Skip to main content

Creating Your First Visual Workflow From Scratch

In the previous tutorial, you learned how to run and modify prebuilt Xircuits workflows by adjusting parameter components. While this is a great starting point, Xircuits offers much more flexibility. You can create your own custom workflows from scratch, combining components from various libraries we support. This tutorial will guide you through the process of building a more complex workflow, introducing you to the component library, how to install remote libraries, and showing you how to add and mix different components to create a powerful, customized solution.

Creating a New Workflow File​

Let's begin by setting up a new workflow in Xircuits:

  1. In the Launcher interface, click on Others > Xircuits File to create a new workflow file.
  2. Name your file MyFirstWorkflow.xircuits.
  3. You'll see a Start and Finish node automatically placed on the canvas. These nodes mark the beginning and end of your workflow.

The component library is your toolbox for building workflows. Let's explore how to use it effectively.

Adding Your First Component​

  1. On the left side of the Xircuits interface, click on the cube-like icon to open the component library tray.
  2. Use the search bar at the top to find the HelloComponent. Drag it onto the canvas.
  3. Connect the Start node to the HelloComponent node by dragging from the Start arrow outPort to the HelloComponent arrow inPort.
  4. Click Save and then Compile to generate a Python script with the same name as your .xircuits file.
  5. Run the workflow by clicking the 'Run' button. You should see an output like "Hello, User!" or a similar greeting.

Expanding Your Component Library​

Sometimes, you might need a library that isn't installed. Here's how to add it:

  1. In the component library tray, look for libraries listed under Available for Installation.
  2. To install a library, click on the ... next to it and select Install. The library will be cloned to the xai_components directory.
  3. You'll be prompted to choose a kernel, which will execute the pip install command to install the library.

For this tutorial, let's install the Tensorflow Keras library:

  1. Click on the ... next to TENSORFLOW_KERAS and select Install.
  2. Follow the prompts to complete the installation.

Once installed, the new components will be available in your library.

Constructing a More Complex Workflow​

Now, let's create a machine learning workflow by adding and connecting multiple components. You can reuse the previous canvas or create a new one.

Assembling the Core Components​

  1. Search for and drag these components onto the canvas, connecting them in this order:

    • ReadKerasDataSet (connect to Start node)
    • TrainTestSplit (connect to ReadKerasDataSet)
    • KerasCreate2DInputModel (connect to TrainTestSplit)
    • KerasTrainImageClassifier (connect to KerasCreate2DInputModel)
    • KerasEvaluateAccuracy (connect to KerasTrainImageClassifier)
    • Connect KerasEvaluateAccuracy to the Finish node to complete the workflow.

    Connecting the Components

    Notice how Xircuits automatically connects some ports for you. For example, when you connect ReadKerasDataSet to TrainTestSplit, the dataset outPort of ReadKerasDataSet is automatically connected to the dataset inPort of TrainTestSplit. A smart linking feature helps connect ports with similar names and compatible types.

Completing Compulsory Connections​

Try running the workflow now. You'll likely encounter an error message: "Please connect all [★]COMPULSORY InPorts." In Xircuits, ports marked with a ★ are compulsory and must be connected.

Take a moment to observe which compulsory inPorts aren't connected yet. Can you make educated guesses on how to connect them?

Here's what you need to do:

  1. Connect TrainTestSplit's train outPort to both KerasCreate2DModel's and KerasTrainImageClassifier's training_data inPort.

  2. Connect TrainTestSplit's test outPort to KerasEvaluateAccuracy's eval_dataset inPort.

    Providing Data between Components

    Hovering over a link or port will highlight the connections.

tip

If you find it difficult to see or manage connections between components, you can create custom routes for better clarity. Simply hold the Ctrl key and click on a link to create waypoints along the connection line, allowing you to route it around other components or organize your workflow visually.

Modifying Workflow Behavior with Parameter Components​

Parameter components allow you to set fixed values that can be used by other components in the workflow. Let's add some parameters using two different methods:

Method 1: Dragging Components from the Tray​

  1. Drag a Literal String component onto the canvas from the component tray.
  2. Set its value to "mnist".
  3. Connect this Literal String component to the dataset_name input port of the ReadKerasDataSet component.
  1. Click on the training_epochs input port of the KerasTrainImageClassifier component.
  2. Drag outwards to create a link. This will automatically spawn the appropriate input dialogue.
  3. In the dialogue that appears, enter the integer value 5.
  4. Press Enter to confirm the value.

Supplying Parameter Components

Both methods achieve the same result, creating parameter components that feed values into your workflow. The second method can be quicker as it automatically creates the correct type of literal component based on the port you're connecting to.

Using parameter components like these helps in making workflows flexible and easily configurable. You can easily change these values to adjust your workflow's behavior without modifying the core components.

Try both methods and see which one you prefer. As you become more comfortable with Xircuits, you'll likely find yourself using a combination of both techniques depending on the situation.

Launching Your Workflow​

Now it's time to see your creation in action:

  1. Press the Save & Compile button.
  2. Select the Python kernel to execute the workflow.

You should see output similar to this:

======================================
__   __  ___                _ _
\ \  \ \/ (_)_ __ ___ _   _(_) |_ ___
 \ \  \  /| | '__/ __| | | | | __/ __|
 / /  /  \| | | | (__| |_| | | |_\__ \
/_/  /_/\_\_|_|  \___|\__,_|_|\__|___/
======================================
Xircuits is running...
Executing: ReadKerasDataSet
Executing: TrainTestSplit
Executing: KerasCreate2DInputModel
Executing: KerasTrainImageClassifier
Epoch 1/5
...
Executing: KerasEvaluateAccuracy
{'loss': '0.07967730611562729', 'accuracy': '0.975803554058075'}
Finish Executing

Congratulations! You've just created a complex workflow from scratch, installed necessary libraries, and learned how to use literal components to customize your workflow parameters. As you become more comfortable with Xircuits, try exploring more components and creating more intricate workflows.