Azure Document Intelligence
The Azure Document Intelligence component connects Robility to Microsoft Azure AI Document Intelligence to analyze documents and extract text and structured information.
The component uses the Prebuilt Layout model to extract text, tables, selection marks, and document structure from supported file formats. The extracted information can be used in downstream automation workflows.
Note: Available extraction capabilities depend on the input file format and Azure service tier.
Prerequisites
Before using the component, ensure you have:
a. An Azure subscription and an Azure Document Intelligence resource.
b. The Azure Endpoint URL and API Key for the resource.
c. A supported input file accessible to the execution environment.
Extraction Capabilities
The Prebuilt Layout model can extract text, tables, document structure, and selection marks from structured and unstructured documents.
Use Fields to Extract to specify the required field names. Only successfully extracted fields that meet the configured confidence threshold are included in the output.
Document and Image Requirements
1. Document quality: The component can process structured, unstructured, and poorly formatted documents. However, unclear content may affect extraction results. Only fields with a confidence score of 0.80 or higher are included in the output.
2. Image dimensions: Ensure images meet Azure’s supported dimension requirements for the applicable file format and API version.
3. File restrictions: Ensure the input file uses a supported format and complies with the applicable file-size and page limits.
4. Page limits: The number of pages processed depends on the Azure service tier, file format, and component restrictions.
For supported formats, image requirements, and processing limits, refer to the Azure Document Intelligence model overview.
Parameters
| Parameter | Description |
|---|---|
| Azure Endpoint URL* | Specifies the endpoint URL of the Azure Document Intelligence resource. |
| Azure Key* | Specifies the API key associated with the Azure resource. |
| Upload File | Specifies a file to upload from the local machine. This parameter is ignored when File Path is provided. |
| File Path | Specifies the path of a file available in Flow storage. When both File Path and Upload File are provided, File Path takes priority. |
| Fields to Extract | Specifies the field names to extract as a comma-separated list. Leave empty to request all key-value pairs supported by the component and available from the extraction results. |
| Timeout (Seconds) | Specifies how long the component allows the operation to run before a timeout error occurs. Default: 30 seconds. |
| Retry Count | Specifies the number of retry attempts configured for the operation. Default: 1. |
| Delay Between Retries (Seconds) | Specifies the waiting time between retry attempts. Default: 1 second. |
| Delay Before Execution (Seconds) | Specifies the waiting time before the operation starts. Default: 0 seconds. |
| Delay After Execution (Seconds) | Specifies the waiting time after the operation completes. Default: 0 seconds. |
| Continue on Error |
Defines the workflow behavior when component execution fails.
|
Outputs
| Output | Description |
|---|---|
| Message | Returns a text summary of the extracted fields. |
| JSON Data | Returns the extracted information as structured JSON for downstream processing. |
| DataFrame | Returns the extracted information in tabular form for further analysis or processing. |
Confidence Scores
The component evaluates confidence scores at the individual field level. Only fields with a confidence score of 0.80 or higher are included in the output.
a. Score of 0.80 or higher: The field is included in the output.
b. Score below 0.80: The field is omitted from the output.
Confidence-score behavior depends on the input file format. If an expected field is missing from the output, it may not have been extracted or may have fallen below the confidence threshold. A missing field does not necessarily mean the information is absent from the source document.
Processing Time and Timeout
The Timeout (Seconds) parameter specifies how long the component allows the operation to run. The default is 30 seconds, and the timeout can be increased as required.
a. If a timeout occurs, verify the configured duration and input file.
b. Configure retry settings as required by the workflow.
A timeout indicates that processing did not finish within the configured duration. It does not necessarily mean the file format is unsupported.
Retrieve Azure Endpoint URL and API Key
Before configuring the component, retrieve the Azure Document Intelligence resource’s Endpoint URL and API Key from the Azure Portal.
For step-by-step instructions, see Document Intelligence – Azure Setup and Configuration.
Once you have the credentials, enter the endpoint URL in Azure Endpoint URL and the API key in Azure Key.
Security Note: Treat API keys as sensitive credentials. Store them securely and do not expose them in public documentation, screenshots, source code, or logs.
Troubleshooting
| Issue | Possible Cause and Resolution |
|---|---|
| Authentication Failure | Verify the Azure Endpoint URL and API Key. |
| Unsupported File Format | Check the format against the supported formats for the selected Azure service tier and API version. |
| File Exceeds Size Limits | Ensure the file is within both the component's 50 MB limit and Azure's applicable limits. |
| Expected Fields Are Missing | Check the document content, the Fields to Extract parameter, and the confidence threshold. |
| Incomplete or Inaccurate Extraction | Verify document readability and ensure the input meets Azure's requirements. |
| Timeout Error | Processing exceeded the configured timeout. Increase the timeout if required and verify the input file. |
