstructured files.

Configuration

Operation

Selects which Amazon Textract API call to make. Required.

Detect document text detects lines and words of text in the document (no structured analysis). It requires s3Bucket and s3Name.

Analyze document analyzes the document for structured features such as tables and forms. It requires s3Bucket, s3Name, and optionally featureTypes.

S3 Bucket

The name of the S3 bucket that contains the document. Visible for both operations.

• Type: short-input
• Placeholder: my-bucket
• Required for both operations.

S3 Object Name

The S3 object key (path) of the document within the bucket. Visible for both operations.

• Type: short-input
• Placeholder: path/to/document.pdf
• Required for both operations.

Feature Types

A JSON array of feature types to analyze. Only shown when Operation is textract_analyze_document.

• Type: long-input
• Placeholder: ["TABLES","FORMS"]
• Optional — defaults to ["TABLES","FORMS"] if omitted.
• Valid values: "TABLES" (extract table cells) and "FORMS" (extract key-value pairs from form fields).

AWS Region

The AWS region where the Textract service will be called (must match the region of your S3 bucket for cross-service access). Required.

• Type: short-input
• Placeholder: us-east-1

AWS Access Key ID

The AWS IAM access key ID used for SigV4 request signing. Stored as a password field (masked). Required.

AWS Secret Access Key

The AWS IAM secret access key used for SigV4 request signing. Stored as a password field (masked). Required.

Inputs & Outputs

Inputs:

• operation (string) — Operation to perform (textract_detect_document_text or textract_analyze_document).
• awsRegion (string) — AWS region (e.g. us-east-1).
• awsAccessKeyId (string) — AWS access key ID.
• awsSecretAccessKey (string) — AWS secret access key.
• s3Bucket (string) — S3 bucket name containing the document.
• s3Name (string) — S3 object key (name/path) of the document.
• featureTypes (json) — Feature types to analyze (only used with textract_analyze_document), e.g. ["TABLES","FORMS"].

Outputs:

• data (json) — The raw Textract API response, containing the Blocks array with detected text, lines, words, tables, cells, and key-value pairs depending on the operation used.

Tools

Textract Detect Document Text (textract_detect_document_text) — Calls the Textract.DetectDocumentText API to detect all lines and words of printed or handwritten text in a document stored in S3. Returns a Blocks structure with LINE and WORD block types.

Textract Analyze Document (textract_analyze_document) — Calls the Textract.AnalyzeDocument API to analyze a document in S3 for structured content. Extracts tables (as TABLE, CELL, COLUMN_HEADER blocks) and form fields (as KEY_VALUE_SET blocks) based on the requested featureTypes. Defaults to ["TABLES","FORMS"] if no feature types are specified.

YAML Example

textract_1:
  type: textract
  name: "Amazon Textract"
  inputs:
    operation: "textract_analyze_document"
    awsRegion: "{{AWS_REGION}}"
    awsAccessKeyId: "{{AWS_ACCESS_KEY_ID}}"
    awsSecretAccessKey: "{{AWS_SECRET_ACCESS_KEY}}"
    s3Bucket: "{{upload_trigger.bucket}}"
    s3Name: "{{upload_trigger.objectKey}}"
    featureTypes: '["TABLES","FORMS"]'
  connections:
    outgoing:
      - target: next-block-id