Visualize FastAPI endpoints with FastAPI-Voyager
https://www.newsyeah.fun/voyager/
#HackerNews #FastAPI #FastAPIVoyager #APIVisualization #DeveloperTools
what's up, it's me, jonny #python developer. #fastAPI has a totally dysfunctional development process where all code has to route through a single point of control. look at the PR velocity on it and all related projects and see how many non-bot PRs have been merged in the past year. that tells the whole story.
it's really just a very simple wrapper around starlette, has a ton of perf bottlenecks that will never be resolved because of said dysfunctional dev process, and if i were starting a new project i would not use any code used by that whole group. there are better options, and we do indeed have to choose to starve the thing eating all the oxygen in the room
it's taken me 9 months to go from reporting an obvious bug to having the 7 lines of code it changed be usurped by a core dev and i dont wanna fight anymore
https://www.youtube.com/watch?v=Yj3WhcGoSA4
https://github.com/fastapi/fastapi/discussions/13380
https://github.com/fastapi/fastapi/pull/14283#issuecomment-3479009522
so after the almost yearlong process where i added all those tests because another maintainer asked me to add them, this PR does the same changes I made, except for worse, duplicating functionality across the different arms of the if form_input branches so that it's possible for those behaviors to diverge, and excluding the docs that explain the behavior. this is how fastAPI's contribution process works?
edit: i also, after explaining at length why the other changes were necessary for the feature to address the actual problem that was described, removed all the changes you described as unnecessary. and still that wasn't enough? this is really quite something.
sneakers-the-rat
on Feb 19
Author
" i don't want to fight anymore "
https://www.youtube.com/watch?v=Yj3WhcGoSA4
Fa temps que vaig llegint que Mastodon (el servidor) menja molts recursos de màquina i que hi ha altres programaris fediversals molt més lleugers. Cert, per exemple, Mastodon no el pots instal·lar en una Raspberry Pi 3B+, però sí que pots posar-li Pleroma o Akkoma.
El que em fa sentir orgullós és que el meu propi servidor que estic programant amb #fastAPI ( #Python), #appy, és encara més lleuger que Pleroma o Akkoma.
En el mateix servidor, Pleroma consumeix 2,5% de CPU, Akkoma 1,5% i appy 0%
I això que el servidor Pleroma només té 1 usuari que segueix a tres usuaris remots, Akkoma el mateix, però el servidor appy té 2 usuaris amb 35 seguidors i 34 seguits.
A més usuaris remots seguits, més activitat rep el servidor i més feina té (més carrega).
He encertat plenament al triar #fastAPI i #Python per a desenvolupar un servidor ActivityPub.
Fa temps que vaig llegint que Mastodon (el servidor) menja molts recursos de màquina i que hi ha altres programaris fediversals molt més lleugers. Cert, per exemple, Mastodon no el pots instal·lar en una Raspberry Pi 3B+, però sí que pots posar-li Pleroma o Akkoma.
El que em fa sentir orgullós és que el meu propi servidor que estic programant amb #fastAPI ( #Python), #appy, és encara més lleuger que Pleroma o Akkoma.
En el mateix servidor, Pleroma consumeix 2,5% de CPU, Akkoma 1,5% i appy 0%
Oh! el meu projecte #appy ja té 8 estrelles i 3 forks. 😀
El desenvolupament del servidor #ActivityPub #appy ( :appy: ) no s'atura. Fa quatre dies vaig alliberar la versió v0.3.0:
https://codeberg.org/spla/appy/releases/tag/v0.3.0
- 3 servidors #appy necessiten 192+173+117 = 482 megues de RAM
- 1 servidor #Mastodon 322+300+290+290 = 1.202 megues de RAM
- 1 servidor #Pleroma (veure imatge) 491 megues de RAM
- 1 servidor #Akkoma (veure imatge) 395,3 de RAM
Tots els servidors tenen un sol usuari actiu.
#appy és la millor opció, clarament (#Python #fastAPI) 😀
consum de RAM d'un sol servidor Pleroma: 491 megues de RAM
consum de RAM d'un sol servidor Akkoma: 395,3 megues de RAM
consum de RAM d'un sol servidor Mastodon: 1,202 megues de RAM (1,2G)
consum de RAM de tres servidors appy: 482 megues de RAM
Over the past few days in the small moments when I've been able to work I've been experimenting with porting Hasher-Matcher-Actioner from Flask to FastAPI, and I now have it mostly working with an existing database.
It's been interesting working with python because I don't work with it much, but FastAPI still feels very much like Express in terms of how things are written — lots of building your own framework essentially, instead of "here's the given path", which can be daunting.
I still need to work out Alembic and migrations, but I have it working against an existing database with the same schema. Also need to figure out building CLI tools that accompany the server, where flask has that built-in
