OCR

1. Overview

The OCR component (ocr) extracts text from an image using optical character recognition (OCR). It's useful for reading scanned documents, receipt photos, screenshots with text, and more.

Use it when the flow needs to turn an image into searchable text. Not suited for PDFs with native text (in that case, the PDF Reader component is more appropriate, since it reads the text directly without needing OCR).

2. Prerequisites

No external credentials — text recognition runs locally (Tesseract engine), with no calls to external services.

3. Authentication and Connection

Not applicable.

4. Configuration / Supported Operations

The component has a single operation: extract text from an image.

Field	Required	Type	Description	Example
image	Yes	Text	Name of the message property where the image is.	"receipt"
format	No	Text: text or markdown	Format of the extracted text. markdown adds Markdown-compatible line breaks. Default value: "text".	"markdown"
language	No	Text	Language code of the text in the image. Default value: "eng" (English).	"por" (Portuguese)
dpi	No	Number	Resolution (DPI) provided to the OCR engine. Default value: 300.	300
pageSegMode	No	Number	OCR engine's page segmentation mode (controls how text is identified in the image layout). Default value: 6.	6
blockSize	No	Number	Technical parameter for image pre-processing (binarization). Default value: 15.	15
thresholdOffset	No	Number	Technical parameter for image pre-processing (binarization). Default value: 10.	10

⚠️ TO CONFIRM: pageSegMode values follow Tesseract's standard numbering (e.g., 6 = "uniform block of text"), but the code doesn't document the complete table of possible values.

5. Practical Examples

Simple example: extract the text from a payment receipt image, received in Portuguese.

Input (image already attached by a previous step, e.g. an upload):

{
  "receipt": "<image file, made available by a previous step>"
}

Component configuration:

{
  "componentName": "ocr",
  "configurations": {
    "image": "receipt",
    "language": "por"
  }
}

Response:

{
  "text": "PAYMENT RECEIPT\nAmount: R$ 150.00\nDate: 07/20/2026",
  "format": "text",
  "language": "por",
  "length": 52
}

Advanced example: extract text from an image already targeting Markdown format (to display preserving line breaks), adjusting pre-processing parameters for a low-quality image.

Component configuration:

{
  "componentName": "ocr",
  "configurations": {
    "image": "documentPhoto",
    "language": "por",
    "format": "markdown",
    "blockSize": "25",
    "thresholdOffset": "15"
  }
}

Response:

{
  "text": "ID DOCUMENT  \nName: Maria Silva  \nTax ID: 123.456.789-00",
  "format": "markdown",
  "language": "por",
  "length": 59
}

6. Common Errors and Troubleshooting

Error / Symptom	Likely Cause	How to Fix
"An image file is required but has not been configured."	The image field wasn't filled in.	Fill in image with the name of the property that contains the image.
"File [name] not found inside the execution context."	The referenced image isn't available in the message under that name.	Confirm a previous step made the image available under that exact name.
"Unsupported or corrupted image format"	The submitted file isn't a valid image, or is corrupted.	Check whether the file is really an image in a supported format (e.g., JPG, PNG).
"OCR processing failed: ..."	Internal OCR engine failure processing the image.	Try with a higher quality/resolution image, or review the pre-processing parameters.
Extracted text has a lot of errors/noise	Low-quality image, incorrectly configured language, or pre-processing parameters unsuited to the image type.	Adjust language to the text's correct language and test different blockSize/thresholdOffset values.