Back to Repositories

Testing ChatInterface Streaming with Multimodal Support in gradio-app/gradio

This test suite validates the streaming functionality of Gradio’s ChatInterface component with multimodal tuple support. It implements a simple echo service that demonstrates progressive text generation and maintains a global run counter for verification purposes.

Test Coverage Overview

The test provides coverage for Gradio’s ChatInterface streaming capabilities with multimodal input handling.

Key areas tested include:

Progressive text streaming response generation
Multimodal message handling with tuple support
Global state management across chat interactions
Run counter reset functionality

Implementation Analysis

The implementation uses a straightforward approach with a global counter to track interaction runs. The slow_echo function demonstrates streaming by yielding progressive text responses character by character, while maintaining run count state. The test leverages Gradio’s ChatInterface with multimodal=True and type=’tuples’ configuration for enhanced message handling.

Notable patterns include:

Generator-based streaming response
Global state management for run tracking
Unload event handler for state reset

Technical Details

Testing components include:

Gradio ChatInterface with multimodal support
Blocks context for component rendering
Global variable for state management
Generator function for streaming responses
Unload event handler for cleanup

Best Practices Demonstrated

The test exemplifies several testing best practices in Gradio application development.

Notable practices include:

Isolation of test state through reset functionality
Clear separation of interface setup and launch logic
Proper cleanup handling with unload events
Progressive response generation for realistic chat simulation

gradio-app/gradio

demo/test_chatinterface_streaming_echo/multimodal_tuples_testcase.py

            
import gradio as gr

runs = 0

def reset_runs():
    global runs
    runs = 0

def slow_echo(message, history):
    global runs  # i didn't want to add state or anything to this demo
    runs = runs + 1
    for i in range(len(message['text'])):
        yield f"Run {runs} - You typed: " + message['text'][: i + 1]

chat = gr.ChatInterface(slow_echo, multimodal=True, type="tuples")

with gr.Blocks() as demo:
    chat.render()
    # We reset the global variable to minimize flakes
    # this works because CI runs only one test at at time
    # need to use gr.State if we want to parallelize this test
    # currently chatinterface does not support that
    demo.unload(reset_runs)

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