Create a Pipeline

To create a paragraph using pipeline:

1. Navigate to the Pipeline Designer page.
   The Start widget is displayed by default in the canvas screen.
2. Hover over the Start widget and new nodes can be added by using the Add button in the Pipeline Canvas.
   When the Add button is clicked, a menu opens that contains all the widgets sectioned into different categories. The spectra diagram builder is based on row and column positioning.

   Figure 11-85 Create a Pipeline menu

   
   

   

   
   

   Note:

   Click each menu option to select the required widgets.
   You cannot edit or delete the Start widget. Whenever a new draft is created, not by importing dump files, the default paragraph that is created is converted into a Start widget. The visibility of code or result or the title in the notebook of this node will be invisible.
   Whenever the notebook is opened, the script execution, including the workspace attachment will happen in this node.

   Table 11-17 Widgets

   Widget	Description
   Load Dataset Widget	Push Down Filtering (SQL Filters) and Python Script Filtering is available for the Load Dataset widget in the pipeline. Figure 11-86 Pushdown (SQL) Filter and Python Script Filter Examples: Pushdown (SQL) Filter: - column1 > 100 and column2 < 1000 Python Script Filtering: df_out = df_prev[df_prev['column'] > 100]
   Start Widget	Publish, download, import, and promoting a model to production can be done via this widget. When you publish a model from the Model Summary page, you can select the Start widget paragraph from the list of paragraphs and then the start widget will appear in the published model.
   Exploratory Data Analysis (EDA) Widget	This widget provides a one-click, automated evaluation of any dataset within OFS MMG. It provides a rapid insight into the structure, quality, and characteristics of data, and generates visual summaries, statistics, and distribution plots without requiring any coding or a manual setup. This widget automatically generates the following analyses for a dataset: Missing Value & Schema Summary Creates a table that lists each feature or column. Shows the number and percentage of missing values per column. Displays the inferred data type (e.g., numeric, categorical, text). This helps users to quickly identify columns with significant missing data or non-standard data types that may need cleaning or attention. Numeric Feature Analysis (For Columns Detected as Numeric) Distribution Plots: Histograms combined with a smooth density (KDE) curve display how data points are spread across ranges of values. Box Plots: Allows quick detection of outliers, medians, and interquartile ranges. Purpose: Enables users to see which values are common, detect skewness or unusual patterns, and spot potential anomalies. Correlation Heatmap Displays a color-coded matrix showing how strongly pairs of numeric features are correlated. Helps in detecting strong associations, redundancy (collinearity), or identifying which variables might influence each other. Categorical Feature Analysis (For any Column Detected as Categorical or Text:) Frequency Plots: Bar charts that display how often each category appears within the column. Purpose: Visualizes dominant and rare categories, helping to find class imbalance or need for further grouping.

3. Click the node to add the basic details.
   The Basic Details page is displayed.

   Figure 11-87 Basic Details for the Paragraph

   
   

   

   
   
4. Provide the following details:

   Table 11-18 Adding Details

   Field	Description
   Activity Name	Enter the activity name
   Description	Enter the description of the activity
   Task Type	Select the task type. For example, model training, data analysis and so on. You can also search the task type.
   Track Output	If this option is selected for any paragraph, then during the model comparison the output details are displayed. Keep the Track Output as ON in case you want to execute the paragraph and view the result from the Dashboard tab.
   Script	Displays the script. You can edit the script in this screen or in the Script tab. This script can also be saved as the script template.

   Note:

   If is a conditional node, whose behavior is based on the execution of the user. It has a script that evaluates True or False and based on the value, it chooses either of them. Additionally, you can also add a new node that comes with a default link. The default path will be executed irrespective of the result.
5. To execute a node, click the Execute icon. If no active or a selected session is available, then the Session Required dialog box appears, informing you that it cannot be executed with no selected session and you need to select a valid session or create a new one. Either click the Create New button to create a new session, or click the Cancel button, select a session and then click Execute.

   Note:

   The Session Required dialog box will not appear if the session is active or if the session has been selected.