In the rapidly evolving landscape of enterprise software, organizations often face a dilemma: they want the productivity benefits of Generative AI, but they cannot compromise on data sovereignty, security compliance, or cost control. Visual Paradigm addresses this challenge through its Custom AI Endpoint Support architecture. This feature transforms the standard, cloud-dependent AI workflow into a flexible, organization-managed infrastructure.

This tutorial explores the technical architecture behind Custom AI Endpoints, explaining how workspace managers can configure a secure gateway that routes AI requests to specific providers or self-hosted models. We will break down the configuration parameters, the supported features, and the strategic advantages of this architecture.

1. The System Architecture: Decoupling the Workspace from the Provider

Traditionally, AI-powered applications are hard-coded to connect to a specific vendor’s API. Visual Paradigm’s architecture, however, introduces a configurable middleware layer known as the Custom AI Endpoint.

Think of this architecture as a “Smart Router” for your AI traffic. Instead of the Visual Paradigm application attempting to connect directly to a public AI API, it first consults the Workspace Configuration. If a Custom Endpoint is enabled, the request is redirected through the organization’s approved channel.

The flow operates as follows:

1. The Trigger: A user initiates an AI-supported action (e.g., generating a diagram or asking a chatbot question) within the Visual Paradigm interface.
2. The Routing Check: The system checks the Workspace settings. If “Routing” is enabled, the application retrieves the configured Endpoint URL, Model ID, and API Key.
3. The Request: The application sends the prompt to the specified Custom Provider or self-hosted service.
4. The Response: The custom service processes the request and returns the output to the Visual Paradigm client.

This architecture ensures that the actual AI processing—and the associated data—never leaves the organization’s preferred infrastructure stack.

Key Configuration Parameters

When configuring a Custom AI Endpoint, the system accepts several specific parameters that define the connection’s behavior. Understanding these fields is crucial for IT administrators:

• Provider: The entity hosting the AI service (e.g., a public cloud provider like Azure or AWS, or an internal private cloud).
• Endpoint URL: The specific API address (e.g., https://api.your-endpoint.com) where requests are sent. This allows for self-hosting or private cloud instances.
• API Key: The credential used to authenticate the request. Visual Paradigm stores this securely within the workspace, ensuring it is never shared with third parties.
• Model ID: The specific identifier for the AI model (e.g., vp-model-001). This allows the organization to switch between different model versions (e.g., switching from a standard model to a reasoning-heavy model) without changing the infrastructure.
• Routing: A binary toggle (Enabled/Disabled). When Enabled, all supported AI requests in the workspace are forced through this configuration. When Disabled, the system falls back to Visual Paradigm’s default model.

2. Supported Features and Capabilities

Custom AI Endpoints are not just for general chat; they are integrated deeply into the modeling lifecycle. The architecture supports routing for specific high-value features across different Visual Paradigm applications.

Application-Specific Routing

The following features can be configured to use the custom endpoint:

• Visual Paradigm AI Diagramming: Supports chatbot interactions and AI-assisted diagramming workflows. This allows users to generate diagrams via natural language, with the generation logic running on the organization’s approved model.
• VPasCode: Handles complex code-related tasks, including diagram generation, diagram modification, code error fixing, and code translation. This is critical for developers who need to ensure their proprietary code logic is not sent to public AI servers.
• Visual Paradigm Vpp Chatbot: AI-assisted interactions specifically within the Vpp Chatbot environment.
• Visual Paradigm Desktop AI: Supports AI-generated diagrams directly within the desktop application, ensuring offline or local-first workflows remain secure.

Advanced Settings: Token Limits and Reasoning

Beyond simple connectivity, the architecture allows for performance tuning. Workspace managers can configure:

1. Maximum Output-Token Limit: Restrict the length of the AI’s response to control costs or fit specific output formats.
2. Reasoning/Thinking Settings: Where supported by the custom model, administrators can toggle reasoning capabilities to enhance the complexity of the AI’s output.

3. Strategic Benefits: Why Use a Custom Endpoint?

The decision to implement a Custom AI Endpoint is driven by four primary architectural and business requirements:

A. Security and Governance

By routing requests through an approved provider or self-hosted instance, organizations ensure that sensitive modeling data and intellectual property are governed by internal compliance policies. This eliminates the risk of data leakage to public AI training sets.

B. Infrastructure Control

IT teams gain full control over provider relationships. They can manage API keys, quotas, and rate limits centrally. This setup allows for the integration of AI services into the organization’s existing Identity and Access Management (IAM) and monitoring systems.

C. Model Flexibility and Performance

Not all models are created equal. A custom endpoint allows the organization to select a model based on specific performance metrics:

• Latency: Choosing a provider geographically closer to the workspace for faster response times.
• Capability: Selecting models with superior reasoning or language support.
• Cost: Switching to more cost-effective models for high-volume, low-complexity tasks.

D. Operational Consistency

The configuration is applied at the Workspace Level. This means the workspace manager establishes a “single source of truth.” Users do not need to configure their own API keys; they simply use the tools, and the system automatically routes traffic through the secure, pre-approved channel.

4. Implementation Requirements

To leverage this architecture, specific prerequisites must be met:

• Licensing: The workspace requires an Online Deluxe or Desktop Professional license.
• Permissions: Only the Workspace Manager has the authority to configure endpoints. Regular users cannot override these settings.
• Graceful Fallback: If the custom routing is disabled or the endpoint is unreachable, the system is designed to fall back to the system-default model (depending on the specific routing configuration), ensuring workflow continuity.

Conclusion

Visual Paradigm’s Custom AI Endpoint Support represents a significant step forward in enterprise-grade AI integration. By decoupling the application interface from the underlying AI infrastructure, it empowers organizations to harness the power of Generative AI while maintaining strict control over security, cost, and performance. Whether connecting to a massive cloud provider or a private, self-hosted LLM, the architecture provides the flexibility needed for modern, secure development environments.