Skip to main content

Developing a Plugin

Prerequisites​

Before developing a plugin, ensure you understand:

  1. What is a Plugin? — High-level overview of plugin concepts and capabilities
  2. Plugin Structure — How plugins are organized and what DashAI requires

Setup Your Development Environment​

To create a plugin, you need access to DashAI's Python classes. There are two approaches:

Option 1: DashAI as an Installed Library​

Install DashAI as a package in your Python virtual environment:

pip install dashai

Note: This approach is simpler but limits your ability to test the plugin interactively within DashAI.

Clone the DashAI repository for full development and testing capabilities:

git clone https://github.com/DashAISoftware/DashAI.git
cd DashAI

Create a plugins folder in the repository root for your plugin development:

DashAI/
├── DashAI/
├── plugins/                  ← Create this folder
├── tests/
├── ...

This approach lets you test your plugin in real-time as you develop.

Step 1: Identify What You Want to Build​

Plugins can extend DashAI with:

  • Machine Learning Models (TabularClassificationModel, RegressionModel, TextGenerationModel, etc.)
  • Data Loaders (support for additional dataset formats)
  • Data Converters (preprocessing, feature engineering, transformations)
  • Explorers (data visualization and analysis tools)
  • Explainers (model interpretability tools)
  • Custom Tasks (entirely new ML problem types)
  • Custom Metrics (evaluation metrics)

To ensure compatibility, your plugin classes must extend the appropriate DashAI base class:

# Example: Creating a tabular classification model
from DashAI.back.models.tabular_classification_model import TabularClassificationModel

class MyCustomModel(TabularClassificationModel):
    def train(self, X, y):
        # Your training logic
        pass

Step 2: Implement Your Plugin​

  1. Create the plugin folder inside the plugins directory (if using Option 2 above)
  2. Write your Python classes extending appropriate DashAI base classes
  3. Create any required JSON configuration files
  4. Add the pyproject.toml with proper entry points (see Plugin Structure)
  5. Write a README describing your plugin

Recommendations​

Review Existing Plugins​

Study working examples to understand best practices:

Test Your Plugin During Development​

  1. Create the Python and JSON files locally (Option 2: clone repository)
  2. Place them in the /plugins folder
  3. Start DashAI and verify your components appear and work correctly
  4. Iterate until you're confident in the implementation

Check for Library Conflicts​

After installing new dependencies, verify no package conflicts exist:

pip check

No conflicts:

No broken requirements found.

With conflicts:

fastapi 0.106.0 has requirement starlette<0.28.0,>=0.27.0, but you have starlette 0.20.0.

Resolve any conflicts before publishing your plugin.

It is important that when installing a new library for a plugin or package, it must be included among the dependencies of the plugin or package.

  1. Use prints in the flow of the added components

    Including prints in the flow of the component you created can be very useful to verify the proper functioning of the component you have created in the software.