Streaming response

Quart supports responses that are meant to be streamed to the client, rather than received in one block. If you are interested in streaming the request data see Consuming the request body or for duplex streaming see Using websockets.

To stream a response the view-function should return an asynchronous generator that yields bytes. This generator can be returned with a status code and headers as normal. For example to stream the time every second,

async def stream_time():
    async def async_generator():
        time = datetime.isoformat()
        yield time.encode()
    return async_generator(), 200, {'X-Something': 'value'}


Quart by default will timeout long responses to protect against possible denial of service attacks, see Denial Of Service mitigations. This may be undesired for streaming responses, e.g. an indefinite stream. The timeout can be disabled globally, however this could make other routes DOS vulnerable, therefore the recommendation is to set the timeout attribute on a specific response to None,

from quart import make_response

async def stream_time():
    response = await make_response(async_generator())
    response.timeout = None  # No timeout for this route
    return response

Server Sent Events

See the Tutorial: Broadcast Server Side Events for details on how to utilise server sent events.


The test client get() and associated methods will collate the entire streamed response. If you want to test that the route actually streams the response, or to test routes that stream until the client disconnects you will need to use the request() method,

async def test_stream() -> None:
    test_client = app.test_client()
    async with test_client.request(..) as connection:
        data = await connection.receive()
        assert data ...
        assert connection.status_code == 200
        await connection.disconnect()  # For infinite streams