Powered by Zoomin Software. For more details please contactZoomin

Semaphore 5.13 Knowledge Model Management (KMM) How-To Articles

Creating a Model from a Specification

  • Last Updated: November 20, 2026
  • 5 minute read
    • Semaphore
    • Version 5.13
    • Documentation

Starting with Semaphore 5.13.0, you can use the Model from Spec AI workflow to create a knowledge model from a natural-language specification and optional supporting documents.

The workflow uses a configured Large Language Model (LLM) to generate model content based on your requirements.

You can create a model directly in its master version or generate content within a task so that you can review the results before incorporating them into the master model.

Before you begin

Before creating a model from a specification:

  • Ensure that an administrator has configured the AI Model Builder in Semaphore Studio. See Configure the AI Model Builder.
  • Ensure that you have permission to create knowledge models in KMM.
  • Prepare a description of the model you want to create, including its purpose, subject area, and expected content.
  • Optionally, prepare supporting documents containing terminology, requirements, definitions, or other relevant information.

Note: Use Create a new model with a task if you want to review AI-generated content before incorporating it into the master model.

Create a model from a specification

  1. Open Semaphore Knowledge Model Management (KMM).

  2. On the model list page, select New from Spec.

  3. Enter a prompt describing the knowledge model you want to create.

    Include information such as:

    • The purpose and subject area of the model.
    • The types of concepts you want to include.
    • The expected hierarchy and relationships.
    • Any terminology, definitions, or modeling requirements.
    • Any content that should be excluded.
  4. Optionally, upload supporting documents to provide additional context for model generation.

  5. Choose how you want to create the model:

    • Create a new model: Creates a new model and applies generated content directly to its master version.
    • Create a new model with a task: Creates a new model using a task-based workflow, allowing you to review generated content before incorporating it into the master model.
  6. If you want to generate only the model structure, select Structure only. This option generates concept schemes, classes, relationship types, and metadata types without generating the concept taxonomy.

  7. Start the workflow.

  8. Monitor the execution progress while Semaphore generates the model content.

  9. When the workflow finishes, open the generated model.

  10. Review the generated structure, concepts, labels, and relationships against your specification.

  11. If you selected Create a new model with a task, review the generated changes in the task before incorporating them into the master model.

Important: AI-generated content may require corrections. Review the generated model before using it in production.

Choose a model creation option

Model from Spec provides two ways to create a model.

Option When to use it
Create a new model When you want generated content to be added directly to the new model's master version.
Create a new model with a task When you want to inspect and refine generated content before committing it to the master model.

If you are generating a model from unfamiliar source material or expect to make substantial changes, consider using a task-based workflow.

Generate the structure only

Select Structure only when you want to generate the initial model structure without creating the concept taxonomy.

This can be useful when you want to establish the model's organization before adding concepts manually or through subsequent enrichment operations.

Supported source documents

You can upload supporting documents to help the AI workflow interpret your specification.

The following file formats are supported:

Category File formats
Text and documents .txt, .md, .pdf, .docx
Semantic web and RDF .ttl, .nt, .n3, .trig, .nq, .rdf, .jsonld

Upload limits:

  • Maximum of five files.
  • Maximum file size of 4 MB per file.

Use documents that contain information relevant to the model you want to create. For example, you might provide a terminology list, a requirements document, or a domain-specific specification.

Supporting documents provide additional context, but they do not guarantee that all their contents will be represented accurately in the generated model.

Example: Create a renewable energy knowledge model

Suppose you want to create a knowledge model describing renewable energy technologies.

You can use the following prompt:

Create a knowledge model for renewable energy technologies.

Include the following main subject areas:

  • Solar energy
  • Wind energy
  • Hydropower
  • Geothermal energy

Organize the concepts into a meaningful hierarchy.

For each subject area, include relevant concepts and identify appropriate relationships between them.

Use consistent terminology and avoid duplicate concepts.

Generate clear, concise definitions where appropriate.

To create the model:

  1. On the KMM models page, select New from Spec.
  2. Enter the prompt.
  3. Optionally, upload supporting documents containing technical terminology or definitions.
  4. Select Create a new model with a task.
  5. Start the workflow.
  6. Open the generated model when the operation finishes.
  7. Review the generated concepts, hierarchy, labels, and relationships.
  8. Correct any inaccurate or unnecessary content.
  9. When the content meets your requirements, incorporate the reviewed task changes into the master model.

The generated content depends on the prompt, supporting documents, and configured LLM provider. Results may vary.

Review the generated model

After the workflow finishes, inspect the generated content in Knowledge Model Management.

Check the following:

Area What to review
Model structure Does the model reflect the intended subject area and organization?
Concepts Are the concepts relevant, correctly named, and sufficiently specific?
Labels Are preferred and alternative labels accurate and consistent?
Relationships Do hierarchical and associative relationships reflect the intended meaning?
Metadata Are generated definitions and other metadata values accurate?
Completeness Does the model cover the requirements in your prompt and supporting documents?

Pay particular attention to duplicate concepts and relationships that may not reflect the intended hierarchy.

If you created the model with a task, review and correct the generated content before committing the changes to the master model.

Troubleshooting

The generated model does not match the specification

The quality of generated content depends on the configured LLM provider, the clarity of your prompt, and the supporting information you provide.

If the result does not meet your requirements:

  1. Compare the generated model with your original specification.
  2. Identify missing, inaccurate, or unnecessary content.
  3. Refine the prompt to make your requirements more specific if you decide to run the workflow again.
  4. Review the generated content and correct it as necessary.

A more detailed prompt can help guide the generation process, but it does not guarantee accurate results.

The workflow reports an error

An execution error may occur after some changes have already been applied.

If the workflow fails:

  1. Check whether a new model or task was created.
  2. Open the model, if available, and inspect its content.
  3. Review the model history for changes that may have been applied before the error.
  4. Avoid repeating the operation until you understand what was created.
  5. Contact your Semaphore administrator if the error persists.

Some LLM providers impose request or rate limits that may interrupt execution.

Alert