Schema Copilot

Build ontology schemas through natural language conversation with AI.

Schema Copilot

The Schema Copilot is a chat-based interface in the Graphora web UI that helps you create and refine ontology schemas through natural language conversation. Instead of writing YAML from scratch, describe your domain in plain language and let AI generate the schema for you.

How It Works

1. Open the AI Copilot page

   Navigate to the AI Copilot page in the Graphora dashboard.

2. Describe your domain

   Tell the copilot what you want to model. For example:

   “I want to extract information about companies, their products, and the people who work there. Companies have names and industries. People have names, titles, and email addresses. Products have names and descriptions.”

   The copilot uses a freeflow chat interface with streaming responses, so you see the schema being built in real time.

3. Review the generated schema

   The copilot generates a complete ontology YAML with entities, properties, relationships, and quality rules. Review it in the editor.

   version: 1
   entities:
     Company:
       properties:
         name:
           type: string
           required: true
           unique: true
         industry:
           type: string
       relationships:
         HAS_EMPLOYEE:
           target: Person
           cardinality: one_to_many
         HAS_PRODUCT:
           target: Product
           cardinality: one_to_many
     Person:
       properties:
         name:
           type: string
           required: true
         title:
           type: string
         email:
           type: string
       relationships:
         WORKS_FOR:
           target: Company
     Product:
       properties:
         name:
           type: string
           required: true
           unique: true
         description:
           type: string
   
4. Refine through conversation

   Continue the conversation to adjust the schema:

   “Add a foundedYear property to Company as an integer. Also add a COMPETES_WITH relationship between companies.”

   The copilot updates the schema incrementally based on your feedback.

5. Save and use

   Once you are satisfied, save the ontology. It is immediately available for document transformations.

Tips for Better Results

• Be specific about property types. Instead of “companies have revenue”, say “companies have revenue as an integer representing USD.”
• Mention cardinality. For example: “A company can have many products, but a product belongs to one company.”
• Include constraints. Say things like “company name should be unique and required.”
• Iterate. Start with core entities and add complexity through follow-up messages.
• Ask for quality rules. Tell the copilot: “Add quality rules to validate email format and ensure company names are not empty.”

When to Use the Copilot vs. Manual Editing

Scenario	Recommended Approach
Starting from scratch	Schema Copilot
Quick domain exploration	Schema Copilot
Precise control over every field	Manual YAML editing
Complex quality rules	Start with Copilot, refine in YAML editor
Existing schema to import	Upload YAML directly

Next Steps

YAML Syntax Reference

Understand the full ontology YAML format for manual editing

Quality Rules

Add validation rules to ensure data quality

Ontology Examples

Browse complete ontology examples for different domains

Transform Documents

Use your schema to extract knowledge graphs from documents