Slug: quality-gate Version: 1.0.0 Infer: Yes

Overview

Quality Gate Model

The Quality Gate model analyzes documents to detect quality issues that might interfere with downstream document processing. Currently, it specializes in detecting handwritten numeric values (such as invoice amounts, balances, etc.) that might require manual processing.

How It Works

The model converts the input document to grayscale to optimize handwriting detection
It uses AI to identify specific document types based on content and layout
For qualifying documents, it analyzes each page to detect handwritten numeric values
Documents with handwriting are labeled accordingly, allowing for special handling

Options Configuration

Option	Description
should_check_for_handwriting	Enables or disables handwriting detection (default: true)
included_document_types	Document types to check for handwriting (e.g., “invoice, receipt”)
excluded_document_types	Document types to exempt from handwriting checks (e.g., “bill of lading, weight slip”)
limited_to_first_page	When true, only checks the first page for handwriting (default: false)
handwriting_detector_model	The AI model used for handwriting detection (default: “gemini-2.0-flash-001”)

Process Flow

Example Usage

The Quality Gate model is especially useful for:

Automated document processing pipelines that require high-quality machine-readable content
Identifying documents that need manual review versus those that can be fully automated
Creating quality control steps before document extraction
Filtering invoices with handwritten amounts that might lead to extraction errors

Configuration Example

To check only invoices and receipts for handwriting:

should_check_for_handwriting: true
included_document_types: "invoice, receipt"
excluded_document_types: "packing slip, bill of lading"
limited_to_first_page: false
handwriting_detector_model: "gemini-2.0-flash-001"

Handwriting Detection Logic

The handwriting detection algorithm:

Focuses specifically on numeric values (amounts, quantities, totals)
Ignores dates, times, and non-numeric handwritten text
Verifies multiple handwritten numeric values or the same value in multiple places
Uses a two-pass verification process to reduce false positives

Inference Options

The following options can be configured when using this model for inference:

Name	Label	Type	Description	Default	Required
`should_check_for_handwriting`	Should Check For Handwriting	boolean	Whether to check for handwriting related to numeric data (e.g. invoice amounts, balances, or amounts paid	True	No
`included_document_types`	Included Document Types	string	The document types to check for handwriting (e.g. invoice, receipt, etc.)	-	Yes
`excluded_document_types`	Excluded Document Types	string	The document types to specifically exclude from handwriting checking (e.g. bill of lading, weight slip, etc.)	-	No
`limited_to_first_page`	Limit to First Page	boolean	Whether to limit the handwriting detection to just the first page	False	No
`handwriting_detector_model`	Handwriting Detector Model	string	The model to use to identify pages with handwritten billing information	gemini-2.0-flash-001	No

Model Details

Provider: Kodexa

Introduction

​Overview

​Quality Gate Model

​How It Works

​Options Configuration

​Process Flow

​Example Usage

​Configuration Example

​Handwriting Detection Logic

​Inference Options

​Model Details