Intro to custom components

The first step in developing a Streamlit Component is deciding whether to create a static component (i.e. rendered once, controlled by Python) or to create a bi-directional component that can communicate from Python to JavaScript and back.

If your goal in creating a Streamlit Component is solely to display HTML code or render a chart from a Python visualization library, Streamlit provides two methods that greatly simplify the process: components.html() and components.iframe().

If you are unsure whether you need bi-directional communication, start here first!

While st.text, st.markdown and st.write make it easy to write text to a Streamlit app, sometimes you'd rather implement a custom piece of HTML. Similarly, while Streamlit natively supports many charting libraries, you may want to implement a specific HTML/JavaScript template for a new charting library. components.html works by giving you the ability to embed an iframe inside of a Streamlit app that contains your desired output.

Example

import streamlit as st import streamlit.components.v1 as components # bootstrap 4 collapse example components.html( """ <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/css/bootstrap.min.css" integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous"> <script src="https://code.jquery.com/jquery-3.2.1.slim.min.js" integrity="sha384-KJ3o2DKtIkvYIK3UENzmM7KCkRr/rE9/Qpg6aAZGJwFDMVNA/GpGFF93hXpG5KkN" crossorigin="anonymous"></script> <script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/js/bootstrap.min.js" integrity="sha384-JZR6Spejh4U02d8jOt6vLEHfe/JQGiRRSQQxSfFWpi1MquVdAyjUar5+76PVCmYl" crossorigin="anonymous"></script> <div id="accordion"> <div class="card"> <div class="card-header" id="headingOne"> <h5 class="mb-0"> <button class="btn btn-link" data-toggle="collapse" data-target="#collapseOne" aria-expanded="true" aria-controls="collapseOne"> Collapsible Group Item #1 </button> </h5> </div> <div id="collapseOne" class="collapse show" aria-labelledby="headingOne" data-parent="#accordion"> <div class="card-body"> Collapsible Group Item #1 content </div> </div> </div> <div class="card"> <div class="card-header" id="headingTwo"> <h5 class="mb-0"> <button class="btn btn-link collapsed" data-toggle="collapse" data-target="#collapseTwo" aria-expanded="false" aria-controls="collapseTwo"> Collapsible Group Item #2 </button> </h5> </div> <div id="collapseTwo" class="collapse" aria-labelledby="headingTwo" data-parent="#accordion"> <div class="card-body"> Collapsible Group Item #2 content </div> </div> </div> </div> """, height=600, )

components.iframe is similar in features to components.html, with the difference being that components.iframe takes a URL as its input. This is used for situations where you want to include an entire page within a Streamlit app.

Example

import streamlit as st import streamlit.components.v1 as components # embed streamlit docs in a streamlit app components.iframe("https://example.com", height=500)

A bi-directional Streamlit Component has two parts:

  1. A frontend, which is built out of HTML and any other web tech you like (JavaScript, React, Vue, etc.), and gets rendered in Streamlit apps via an iframe tag.
  2. A Python API, which Streamlit apps use to instantiate and talk to that frontend

To make the process of creating bi-directional Streamlit Components easier, we've created a React template and a TypeScript-only template in the Streamlit Component-template GitHub repo. We also provide some example Components in the same repo.

To build a Streamlit Component, you need the following installed in your development environment:

  • Python 3.8 - Python 3.12
  • Streamlit 1.11.1 or higher
  • nodejs
  • npm or yarn

Clone the component-template GitHub repo, then decide whether you want to use the React.js ("template") or plain TypeScript ("template-reactless") template.

  1. Initialize and build the component template frontend from the terminal:

    # React template template/my_component/frontend npm install # Initialize the project and install npm dependencies npm run start # Start the Webpack dev server # or # TypeScript-only template template-reactless/my_component/frontend npm install # Initialize the project and install npm dependencies npm run start # Start the Webpack dev server
  2. From a separate terminal, run the Streamlit app (Python) that declares and uses the component:

    # React template cd template . venv/bin/activate # or similar to activate the venv/conda environment where Streamlit is installed pip install -e . # install template as editable package streamlit run my_component/example.py # run the example # or # TypeScript-only template cd template-reactless . venv/bin/activate # or similar to activate the venv/conda environment where Streamlit is installed pip install -e . # install template as editable package streamlit run my_component/example.py # run the example

After running the steps above, you should see a Streamlit app in your browser that looks like this:

Streamlit Component Example App

The example app from the template shows how bi-directional communication is implemented. The Streamlit Component displays a button (Python → JavaScript), and the end-user can click the button. Each time the button is clicked, the JavaScript front-end increments the counter value and passes it back to Python (JavaScript → Python), which is then displayed by Streamlit (Python → JavaScript).

Because each Streamlit Component is its own webpage that gets rendered into an iframe, you can use just about any web tech you'd like to create that web page. We provide two templates to get started with in the Streamlit Components-template GitHub repo; one of those templates uses React and the other does not.

push_pin

Note

Even if you're not already familiar with React, you may still want to check out the React-based template. It handles most of the boilerplate required to send and receive data from Streamlit, and you can learn the bits of React you need as you go.

If you'd rather not use React, please read this section anyway! It explains the fundamentals of Streamlit ↔ Component communication.

React

The React-based template is in template/my_component/frontend/src/MyComponent.tsx.

  • MyComponent.render() is called automatically when the component needs to be re-rendered (just like in any React app)
  • Arguments passed from the Python script are available via the this.props.args dictionary:
# Send arguments in Python: result = my_component(greeting="Hello", name="Streamlit")
// Receive arguments in frontend: let greeting = this.props.args["greeting"]; // greeting = "Hello" let name = this.props.args["name"]; // name = "Streamlit"
  • Use Streamlit.setComponentValue() to return data from the component to the Python script:
// Set value in frontend: Streamlit.setComponentValue(3.14);
# Access value in Python: result = my_component(greeting="Hello", name="Streamlit") st.write("result = ", result) # result = 3.14

When you call Streamlit.setComponentValue(new_value), that new value is sent to Streamlit, which then re-executes the Python script from top to bottom. When the script is re-executed, the call to my_component(...) will return the new value.

From a code flow perspective, it appears that you're transmitting data synchronously with the frontend: Python sends the arguments to JavaScript, and JavaScript returns a value to Python, all in a single function call! But in reality this is all happening asynchronously, and it's the re-execution of the Python script that achieves the sleight of hand.

  • Use Streamlit.setFrameHeight() to control the height of your component. By default, the React template calls this automatically (see StreamlitComponentBase.componentDidUpdate()). You can override this behavior if you need more control.
  • There's a tiny bit of magic in the last line of the file: export default withStreamlitConnection(MyComponent) - this does some handshaking with Streamlit, and sets up the mechanisms for bi-directional data communication.

TypeScript-only

The TypeScript-only template is in template-reactless/my_component/frontend/src/MyComponent.tsx.

This template has much more code than its React sibling, in that all the mechanics of handshaking, setting up event listeners, and updating the component's frame height are done manually. The React version of the template handles most of these details automatically.

  • Towards the bottom of the source file, the template calls Streamlit.setComponentReady() to tell Streamlit it's ready to start receiving data. (You'll generally want to do this after creating and loading everything that the Component relies on.)
  • It subscribes to Streamlit.RENDER_EVENT to be notified of when to redraw. (This event won't be fired until setComponentReady is called)
  • Within its onRender event handler, it accesses the arguments passed in the Python script via event.detail.args
  • It sends data back to the Python script in the same way that the React template does—clicking on the "Click Me!" button calls Streamlit.setComponentValue()
  • It informs Streamlit when its height may have changed via Streamlit.setFrameHeight()

Working with Themes

push_pin

Note

Custom component theme support requires streamlit-component-lib version 1.2.0 or higher.

Along with sending an args object to your component, Streamlit also sends a theme object defining the active theme so that your component can adjust its styling in a compatible way. This object is sent in the same message as args, so it can be accessed via this.props.theme (when using the React template) or event.detail.theme (when using the plain TypeScript template).

The theme object has the following shape:

{ "base": "lightORdark", "primaryColor": "someColor1", "backgroundColor": "someColor2", "secondaryBackgroundColor": "someColor3", "textColor": "someColor4", "font": "someFont" }

The base option allows you to specify a preset Streamlit theme that your custom theme inherits from. Any theme config options not defined in your theme settings have their values set to those of the base theme. Valid values for base are "light" and "dark".

Note that the theme object has fields with the same names and semantics as the options in the "theme" section of the config options printed with the command streamlit config show.

When using the React template, the following CSS variables are also set automatically.

--base --primary-color --background-color --secondary-background-color --text-color --font

If you're not familiar with CSS variables, the TLDR version is that you can use them like this:

.mySelector { color: var(--text-color); }

These variables match the fields defined in the theme object above, and whether to use CSS variables or the theme object in your component is a matter of personal preference.

Other frontend details

  • Because you're hosting your component from a dev server (via npm run start), any changes you make should be automatically reflected in the Streamlit app when you save.
  • If you want to add more packages to your component, run npm add to add them from within your component's frontend/ directory.
npm add baseui
  • To build a static version of your component, run npm run export. See Prepare your Component for more information

components.declare_component() is all that's required to create your Component's Python API:

import streamlit.components.v1 as components my_component = components.declare_component( "my_component", url="http://localhost:3001" )

You can then use the returned my_component function to send and receive data with your frontend code:

# Send data to the frontend using named arguments. return_value = my_component(name="Blackbeard", ship="Queen Anne's Revenge") # `my_component`'s return value is the data returned from the frontend. st.write("Value = ", return_value)

While the above is all you need to define from the Python side to have a working Component, we recommend creating a "wrapper" function with named arguments and default values, input validation and so on. This will make it easier for end-users to understand what data values your function accepts and allows for defining helpful docstrings.

Please see this example from the Components-template for an example of creating a wrapper function.

Python → Frontend

You send data from Python to the frontend by passing keyword args to your Component's invoke function (that is, the function returned from declare_component). You can send the following types of data from Python to the frontend:

  • Any JSON-serializable data
  • numpy.array
  • pandas.DataFrame

Any JSON-serializable data gets serialized to a JSON string, and deserialized to its JavaScript equivalent. numpy.array and pandas.DataFrame get serialized using Apache Arrow and are deserialized as instances of ArrowTable, which is a custom type that wraps Arrow structures and provides a convenient API on top of them.

Check out the CustomDataframe and SelectableDataTable Component example code for more context on how to use ArrowTable.

Frontend → Python

You send data from the frontend to Python via the Streamlit.setComponentValue() API (which is part of the template code). Unlike arg-passing from Python → frontend, this API takes a single value. If you want to return multiple values, you'll need to wrap them in an Array or Object.

Custom Components can send JSON-serializable data from the frontend to Python, as well as Apache Arrow ArrowTables to represent dataframes.

forum

Still have questions?

Our forums are full of helpful information and Streamlit experts.