​
1. Accessing the Column Creation Interface

To create a conditional node column, go to the Data tab in your dataset and click the + Add Columns button. In the Dynamic Columns section, select Conditional Node.

​
2. Configuring the Conditional Node

Once selected, configure the following settings:

  • Name – Assign a name to this new column.
  • Each row in the dataset is processed based on the branching logic defined in the conditional node:
    • If Condition – The first condition to check.
    • Elif Conditions (optional) – Additional conditions checked sequentially if the first condition is false.
    • Else Condition (optional) – The default fallback when none of the conditions match.
  • Choosing an Operation Type: The system allows various operations when conditions are met
    • Run Prompt – Generates AI-driven responses using custom LLM prompts.
    • Retrieval – Fetches relevant data from a vector database via similarity search.
    • Extract Entities – Identifies and extracts key information from text columns.
    • Extract JSON Key – Retrieves specific values from JSON-formatted dataset columns.
    • Execute Custom Code – Runs Python scripts for custom row-level transformations.
    • Classification – Assigns labels to dataset rows using a pre-trained AI model.
    • API Calls – Integrates external APIs to fetch and populate dynamic column data.

Once created, the system evaluates each row, applying the conditional logic in sequence:

  1. Evaluates Conditions – Checks if, elif, and else in order.
  2. Executes Matching Operation – Applies the corresponding transformation.
  3. Stores Results – Saves the generated values in the new column.

​
Best Practices for Conditional Nodes

  • Ensure clear condition hierarchy (if → elif → else) to prevent logical conflicts.
  • Match data type with the intended operation to avoid conversion issues.
  • Use text transformation for modifying string data dynamically.
  • Apply classification logic for structured labelling of dataset rows.
  • If integrating API calls, ensure external sources return expected results.

Conditional nodes enable flexible and automated data transformations, allowing datasets to adapt dynamically based on logic-driven workflows.