AI Diagram Assistant
The AI Diagram Assistant helps you create a system diagram by describing your architecture in natural language. It accelerates early threat modeling, reduces manual diagramming, and supports iteration as your understanding of the system evolves.
Use the AI Diagram Assistant when you want to:
- Create a first-pass diagram quickly
- Translate written architecture descriptions into a visual model
- Iterate on a design before committing it to a threat model canvas
What the AI Diagram Assistant does
The assistant:
- Converts a written description of your system into a structured diagram
- Identifies likely processes, data stores, external entities, and connections
- Generates a preview diagram you can refine before adding it to your threat model canvas
The output is a starting point, not a finished model. You should always review and adjust the diagram before generating threats.
How it works
Step 1: Describe your system
Start by explaining your system architecture in plain language. Focus on:
- Major components or services
- How data flows between them
- External users or third-party systems
- Key boundaries or trust assumptions
Example prompt:
A web application used by authenticated users. Traffic goes through a load balancer to a frontend service, which calls a backend API. The API reads and writes to a database. The system integrates with a third-party payment provider.
You do not need special syntax or formatting.
Step 2: Review the preview diagram
Devici generates a diagram preview based on your prompt.
In the preview, you can:
- Inspect the identified elements and connections
- Adjust layout and positioning for clarity
- Validate that components and data flows reflect reality
At this stage, the diagram is not yet part of your threat model canvas.
Step 3: Iterate on your prompt
You can refine the diagram by updating your prompt. Common iterations include:
- Adding missing components or services
- Clarifying data flows or protocols
- Introducing external systems
- Simplifying or restructuring the model
Each iteration updates the preview so you can see changes immediately.
Tip
Treat this like a conversation. Start simple, then add detail gradually as the diagram improves.
Tip
If you expect to reuse the same architecture description, consider saving your prompt as a preset. Presets allow you to quickly regenerate diagrams for similar systems or updated designs.
Step 4: Add attributes
Before adding the diagram to the canvas, you can generate and review attributes for the elements in the model.
To do this, select Add Attributes.
Devici will generate a set of suggested attributes based on the diagram structure. You can then review and adjust these attributes before continuing.
During this step, you can:
- Review attributes suggested for each element
- Add additional attributes where needed
- Remove attributes that do not apply
- Refine attribute selections to better match the system design
Tip
Reviewing attributes before adding the diagram helps ensure the generated threats accurately reflect the system.
Step 5: Add the diagram to the canvas
Once you are satisfied with the preview and attribute configuration:
- Add the diagram to the threat model canvas
- Place it anywhere on the canvas
- Continue editing using standard canvas tools
After adding it to the canvas, the diagram behaves like any other manually created model.
What to do after adding the diagram
After the diagram is on the canvas, continue with standard threat modeling steps:
- Add or refine trust boundaries and zones
- Apply attributes to elements and data flows
- Review generated threats and mitigations
- Adjust the structure as your understanding improves
The AI Diagram Assistant does not replace these steps. It accelerates getting to them.
Best practices
- Start with a high-level description, then iterate
- Don’t aim for perfect output on the first prompt
- Always review and validate generated diagrams
- Review attributes before adding the diagram to the canvas
- Use presets for architectures you expect to revisit