Preprocessing
Configure transformations on input data before sending it to your AI/ML model.
Preprocessing allows you to transform or enrich the input data before it is sent to your AI/ML model endpoint. This is useful when your source data requires formatting, restructuring, or enhancement to match the model’s expected input.
With AI Squared, preprocessing is fully configurable through a no-code interface or optional custom logic for more advanced cases.
When to Use Preprocessing
- Format inputs to match the model schema (e.g., convert a date to ISO format)
- Add additional metadata required by the model
- Clean raw input (e.g., remove special characters from text)
- Combine or derive fields (e.g., full name = first + last)
How Preprocessing Works
Each input field can be passed through one or more transformations before being sent to the model. These transformations are applied in the order defined in the UI.
⚠️ Preprocessing does not modify your original data — it only adjusts the payload sent to the model for that request.
Common Use Cases
| Example Use Case | Transformation |
|---|---|
Format created_at timestamp | Convert to ISO 8601 |
| Combine first and last name | Join with space |
| Normalize text input | Lowercase, remove punctuation |
| Apply static fallback | Use default if no value found |
📸 Placeholder for: Screenshot of preprocessing config screen
Dynamic Input + Preprocessing
Preprocessing is often used alongside Dynamic Input Values to shape data pulled from apps like Salesforce, ServiceNow, or custom web tools.
📘 Example:
If you’re harvesting a value like deal_amount from a CRM, you might want to round it or convert it into another currency before sending it to the model.
Optional Scripting (Advanced)
In upcoming versions, advanced users may have the option to inject lightweight transformation scripts for more customized logic. Contact support to learn more about enabling this feature.
What’s Next
Now that your inputs are prepared, it’s time to define how your model’s responses are structured.
👉 Proceed to Output Schema to configure your response handling.