This section answers frequently asked questions about configuring AI models within AI Squared—including how to define schemas, preprocess inputs, and ensure model compatibility.

​
Input & Output Schema

​
What is the input schema used for?

The input schema defines the structure of data sent to your model. Each key in the schema maps to a variable used by the model endpoint.
  • Ensure all required fields are covered
  • Match data types (string, integer, float, boolean) exactly
  • Use dynamic/static value tagging depending on your use case

​
What is the output schema used for?

The output schema maps the model’s prediction response to fields that can be used in visualizations and downstream applications.
  • Identify key-value pairs in the model response
  • Use flat structures (nested objects not supported currently)
  • Label predictions clearly for user-facing display

​
Preprocessing & Formatting

​
How do I clean or transform inputs before sending them to the model?

Use AI Squared’s built-in Preprocessing Rules, which allow no-code logic to:
  • Format strings or numbers (e.g., round decimals)
  • Apply conditional logic (e.g., replace nulls)
  • Normalize or scale inputs
Preprocessors are configurable per input field.

​
Updating a Model Source

​
Can I update a connected model after initial setup?

Yes, you can:
  • Edit endpoint details (URL, auth, headers)
  • Modify input/output schemas
  • Add or update preprocessing rules
Changes will reflect in associated Data Apps using the model.

​
Debugging Model Issues

​
How can I test if my model responds correctly?

  1. Navigate to the AI Modeling section and click on Test Model
  2. Provide sample input data based on your schema
  3. Review the response payload
Common issues include:
  • Auth failures (invalid API keys or tokens)
  • Incorrect input field names or types
  • Mismatched response format from expected schema