PB Layout Editor AI – System Architecture
Table of Contents
PB Layout Editor AI – System Architecture
1. Overview
PB Layout Editor AI is an AI-assisted capability for analyzing PDF documents and extracting their layout structure. The cloud service receives source documents from PBDesigner, analyzes the document layout, and returns structured layout information that PBDesigner can use to create production templates automatically.
The core cloud component is the PDF Layout Extractor, a SaaS service hosted on AWS. PBDesigner, which is installed in the customer environment, communicates with this service over a secured HTTPS channel. The returned layout description helps PBDesigner identify visual and structural document elements such as text blocks, tables, and other content areas.
2. Scope and Deployment Boundaries
The architecture separates the customer-side design tool from the provider-managed cloud analysis service.
| Area | Location | Responsibility |
|---|---|---|
| PBDesigner | Customer environment | Used by the implementer to design, edit, manage, and maintain production templates. |
| PDF Layout Extractor | Provider-managed SaaS environment on AWS | Receives documents for layout analysis and returns structured layout information. |
| Third-party AI processing services | Provider-controlled cloud integration | Used only by the provider-managed service. They are not installed in the customer environment. |
3. Architecture Diagram
The architecture diagram shows PBDesigner connecting to the provider-managed AWS SaaS environment. Inside AWS, traffic passes through the front-door and protection layers, then reaches the Kubernetes-based service layer where the PDF Layout Extractor container and Keycloak identity component are located. The service integrates with Microsoft Azure Document Intelligence in the Europe region for OCR and document-analysis capabilities.
3.1 High-level flow
- PBDesigner sends a PDF document to the PDF Layout Extractor using secured REST APIs over HTTPS.
- The request enters the provider-managed AWS environment through the exposed service layer and security controls.
- The service layer runs on Amazon EKS and uses Ingress NGINX, AWS WAF, and Cloudflare as part of the access and protection path.
- Inside the Kubernetes cluster, an internal shield/proxy layer routes the request to the relevant components.
- Keycloak handles identity and access-management functions.
- PDF Layout Extractor performs orchestration and layout-processing logic.
- The extractor calls Azure Document Intelligence for OCR, document analysis, and structured content extraction.
- The resulting structured layout is returned to PBDesigner and used for template creation.
4. Processing Workflow
The end-to-end process is designed to keep the implementation work inside PBDesigner while using AI services only for analysis of the source PDF structure.
| Step | Description | Output |
|---|---|---|
| 1 | The implementer selects or sends a source PDF document from PBDesigner. | PDF document submitted for analysis. |
| 2 | PBDesigner communicates with the cloud service through secured REST/HTTPS APIs. | Authenticated and encrypted request. |
| 3 | The PDF Layout Extractor receives the request and coordinates the processing flow. | Processing job inside the provider-managed SaaS environment. |
| 4 | Azure Document Intelligence is used for OCR, document analysis, and extraction of structured information. | Detected text, tables, layout objects, and document structure. |
| 5 | The extractor converts the analysis result into a structured layout description. | Layout structure returned to PBDesigner. |
| 6 | PBDesigner displays the detected content as editable graphical objects. | Objects such as text blocks, tables, position, size, style, and structure. |
| 7 | The implementer uses PBDesigner to complete template design and maintenance. | Production template, including FMT output definition. |
5. Component Breakdown
5.1 PBDesigner
PBDesigner is installed and operated in the customer environment. It provides a visual workspace for planning, editing, and managing production templates for PDF documents. In the AI-assisted workflow, PBDesigner uses the PDF Layout Extractor to analyze source PDFs and then displays the returned document elements as graphical objects.
Typical objects returned and displayed in PBDesigner include:
- Text blocks and labels
- Tables and tabular areas
- Additional content elements
- Object position and size
- Formatting and structural properties
5.2 PDF Layout Extractor
PDF Layout Extractor is the central service layer. It provides secured APIs for PBDesigner and other approved system integrations, receives analysis requests, coordinates the required processing, and returns the structured layout result.
The service runs as a container in the provider-managed AWS environment and is shown in the diagram as part of the Kubernetes cluster on Amazon EKS.
5.3 Keycloak
Keycloak provides identity and access-management capabilities. The source architecture describes support for modern authentication and authorization standards, including OpenID Connect (OIDC) and OAuth 2.0, with support for SSO and integration with enterprise identity systems.
5.4 Azure Document Intelligence
Azure Document Intelligence is used for OCR, document analysis, and extraction of structured information. The architecture document describes this integration as operating through secure TLS/HTTPS communication between AWS and Azure.
5.5 AWS Secrets Manager
The system authentication details and permissions used for Azure integration are managed through a secure secrets-management mechanism. The source architecture identifies AWS Secrets Manager as the place where these credentials and permissions are stored and managed according to organizational identity and secret-management policies.
6. Security and Data Handling
| Security Area | Architecture Description |
|---|---|
| Transport security | Communication between PBDesigner, the AWS service, and Azure services is performed over secured HTTPS/TLS channels. |
| Public exposure reduction | The architecture describes usage of private endpoint/private channel mechanisms so the services are not exposed directly to the public internet and are reachable only from authorized networks and addresses. |
| Temporary storage | During processing, documents are stored temporarily only in the Microsoft Azure service environment. |
| Region | The Azure processing region described in the source document is Europe West. |
| Encryption at rest and during processing | The source document states that data is encrypted during storage and processing according to Azure security mechanisms. |
| Post-processing deletion | After processing is completed, documents are deleted according to the service policy described in the source document. |
| AI model training | The source document states that Microsoft AI services are used only for document processing and that Microsoft does not use the transferred content to train, improve, or customize its AI models. |
| Secret management | Azure credentials and permissions are stored and managed through AWS Secrets Manager. |
7. Template Generation in PBDesigner
The purpose of the AI-assisted architecture is not to replace PBDesigner. Instead, the PDF Layout Extractor provides PBDesigner with a structured understanding of the source PDF. PBDesigner then uses this information to accelerate the creation of production templates, including FMT definitions.
After the layout is returned, PBDesigner can present document components as editable visual objects. This allows the implementer to adjust layout mapping, verify detected elements, and complete the production template faster than with a fully manual setup.
8. PB Digital Context
PB Digital is a document communication management solution for creating, managing, producing, and distributing digital documents. It supports personalized and multi-channel communication with customers, dynamic document production, template management, integration with organizational data sources, digital distribution, and complex regulatory or business workflows.
PB Digital is part of the MAGNUM solution suite, which provides a broader digital customer-experience infrastructure. The suite includes complementary capabilities such as digital forms, digital signatures, API integration, co-browsing, digital document accessibility, and related workflow capabilities.
9. Quick Reference
| Question | Answer |
|---|---|
| Where is PBDesigner installed? | In the customer environment. |
| Where is the PDF Layout Extractor hosted? | In a provider-managed SaaS environment on AWS. |
| How does PBDesigner communicate with the service? | Through secured REST APIs over HTTPS. |
| What does the cloud service return? | A structured description of the document layout. |
| What is the returned layout used for? | Automatic and accelerated creation of production templates, including FMT definitions, in PBDesigner. |
| Which AI/OCR service is used? | Microsoft Azure Document Intelligence. |
| Which Azure region is mentioned? | Europe West. |
| What identity component is shown? | Keycloak. |