This feature may not be available by default and might need to be enabled upon request. Please contact your account manager for more information.
SqlDBM’s Copilot features are designed to enhance your data modeling workflow by automating routine tasks, providing intelligent suggestions, and accelerating schema creation. Copilot functionality is available throughout various parts of SqlDBM, and more features are being added.
Summary of AI Copilot capabilities:
Reverse Engineering - creates schemas from natural language prompts
Documentation - generates descriptions and logical names
Object-level - describes or summarizes objects, explains view logic, performs bulk changes, creates related objects, assigns flags, and more!
Model-level - Object-level features extended with the awareness of the entire project. Useful for understanding large schemas, detecting anomalies, and handling bulk changes.
Full list of current AI Copilot capabilities
You can reference the latest capabilities within the tool anytime by asking, "what can you do?"
Currently, the list includes:
- Database Object Creation
- Create tables with custom columns, primary keys, and data types
- Create views with custom query expressions
- Create schemas to organize database objects
- Add relationships between tables
- Define complex table types (regular, external, dynamic, hybrid, etc.)
- Object Metadata Management
- Add descriptions and comments to tables, views, columns, and schemas
- Apply flags/tags to objects and columns
- Rename objects (physical and logical names)
- Update column types and names
- Add new columns to existing tables
- Diagram and Subject Area Management
- Create and manage database diagrams
- Organize diagrams into subject areas
- Apply different layout styles (top-down, left-right, star layout)
- Place objects on diagrams
- Database Documentation and Data Governance
- Create custom metadata fields
- Set metadata values for tables, views, and columns
- Manage field definitions and bindings
- Add or improve descriptions for an entire schema in a single action.
- Advanced Querying
- Find and filter database objects using precise jq queries
- Retrieve detailed information about tables, views, and relationships
- Analyze database model structure
- Check naming conventions and standards
- Project Metadata Exploration
- Explore project settings
- Review naming conventions
Role-aware AI Experience
On first use, you need to select a specific role to tune the AI to your specific workflow.
Available roles:
Modeler/Architect: Focuses on schema generation, naming conventions, and impact analysis.
BI/Documentation Lead: Optimized for auto-filling descriptions and large-scale documentation.
Data Governance: Specialized in PII scanning, compliance documentation, and policy checks.
Analytics Engineer: Tailored for dbt YAML, semantic layers, and data contracts.
Consumer/Viewer: Provides plain-English explanations and model exploration.
Other/General: Provides full access without specific role filtering
Role-specific Pre-prompts
SqlDBM AI Copilot provides one-click pre-prompts that are dynamically tuned to your selected role. To maintain a clean workspace, these suggestions collapse into a single button after the first message, remaining easily accessible without cluttering the conversation.
The available actions and functionality vary based on the role you have selected:
Design — Generate schemas, build Data Vault structures, or reverse engineer models from business requirements.
Document — Add descriptions to all tables in bulk or automatically improve logical names.
Analyze — Perform impact analysis, check against naming standards, or generate a comprehensive compliance summary.
Govern — Perform automated PII scans to identify and flag sensitive data.
Project Level — Access model-wide bulk actions and global insights.
Project-level Copilot Settings
Project-level Copilot settings allow users to fine-tune the Copilot response to functionality described in this article by providing additional context and customizing the standard pre-prompts.
Access the settings from the main SqlDBM project menu or from the gear icon on the Model Copilot chat window.
The following settings are available
- Project global context: Describe your project, industry, specifications, conventions, or native language. This context will be included with all Copilot calls to fine-tune the Copilot response.
- Object pre-prompts: Customize the pre-prompts in Object Copilot. Provide the short text and instruction text for a maximum of two pre-prompts. Leave blank to disable the pre-prompt buttons.
- Model pre-prompts: Customize the pre-prompts in Model Copilot. Provide the short text and instruction text for a maximum of two pre-prompts. Leave blank to disable the pre-prompt buttons.
Reverse Engineering
Jumpstart your projects by generating complete sample schemas from natural language prompts. This feature works both for starting a new project and extending an existing one from natural language requirements. Use the paperclip icon to attach images or requirements documents to digitize existing designs.
How it works:
Open the Reverse Engineering tool in SqlDBM and click the "Generate with AI" button.
Enter a natural language description of the schema you want (e.g., “Generate a retail business schema using Kimball methodology”).
Optionally, attach images or documents to add instructions or context.
Review the generated schema and adjust as needed.
Tip: add newly created objects to an existing diagram by selecting it in the import options.
Common use cases:
Create sample industry models (e.g., healthcare, logistics, telecom).
Render legacy model images or whiteboard mockups as SqlDBM objects.
Quickly produce sample facts or dimensions to speed up design.
Capture detailed business requirements and upload them as an attachment to jump-start the design.
Generate example schemas for OLTP or OLAP environments.
Illustrate modeling concepts such as normalization, star, or snowflake schemas.
Take natural language requirements and convert them into DDL.
Sample prompts:
- Create a model from the requirement below: ...
- Generate a 3NF normalized database schema for a university management system.
- Create a logical model based on the whiteboard sketch in the attached photo.
- Create a Kimball-style DWH for the automotive industry.
- Create a physical model based on the requirements described in the document attached.
- Using the provided database model, generate a Data Vault model with Links, Hubs, and Sats...
Documentation
Simplify documentation by generating meaningful descriptions and logical names for columns, tables, and schemas — either individually or in bulk. The AI uses the full project context to generate relevant logical names and descriptions.
Available Actions:
- Add descriptions to all tables in one action
- Add or improve descriptions — new and existing
- Generate logical names for columns and tables
Object-level
Purpose:
Interact directly with objects in your diagram to get context-aware assistance, modify structures, or generate related content.
Calling the Copilot from an object provides context about that object and related objects within two hops, based on their relationships.
Common use cases:
Ask questions about a table, view, or its dependencies.
Automatically add or improve descriptions for an object and its columns.
Create new objects based on the selected one.
Modify the structure of an existing object (e.g., bulk convert data types, add fields, fix naming conventions).
Assign flags to highlight PII sensitivity or other criteria.
Ask "what can Copilot do?"
Sample prompts:
Explain this view and list its sources and any relevant filters and business logic.
Will removing Column_X from this object affect its related objects?
Convert all columns with data type "string" to "varchar".
Add a child table and include a date column called "reviewed_date".
Flag any columns with potential PII data with a red flag.
More ideas
| Short Text | Prompt |
| Suggest improvements | Please suggest improvements using the provided database model, including Primary Keys (PKs) and Foreign Keys (FKs), object and column descriptions, and metadata (flags). |
| List flagged elements | List flagged elements in the database model. Make sure to look at flags at the object and column level. |
| Show missing descriptions | List any objects and related columns that are missing a description |
| Summarize model | Summarize the provided database model. Include DDL, primary keys, foreign keys, object and column descriptions, and metadata like flags. |
| Views w/ source changes | List any views that have a changed_on date that is earlier than the changed_on date of one of its sources. |
| Explain object/logic | Explain the purpose and logic of the database object using the provided database model. If the object is a view or contains SQL logic, parse it to understand the sources, outputs, and relevant business logic and filters. |
| Impact analysis | Using the provided database model, perform an impact analysis on the fields of this object and potential impact on its related objects. |
How it works:
Hover over any table to reveal the orange Copilot logo at the top-right corner. Click the logo to open the prompt window.
Choose from available actions or type your request (e.g., “Create a child table for this object”).
If performing an action, the chat window closes upon completion. If asking a question, the chat window remains for further prompts.
Model-level
Model-level Copilot lives in a sidebar chat window and has access to the entire project (latest revision). Modelers and business users alike can ask questions, find objects, suggest improvements, and make changes using natural language. This feature provides the same functionality as object-level Copilot but offers visibility into the entire project for a comprehensive understanding.
To access the Copilot sidebar, click on the orange Copilot logo in the top-right corner of any SqlDBM screen. Click again to dismiss or recall the panel. Your prompt history will be preserved.
Common use cases:
- A consumer discovering the important facts and dimensions in a project shared with them.
- Model validation for standards, exceptions, and anomalies like missing PKs or columns that do not follow established naming standards.
- Suggesting improvements of missing keys, descriptions, or discovering potential duplicate data sources.
- Bulk changes, such as flagging columns or objects across the entire project or replacing data types.
Sample prompts:
- Explain this model, highlighting its key dimensions, facts, and reporting sources, including relevant tables and views.
- Flag all PII fields using the red flag.
- List all tables that are missing a primary key.
- List any tables that don't have a primary key that meets the following criteria: minimum length is 5 characters, must be suffixed with "_key" or "_id".
- Create a denormalized table called X that contains all columns from tables Y and Z. Do not include duplicate columns.
- Create a view called X that contains all columns from tables Y and Z. All columns must be named in table.column format. Format the view logic neatly, on multiple lines, using CTE format.
Conversation History
Automatic saving: Every AI session is automatically saved, allowing you to pick up exactly where you left off across different days.
Organization: History is tied to both the specific user and the project - separate per project
-
Management: Users can browse past conversations via the history icon, view message counts and timestamps, or delete old threads.
AI Security and Data Privacy Summary
SqlDBM’s Copilot features are built with security-first principles to protect your sensitive database information. No AI code or internal models are embedded in the SqlDBM application — all AI functionality is handled through secure, on-demand API calls to Anthropic’s Claude models. Your data is never used to train AI models. Only the specific context you choose to share is sent for processing, and the AI provider does not retain your data after the request completes.
Access controls and user responsibility:
No AI features are available at the account level unless explicitly requested.
No AI processing occurs unless you actively use the feature.
Always review AI-generated content before deploying to production and be mindful of the sensitive data you choose to include.