Skip to main content
Version: 0.0.39

Custom Models

Custom Models within Lakehousecat provide a powerful way to tailor AI interactions to specific business needs by enabling advanced configuration and integration with various data sources. This guide will walk you through the setup and configuration details to optimize your Custom Models.

Creating and Initial Configuration​

Step-by-Step Guide​

  1. Start in the Workspace

    • Navigate to your Lakehousecat workspace.

  2. Go to Models Section

    • Click on the 'Models' menu to explore available model configurations.

  3. Select "Custom Models"

    • Move to the "Custom Models" area to either view existing models or create new ones.

Essential Information​

When setting up a Custom Model, ensure you provide:

  • Model Name: Choose a clear and descriptive name for the model to ensure easy identification.
  • Foundation Model Linkage: Connect your Custom Model to a suitable Foundation Model that will serve as its computational base.
  • Description (Optional): Offer an optional description to elaborate on the model's intended use or capabilities.

In-Depth Configuration​

After initial setup, deeper configuration of Custom Models can enhance its analytical capabilities. The Custom Model area consists of various tabs such as General, Suggestion, Data Sources, Operations, Parameter, Hierarchies, Description, and Trainings.

Tabs Overview​

General​

  • Navigate to your model via Workspace > Models > Custom Models.
  • Configuration Details: View model name, Foundation Model linking, and input the System Prompt.
  • System Prompt: This should be entered using a microphone for optimization. You can also preview the prompt's impact on interactions.

Suggestion​

  • Model Suggestions: Automatically generated from linked data but can be manually added. These suggestions aid users in gaining analytical insights.

Data Sources​

  • Connecting Data Sources: Empower your custom models by linking to pre-defined data sources, which enriches interaction with corporate data.

Operations​

  • Automate processes within the model by creating semantic layers:
    • Create Semantic Model: Execute regularly for ongoing data interactions.
    • Update Semantic Model: Adjust as necessary based on data changes or use cases.
    • Delete Semantic Model: Removes layers but may impact charts/dashboard connections.

Note: Model Operations are available to Builders and Administrators only.

Parameter​

  • Configure model behavior parameters like Temperature and Seed. Leave parameters unchanged if unsure.

Hierarchies & Description​

  • Lists the hierarchies tied to the model and provides metadata in descriptions.

Trainings​

  • Training Data: Use this area for chart and dashboard trainings to refine model responses in context.

Best Practices​

  • System Prompt: Carefully crafted prompts enhance model interaction quality.
  • Data Integration: Ensure robust connections with relevant data sources.
  • Regular Updates: Model adjustments should reflect changes in business needs and data structures.

Troubleshooting​

If you face issues:

  • Data Link Errors: Verify data source registrations and linkages.
  • Permission Issues: Ensure you have the necessary access within the Lakehousecat workspace.

By following these guidelines, you can create, configure, and manage Custom Models effectively within Lakehousecat, further enhancing your analytics capabilities.