I was able to run the start_app function in the background but only when I moved start_app function into a separate py file, imported and run it like this:
import mock_service_app
import multiprocessing as mp
proc = mp.Process(target=mock_service_app.start_app, args=())
proc.start()
proc = mp.Process(target=start_app, args=(), initializer=mute)
are there any working examples of how to use startup/shutdown described in:
https://pgjones.gitlab.io/quart/how_to_guides/startup_shutdown.html
?
https://pgjones.gitlab.io/quart/how_to_guides/startup_shutdown.html
?
@povar81 I just pushed an project where I used before and after serving as a global extension: https://github.com/openbikebox/open-bike-box-connect/blob/master/webapp/common/amqp_queue.py . Perhaps this is an inspiration, too? :)
Uh, and this is also a solution to @PenguinMaster0226 s queue issues: https://github.com/openbikebox/open-bike-box-connect/blob/master/webapp/websocket_connect/WebsocketConnect.py
@pgjones thanks for the hint! The
configure_error_handlers is somehow a standard block which is in all my Flask / Quart projects, but without jsonify it's even more beautiful. Does it respect the JSON encoder I set here without jsonify: https://github.com/openbikebox/connect/blob/master/webapp/common/filter.py#L13 ?
hi, how do i use sqlite3 or ORM like peewee (sqlite3) with quart?
was looking @ https://aiosqlite.omnilib.dev/en/stable/ but couldn't find any example
Hi @shantanoo:matrix.org there isn't anything particularly special you need to do. Which aspects are you unsure about? Here is an sqlite example and pgjones/quart#35 is a good discussion.
5 replies
hey everyone!
hope you are all well
I am trying to learn how to make an ML-based web app using Quart.
I want to build an application that can handle prediction of a batch of images. I want to divide the batch into small chunks, make predictions and return the response to the user. This process should be done until the last chunk
I thought of making an async recursive api call. All I want to do is to make it recursively return the response.
Any help is appreciated.
This is the function that I want to call it recursively.
@app.route('/predict', methods=['POST'])
async def predict():
if request.method == 'POST':
has_files = (await request.files).get("file", None)
if has_files:
files = (await request.files).getlist("file")
responses= {}
# split files into small chunks to place here !
for file, i in zip(files, range(len(files))):
img_bytes = file.read()
class_id, class_name = await get_prediction(model=model,
image_bytes=img_bytes,
imagenet_class_index= imagenet_class_index)
responses[i] = {'class_id': class_id, 'class_name': class_name}
print(responses[i])
return await render_template("results.html", text=str(responses))
else:
return await render_template("results.html", text="please upload image!")
Would something using Server Sent Events, SSE, be what you need? https://gitlab.com/pgjones/quart/-/tree/master/examples/broadcast
Hello, I would like to ask how can I make my Quart API await for API Calls before opening new connections? Cause every time a request is sent to the Quart app, it processes it and opens a new connection after it's done automatically, even if there is no API calls which leads to timeout after 60s. Thank you.
I have previously opened a GitLab issue: pgjones/quart#409
@pgjones when using websocket.close() from latest quart (from master branch), i get an error because quart is trying to close the websocket again in handle_websocket
I made a hotfix for hypercorn that forces only one websocket.close event to be executed, ignoring the rest. I did this in hypercorn because quart is not aware of the websocket state, and so it can not know if the websocket has closed (or is in the process of doing so), does this make sense? Is it correct to MR this fix?
Another question: Is hypercorn going to support HTTP&Websockets ASGI 2.3? reason is not a blocker on my webapp development, but it is wanted in the future (this would also mean support for 2.2, since hypercorn is currently on 2.1)
trio's nurseries break and stop accepting tasks, HTTP requests time out without logging anything to the server
and so on
what am I doing wrong?
I imagine there is a way to catch all exceptions by default and show them to the user and go on
this is the code (if it matters, probably not): https://github.com/lnbits/lnbits
@lun-4 I think that is a good solution, but I'll check pgjones/quart#410.
Yep, Hypercorn will support 2.3 when I next release it
Yep, Hypercorn will support 2.3 when I next release it
hypercorn[497166]: [2021-05-03 18:07:08 +0000] [497166] [ERROR] Error in ASGI Framework
hypercorn[497166]: Traceback (most recent call last):
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/hypercorn/trio/context.py", line 39, in _handle
hypercorn[497166]: await invoke_asgi(app, scope, receive, send)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/hypercorn/utils.py", line 239, in invoke_asgi
hypercorn[497166]: await app(scope, receive, send)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart/app.py", line 2117, in __call__
hypercorn[497166]: await self.asgi_app(scope, receive, send)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart/app.py", line 2140, in asgi_app
hypercorn[497166]: await asgi_handler(receive, send)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/asgi.py", line 18, in __call__
hypercorn[497166]: nursery.start_soon(self.handle_request, nursery, request, send)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/trio/_core/_run.py", line 741, in __aexit__
hypercorn[497166]: raise combined_error_from_nursery
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/asgi.py", line 29, in handle_request
hypercorn[497166]: response = await self.app.handle_request(request)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/app.py", line 135, in handle_request
hypercorn[497166]: return await self.handle_exception(error)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart/app.py", line 1118, in handle_exception
hypercorn[497166]: raise error
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/app.py", line 131, in handle_request
hypercorn[497166]: return await self.full_dispatch_request(request_context)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/app.py", line 153, in full_dispatch_request
hypercorn[497166]: result = await self.handle_user_exception(error)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/app.py", line 166, in handle_user_exception
hypercorn[497166]: return await super().handle_user_exception(error)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart/app.py", line 1104, in handle_user_exception
hypercorn[497166]: raise error
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart_trio/app.py", line 151, in full_dispatch_request
hypercorn[497166]: result = await self.dispatch_request(request_context)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/quart/app.py", line 1933, in dispatch_request
hypercorn[497166]: return await handler(**request_.view_args)
hypercorn[497166]: File "/root/lnbits/lnbits/decorators.py", line 28, in wrapped_view
hypercorn[497166]: return await view(**kwargs)
hypercorn[497166]: File "/root/lnbits/lnbits/core/views/api.py", line 313, in api_payments_sse
hypercorn[497166]: g.nursery.start_soon(payment_received)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/trio/_core/_run.py", line 912, in start_soon
hypercorn[497166]: GLOBAL_RUN_CONTEXT.runner.spawn_impl(async_fn, args, self, name)
hypercorn[497166]: File "/root/lnbits/venv/lib/python3.8/site-packages/trio/_core/_run.py", line 1357, in spawn_impl
hypercorn[497166]: raise RuntimeError("Nursery is closed to new arrivals")
hypercorn[497166]: RuntimeError: Nursery is closed to new arrivals
@pgjones this was the last that we had. it came out of nowhere. it happens in the SSE endpoint handler which uses a global nursery assigned to
g. and once it appeared first then for the next 24 hours it kept repeating so the SSE endpoint didn't work (although the rest of the app seemed to work)
the other times it was different, but maybe related. some other small error in our code (could be anything, like a
KeyError when accessing a dict item, for example) would cause the app to stop responding, but not crash, and no more error would show up in the logs ever. the only thing that would show up in the logs are these exact same nursery errors, always at this exact SSE handler.
I was thinking of catching this
RuntimeError and exiting the app when it happened so it could be restarted, but that sounded like a terrible hack, that's why I'm asking here what is the proper way to do it
is that the recommended way? to assign the nursery to
app?
actually
app.nursery already exists by default
but
app is not exposed anywhere because someone thought that was best practices so I can't access it from my handlers
should I just ignore that and use a global
app?
do the nurseries get passed to blueprints too?