How does Flask streaming simulate real-time response of ChatGPT?

Simulate ChatGPT real-time response using Flask streaming

Many applications, such as real-time chats that simulate ChatGPT or large file downloads, need to generate and transmit data while avoiding long waits on the client. This article demonstrates how to implement this streaming in the Python Flask framework and corrects flaws in the original code.

The original code tried to use yield to implement streaming, but since the response object returned only after the generate() function ended, the browser must wait for all data to be generated before the content is displayed, which does not match the real-time response expectations.

Problem code:

 from time import sleep
from flask import Flask, Response, stream_with_context

app = Flask(__name__)

@app.route('/stream', methods=['GET'])
def stream():
    def generate():
        for i in range(1, 21):
            print(i)
            yield f'this is item {i}\n'
            sleep(0.5)

    return Response(generate(), mimetype='text/plain')


if __name__ == '__main__':
    app.run(debug=True)

Workaround: Use Flask's stream_with_context decorator correctly. This decorator ensures that data is returned to the client immediately every time yield is generated, enabling true streaming. Improved code:

 from flask import stream_with_context, request, jsonify

@app.route('/stream')
def streamed_response():
    def generate():
        yield 'Hello'
        yield request.args.get('name', 'World') # Use get() to avoid KeyError
        yield '!'
    return jsonify({'message': list(stream_with_context(generate()))}) # Return to JSON format

stream_with_context wraps the generate function, causing data to be sent immediately every yield . In the example, data generation is simple. In actual applications, generate function may contain more complex logic (such as database queries or complex calculations), but the function of stream_with_context is still to ensure timely transmission of data. request.args.get('name', 'World') obtains data from request parameters, implements more flexible streaming, and uses the get() method to deal with missing parameters to avoid KeyError errors. Finally, using jsonify to wrap the result into JSON format, which is more suitable for front-end processing.

Through the above improvements, the real-time response effect of ChatGPT can be effectively simulated.

The above is the detailed content of How does Flask streaming simulate real-time response of ChatGPT?. For more information, please follow other related articles on the PHP Chinese website!