Custom Components GalleryNEW

Explore

Blocks and Event Listeners

We briefly descirbed the Blocks class in the Quickstart as a way to build custom demos. Let’s dive deeper.

Blocks Structure

Take a look at the demo below.

import gradio as gr


def greet(name):
    return "Hello " + name + "!"


with gr.Blocks() as demo:
    name = gr.Textbox(label="Name")
    output = gr.Textbox(label="Output Box")
    greet_btn = gr.Button("Greet")
    greet_btn.click(fn=greet, inputs=name, outputs=output, api_name="greet")

demo.launch()

  • First, note the with gr.Blocks() as demo: clause. The Blocks app code will be contained within this clause.
  • Next come the Components. These are the same Components used in Interface. However, instead of being passed to some constructor, Components are automatically added to the Blocks as they are created within the with clause.
  • Finally, the click() event listener. Event listeners define the data flow within the app. In the example above, the listener ties the two Textboxes together. The Textbox name acts as the input and Textbox output acts as the output to the greet method. This dataflow is triggered when the Button greet_btn is clicked. Like an Interface, an event listener can take multiple inputs or outputs.

You can also attach event listeners using decorators - skip the fn argument and assign inputs and outputs directly:

import gradio as gr


with gr.Blocks() as demo:
    name = gr.Textbox(label="Name")
    output = gr.Textbox(label="Output Box")
    greet_btn = gr.Button("Greet")

    @greet_btn.click(inputs=name, outputs=output)
    def greet(name):
        return "Hello " + name + "!"

   

demo.launch()

Event Listeners and Interactivity

In the example above, you’ll notice that you are able to edit Textbox name, but not Textbox output. This is because any Component that acts as an input to an event listener is made interactive. However, since Textbox output acts only as an output, Gradio determines that it should not be made interactive. You can override the default behavior and directly configure the interactivity of a Component with the boolean interactive keyword argument.

output = gr.Textbox(label="Output", interactive=True)

Note: What happens if a Gradio component is neither an input nor an output? If a component is constructed with a default value, then it is presumed to be displaying content and is rendered non-interactive. Otherwise, it is rendered interactive. Again, this behavior can be overridden by specifying a value for the interactive argument.

Types of Event Listeners

Take a look at the demo below:

import gradio as gr

def welcome(name):
    return f"Welcome to Gradio, {name}!"

with gr.Blocks() as demo:
    gr.Markdown(
    """
    # Hello World!
    Start typing below to see the output.
    """)
    inp = gr.Textbox(placeholder="What is your name?")
    out = gr.Textbox()
    inp.change(welcome, inp, out)

demo.launch()

Instead of being triggered by a click, the welcome function is triggered by typing in the Textbox inp. This is due to the change() event listener. Different Components support different event listeners. For example, the Video Component supports a play() event listener, triggered when a user presses play. See the Docs for the event listeners for each Component.

Multiple Data Flows

A Blocks app is not limited to a single data flow the way Interfaces are. Take a look at the demo below:

import gradio as gr

def increase(num):
    return num + 1

with gr.Blocks() as demo:
    a = gr.Number(label="a")
    b = gr.Number(label="b")
    atob = gr.Button("a > b")
    btoa = gr.Button("b > a")
    atob.click(increase, a, b)
    btoa.click(increase, b, a)

demo.launch()

Note that num1 can act as input to num2, and also vice-versa! As your apps get more complex, you will have many data flows connecting various Components.

Here’s an example of a “multi-step” demo, where the output of one model (a speech-to-text model) gets fed into the next model (a sentiment classifier).

from transformers import pipeline

import gradio as gr

asr = pipeline("automatic-speech-recognition", "facebook/wav2vec2-base-960h")
classifier = pipeline("text-classification")


def speech_to_text(speech):
    text = asr(speech)["text"]
    return text


def text_to_sentiment(text):
    return classifier(text)[0]["label"]


demo = gr.Blocks()

with demo:
    audio_file = gr.Audio(type="filepath")
    text = gr.Textbox()
    label = gr.Label()

    b1 = gr.Button("Recognize Speech")
    b2 = gr.Button("Classify Sentiment")

    b1.click(speech_to_text, inputs=audio_file, outputs=text)
    b2.click(text_to_sentiment, inputs=text, outputs=label)

demo.launch()

Function Input List vs Dict

The event listeners you’ve seen so far have a single input component. If you’d like to have multiple input components pass data to the function, you have two options on how the function can accept input component values:

  1. as a list of arguments, or
  2. as a single dictionary of values, keyed by the component

Let’s see an example of each:

import gradio as gr

with gr.Blocks() as demo:
    a = gr.Number(label="a")
    b = gr.Number(label="b")
    with gr.Row():
        add_btn = gr.Button("Add")
        sub_btn = gr.Button("Subtract")
    c = gr.Number(label="sum")

    def add(num1, num2):
        return num1 + num2
    add_btn.click(add, inputs=[a, b], outputs=c)

    def sub(data):
        return data[a] - data[b]
    sub_btn.click(sub, inputs={a, b}, outputs=c)


demo.launch()

Both add() and sub() take a and b as inputs. However, the syntax is different between these listeners.

  1. To the add_btn listener, we pass the inputs as a list. The function add() takes each of these inputs as arguments. The value of a maps to the argument num1, and the value of b maps to the argument num2.
  2. To the sub_btn listener, we pass the inputs as a set (note the curly brackets!). The function sub() takes a single dictionary argument data, where the keys are the input components, and the values are the values of those components.

It is a matter of preference which syntax you prefer! For functions with many input components, option 2 may be easier to manage.

Function Return List vs Dict

Similarly, you may return values for multiple output components either as:

  1. a list of values, or
  2. a dictionary keyed by the component

Let’s first see an example of (1), where we set the values of two output components by returning two values:

with gr.Blocks() as demo:
    food_box = gr.Number(value=10, label="Food Count")
    status_box = gr.Textbox()
    def eat(food):
        if food > 0:
            return food - 1, "full"
        else:
            return 0, "hungry"
    gr.Button("EAT").click(
        fn=eat,
        inputs=food_box,
        outputs=[food_box, status_box]
    )

Above, each return statement returns two values corresponding to food_box and status_box, respectively.

Instead of returning a list of values corresponding to each output component in order, you can also return a dictionary, with the key corresponding to the output component and the value as the new value. This also allows you to skip updating some output components.

with gr.Blocks() as demo:
    food_box = gr.Number(value=10, label="Food Count")
    status_box = gr.Textbox()
    def eat(food):
        if food > 0:
            return {food_box: food - 1, status_box: "full"}
        else:
            return {status_box: "hungry"}
    gr.Button("EAT").click(
        fn=eat,
        inputs=food_box,
        outputs=[food_box, status_box]
    )

Notice how when there is no food, we only update the status_box element. We skipped updating the food_box component.

Dictionary returns are helpful when an event listener affects many components on return, or conditionally affects outputs and not others.

Keep in mind that with dictionary returns, we still need to specify the possible outputs in the event listener.

Updating Component Configurations

The return value of an event listener function is usually the updated value of the corresponding output Component. Sometimes we want to update the configuration of the Component as well, such as the visibility. In this case, we return a new Component, setting the properties we want to change.

import gradio as gr


def change_textbox(choice):
    if choice == "short":
        return gr.Textbox(lines=2, visible=True)
    elif choice == "long":
        return gr.Textbox(lines=8, visible=True, value="Lorem ipsum dolor sit amet")
    else:
        return gr.Textbox(visible=False)


with gr.Blocks() as demo:
    radio = gr.Radio(
        ["short", "long", "none"], label="What kind of essay would you like to write?"
    )
    text = gr.Textbox(lines=2, interactive=True, show_copy_button=True)
    radio.change(fn=change_textbox, inputs=radio, outputs=text)


demo.launch()

See how we can configure the Textbox itself through a new gr.Textbox() method. The value= argument can still be used to update the value along with Component configuration. Any arguments we do not set will use their previous values.

Examples

Just like with gr.Interface, you can also add examples for your functions when you are working with gr.Blocks. In this case, instantiate a gr.Examples similar to how you would instantiate any other component. The constructor of gr.Examples takes two required arguments:

  • examples: a nested list of examples, in which the outer list consists of examples and each inner list consists of an input corresponding to each input component
  • inputs: the component or list of components that should be populated when the examples are clicked

You can also set cache_examples=True similar to gr.Interface, in which case two additional arguments must be provided:

  • outputs: the component or list of components corresponding to the output of the examples
  • fn: the function to run to generate the outputs corresponding to the examples

Here’s an example showing how to use gr.Examples in a gr.Blocks app:

import gradio as gr


def calculator(num1, operation, num2):
    if operation == "add":
        return num1 + num2
    elif operation == "subtract":
        return num1 - num2
    elif operation == "multiply":
        return num1 * num2
    elif operation == "divide":
        return num1 / num2


with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            num_1 = gr.Number(value=4)
            operation = gr.Radio(["add", "subtract", "multiply", "divide"])
            num_2 = gr.Number(value=0)
            submit_btn = gr.Button(value="Calculate")
        with gr.Column():
            result = gr.Number()

    submit_btn.click(calculator, inputs=[num_1, operation, num_2], outputs=[result], api_name=False)
    examples = gr.Examples(examples=[[5, "add", 3],
                                     [4, "divide", 2],
                                     [-4, "multiply", 2.5],
                                     [0, "subtract", 1.2]],
                           inputs=[num_1, operation, num_2])

if __name__ == "__main__":
    demo.launch(show_api=False)

Note: In Gradio 4.0 or later, when you click on examples, not only does the value of the input component update to the example value, but the component’s configuration also reverts to the properties with which you constructed the component. This ensures that the examples are compatible with the component even if its configuration has been changed.

Running Events Consecutively

You can also run events consecutively by using the then method of an event listener. This will run an event after the previous event has finished running. This is useful for running events that update components in multiple steps.

For example, in the chatbot example below, we first update the chatbot with the user message immediately, and then update the chatbot with the computer response after a simulated delay.

import gradio as gr
import random
import time

with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    def user(user_message, history):
        return "", history + [[user_message, None]]

    def bot(history):
        bot_message = random.choice(["How are you?", "I love you", "I'm very hungry"])
        time.sleep(2)
        history[-1][1] = bot_message
        return history

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, chatbot, chatbot
    )
    clear.click(lambda: None, None, chatbot, queue=False)
    
demo.queue()
demo.launch()

The .then() method of an event listener executes the subsequent event regardless of whether the previous event raised any errors. If you’d like to only run subsequent events if the previous event executed successfully, use the .success() method, which takes the same arguments as .then().

Running Events Continuously

You can run events on a fixed schedule using the every parameter of the event listener. This will run the event every number of seconds while the client connection is open. If the connection is closed, the event will stop running after the following iteration. Note that this does not take into account the runtime of the event itself. So a function with a 1 second runtime running with every=5, would actually run every 6 seconds. Also note that this parameter does not apply to the js function, only the Python function associated with the event listener.

Here is an example of a sine curve that updates every second!

import math
import gradio as gr
import plotly.express as px
import numpy as np


plot_end = 2 * math.pi


def get_plot(period=1):
    global plot_end
    x = np.arange(plot_end - 2 * math.pi, plot_end, 0.02)
    y = np.sin(2*math.pi*period * x)
    fig = px.line(x=x, y=y)
    plot_end += 2 * math.pi
    if plot_end > 1000:
        plot_end = 2 * math.pi
    return fig


with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            gr.Markdown("Change the value of the slider to automatically update the plot")
            period = gr.Slider(label="Period of plot", value=1, minimum=0, maximum=10, step=1)
            plot = gr.Plot(label="Plot (updates every half second)")

    dep = demo.load(get_plot, None, plot, every=1)
    period.change(get_plot, period, plot, every=1, cancels=[dep])


if __name__ == "__main__":
    demo.queue().launch()

Gathering Event Data

You can gather specific data about an event by adding the associated event data class as a type hint to an argument in the event listener function.

For example, event data for .select() can be type hinted by a gradio.SelectData argument. This event is triggered when a user selects some part of the triggering component, and the event data includes information about what the user specifically selected. If a user selected a specific word in a Textbox, a specific image in a Gallery, or a specific cell in a DataFrame, the event data argument would contain information about the specific selection.

In the 2 player tic-tac-toe demo below, a user can select a cell in the DataFrame to make a move. The event data argument contains information about the specific cell that was selected. We can first check to see if the cell is empty, and then update the cell with the user’s move.

import gradio as gr

with gr.Blocks() as demo:
    turn = gr.Textbox("X", interactive=False, label="Turn")
    board = gr.Dataframe(value=[["", "", ""]] * 3, interactive=False, type="array")

    def place(board, turn, evt: gr.SelectData):
        if evt.value:
            return board, turn
        board[evt.index[0]][evt.index[1]] = turn
        turn = "O" if turn == "X" else "X"
        return board, turn

    board.select(place, [board, turn], [board, turn], show_progress="hidden")

demo.launch()

Binding Multiple Triggers to a Function

Often times, you may want to bind multiple triggers to the same function. For example, you may want to allow a user to click a submit button, or press enter to submit a form. You can do this using the gr.on method and passing a list of triggers to the trigger.

import gradio as gr

with gr.Blocks() as demo:
    name = gr.Textbox(label="Name")
    output = gr.Textbox(label="Output Box")
    greet_btn = gr.Button("Greet")
    trigger = gr.Textbox(label="Trigger Box")
    trigger2 = gr.Textbox(label="Trigger Box")

    def greet(name, evt_data: gr.EventData):
        return "Hello " + name + "!", evt_data.target.__class__.__name__
    
    def clear_name(evt_data: gr.EventData):
        return "", evt_data.target.__class__.__name__
    
    gr.on(
        triggers=[name.submit, greet_btn.click],
        fn=greet,
        inputs=name,
        outputs=[output, trigger],
    ).then(clear_name, outputs=[name, trigger2])


demo.launch()

You can use decorator syntax as well:

import gradio as gr

with gr.Blocks() as demo:
    name = gr.Textbox(label="Name")
    output = gr.Textbox(label="Output Box")
    greet_btn = gr.Button("Greet")

    @gr.on(triggers=[name.submit, greet_btn.click], inputs=name, outputs=output)
    def greet(name):
        return "Hello " + name + "!"


demo.launch()

You can use gr.on to create “live” events by binding to the change event of components that implement it. If you do not specify any triggers, the function will automatically bind to all change event of all input components that include a change event (for example gr.Textbox has a change event whereas gr.Button does not).

import gradio as gr

with gr.Blocks() as demo:
    with gr.Row():
        num1 = gr.Slider(1, 10)
        num2 = gr.Slider(1, 10)
        num3 = gr.Slider(1, 10)
    output = gr.Number(label="Sum")

    @gr.on(inputs=[num1, num2, num3], outputs=output)
    def sum(a, b, c):
        return a + b + c


demo.launch()

You can follow gr.on with .then, just like any regular event listener. This handy method should save you from having to write a lot of repetitive code!