Module Guide: Plugins

The plugin system is DashAI's extensibility mechanism. It allows new capabilities — models, tasks, data formats, transformations, explainers, and metrics — to be added to the platform without modifying its core code. Plugins are distributed as standard Python packages on PyPI and installed directly from the DashAI interface.

---

How the Plugin System Works

DashAI discovers plugins through Python's entry points mechanism. When a Python package declares a dashai.plugins entry point in its pyproject.toml, DashAI's component registry automatically picks it up on startup and makes its components available in the appropriate sections of the interface.

This means:

• A plugin that adds a new model will appear in the Models module's available model list for the task it supports.
• A plugin that adds a new converter will appear in the Notebook's CONVERT panel.
• A plugin that adds a new dataloader will appear in the upload format selection.

No manual registration or configuration is required — the entry point declaration is sufficient.

---

What Plugins Can Contribute

A single plugin package can contribute any combination of the following component types:

Component Type	Where it appears in DashAI
Models	Models module — available model list for the relevant task
Tasks	Models module — task selection landing page
Data Loaders	Datasets module — upload format selector
Converters	Notebook module — CONVERT tab
Explainers	Models module — EXPLAINABILITY tab
Metrics	Models module — evaluation metrics for the relevant task

---

Installing Plugins

Navigate to the PLUGINS section in the top navigation bar. From there you can:

1. Search for plugins published on PyPI by name or keyword.
2. View plugin descriptions, supported component types, and version information.
3. Install with a single click — DashAI handles the pip installation and component registration automatically.
4. Restart DashAI if prompted — some plugins require a restart to fully register their components.

Once installed, new components appear immediately (or after restart) in their respective sections without any further configuration.

---

Plugin Structure

A DashAI plugin is a standard Python package with a specific structure:

plugin_name/
├── LICENSE
├── pyproject.toml
├── README.md
└── src/
    └── plugin_name/
        ├── my_model.py
        └── MyModel.json

Each component is implemented as a Python class that extends the appropriate DashAI base class, paired with a JSON schema file that describes its parameters for the interface.

The pyproject.toml must declare one entry point per component class:

[project.entry-points.'dashai.plugins']
MyModel = 'plugin_name.my_model:MyModel'

And must include the appropriate keywords so DashAI can categorize the plugin:

[project]
keywords = ["DashAI", "Package", "Model", "Task"]

Valid keywords: DashAI, Package, Task, Model, Dataloader, Converter, Explainer

---

Notable Published Plugins

DashAI's image generation capabilities are themselves distributed as plugins:

Plugin	What it adds
dashai-stable-diffusion-v1-model-package	Stable Diffusion v1 for text-to-image generation
dashai-flux-model-package	Flux model for text-to-image generation
dashai-stable-diffusion-controlnet-canny-model	ControlNet with Canny edge conditioning

This architecture — where even first-party capabilities are plugins — means the core platform stays lean and every feature is opt-in based on your hardware and use case.

---

Developing Your Own Plugins

Building a plugin requires:

1. Creating a Python class that extends the right DashAI base class for your component type (e.g., TabularClassificationModel for a tabular classification model).
2. Creating a JSON schema file that describes the component's parameters — this is what DashAI uses to generate the configuration UI automatically.
3. Packaging the code with proper pyproject.toml entry points.
4. Testing locally by placing the plugin in a plugins/ folder inside your DashAI development directory.
5. Publishing to PyPI when ready.

For a complete development guide including code examples, base class references, and a publishing walkthrough, see the Plugin Development section.

info

The entry point mechanism means any package on PyPI with the correct structure will work — there is no approval process or central registry beyond PyPI itself. You can also install plugins from local paths during development.