Essential Django packages ❤️

Django is sometimes called the batteries-included web framework, but there are a few third-party packages I tend to install for every Django project. I have categorized the packages for different use-cases, although one important note is that I am primarily thinking about side projects and websites with reasonable traffic requirements (i.e. 99% of the websites ever created). I personally have a few sites that have survived the Hacker News home page with most of this stack, so you'll also probably be just fine (and it turns out that judicious use of caching with redis can handle a lot of traffic). 😉

Thanks to Will Vincent for publishing 20 Django Packages That I Use in Every Project which directly inspired this post -- we have a few packages which overlap, but of course there are some differences as well. Also, if you need more awesomeness there is always awesomedjango.org or... if you want to see the whole breadth of options check out djangopackages.org!

I have tried to include links to the project sponsor/donation page if I could find it (please let me know if I missed any). One way to support the libraries you use is financially. Every 6 months or so, I document what I personally sponsor and I encourage you to as well if it's within your means and budget. Or better yet, get your workplace to sponsor the open source software they rely on!

And finally, just a note: lists like this are inherently subjective -- all libraries have their own trade-offs and nothing is perfect for everyone. I built up this list over years of trying things and seeing what worked for me and what I liked. I humbly suggest you come up with your own list, write it up, and let me know -- I'd love to read it and see what packages I'm missing out on!

ruff

I remember maybe 8 years ago when a co-worker tried to convince our team to use yapf and I was dead-set against it. I do not remember the actual conversation, just that I thought it was a horrible idea. I probably said something like, "Why would I want something to change my code automatically?!" or "I don't agree with this one formatting rule, therefore it must be bad."

Boy, has my opinion changed. At first, isort opened my eyes to the wonder of not manually alphabetizing all of my code imports. Then, black came along (with its delightful lack of settings) and I realized that I was now willing to trade "my personal code style" for "never thinking about code formatting ever again". Especially when working in a team where everyone tends to have their "own style".

Then, ruff was released and eventually replaced my linter, black, and isort with lightning fast speed. To be fair, I do not love that ruff is built in Rust and that Astral is VC-backed, but 1) it's not like I ever contributed to the Python code bases of any of the tools it replaced, so pretty sure that point is moot (at least for me personally), and 2) the code is MIT licensed -- not sure what more I could ask for from the team at Astral.

pytest-django

I appreciate that Django has unit testing built-in. However, I do not use it because I have completely converted to the church of pytest. The lack of required classes is appealing and fixtures are wonderful (even if they are sometimes a little too magical 🪄). I appreciate the fine-grained control over the database, especially preventing migrations from running and all of the Django-specific fixtures.

pytest-cov

Integrates pytest with coverage to tell you how much of your code is covered by unit tests. Very straight-forward and useful. I tend to have a separate command for generating coverage reports so it doesn't slow down my normal red/green/refactor test cycles.

Core Django

If I could wave a magic wand, I wish at least a little bit of the functionality of these packages would be incorporated into Django. I (helpfully?) rant about making Django more "batteries-included" in Django Roadmap 2024.

python-dotenv

I tend to reach for python-dotenv although I have also used django-environs for this functionality. There are a bunch out there, but all I really want is a way to load an .env file into Django settings (at least in development). Follow twelve-factor and keep secrets out of your code.

django-cache-memoize

This is a function decorator that integrates with the core Django cache. There are a lot of cache decorator libraries floating around, but I have run into a few gotchas with them. django-cache-memoize works exactly as expected.

django-model-utils

Sponsor django-model-utils ❤️

This is very silly, but I literally install this package for one particular model mixin: TimeStampedModel. I use it everywhere -- basically any model that might be useful to know when it got last updated. Insanely useful for lightweight audit tracking of a model.

whitenoise

Stop messing with S3 buckets and the atrocious AWS console. whitenoise will set the correct cache headers on your static assets automagically and a CDN like Cloudflare or Fastly will serve your assets without ever hitting your Python app. Even without a CDN, an NGINX sidecar to serve static assets will greatly reduce the amount of load for your WSGI server.

Server-side rendering

django-compressor

Compress JavaScript or CSS for production with hashed filenames (which breaks the long-term cache when used with whitenoise). There are a lot of settings, but I just copy and paste the same configuration into every project now and it works perfectly. Pair it with django-libsass to support SASS files without an extra build step.

Sponsor django-widget-tweaks ❤️

Very underrated in my opinion, django-widget-tweaks provides a few template tags to help render forms and fields in HTML. Usually when working with designers or more front-end folks, this allows them to customize HTML without touching Python or understanding the form rendering in Django.

django-allauth

Sponsor pennersr on GitHub ❤️

The premier authentication provider for integrating Django with third-party logins like Google, Facebook, LinkedIn, etc. The number of providers is staggering and always growing. pennersr does a great job of handling the myriad support requests for such an integral part of the ecosystem. django-allauth also has an exciting roadmap for 2024 which makes me hopeful it can continue to solve Django authentication needs far into the future.

API

Django seems to be used just as an API and paired with a frontend framework frequently these days. There are some libraries that I have used to provide APIs in my Django projects.

django-ninja

Sponsor django-ninja on BuyMeACoffee ☕️

django-rest-framework is the 800-pound gorilla when building a REST framework with Django. It has lots of documentation and lots of supporters. Personally, I also find it a little cumbersome to use and grok. I have only used django-ninja for two projects, but I really enjoyed its approach, especially if you have ever used FastAPI or are a fan of Python typing. I hope it continues to evolve and provide an alternative to DRF.

strawberry

Sponsor strawberry-graphql on GitHub ❤️

Similar to django-ninja, strawberry uses Python types to provide a GraphQL interface. I have also only used it for a few projects, but I was very happy with the developer experience (especially compared to Apollo).

Background tasks

django-q2

I have tried celery, django-rq, and huey and have always returned back to django-q2 when I need a background task runner for long-running tasks. My favorite features are the integration with Django admin, the ability to use either redis or a database to store tasks, and a cron-like scheduler.

Running in production

These are packages that I always use on production. They are all battle-tested, stable, fast, and boring technology (in the best possible sense of the phrase).

gunicorn

My go-to WSGI server for production. Extremely stable and when paired with NGINX for static assets it can handle a decent amount of traffic. I have seen benchmarks that other servers might be faster, but gunicorn is rock-solid and configuration is simple -- exactly what I'm looking for.

Psycopg 2

Sponsor dvarrazzo on GitHub ❤️

Django has support for other relational databases, but I switched from MySQL to PostgreSQL a long time ago and never looked back. However, I wouldn't ever want to operate any database in production (especially not PostgreSQL!). I suggest finding a cloud provider who has a managed PostgreSQL offering. It's worth it.

Psycopg 3 is now released, but I have not used it in production... yet.

redis + hiredis

redis is fast, it's got minimal resource requirements, and it's my go-to cache. As opposed to PostgreSQL I do run my own redis in production because I treat the data as ephemeral. If my lack of DevOps-fu breaks it, it shouldn't matter.

Note: as of Django 4, support for redis cache is included in Django. If you are in a previous version (or need some more advanced features), django-redis would also be needed.

Sentry

Sentry was originally built on top of Django and still has the best error reporting and monitoring I've ever seen for the web framework (although they support a dizzying amount of other frameworks as well).

Unfortunately, the free tier is not all that useful with any sort of traffic so I end up paying $26/month which is slightly annoying. I did self-host GlitchTip for a brief amount of time, but it wasn't worth the effort personally. However, I have gotten very used to the details available in Sentry and it's challenging for me to troubleshoot production problems without it now!

Will Vincent's 20 Django Packages That I Use in Every Project

Big thanks to Sangeeta Jadoonanan for proof-reading this article and giving me some ✨super✨ helpful feedback.

Hi, I'm Adam 👋

I've been a backend programmer for ~20 years in a variety of different languages before I discovered Python 10 years ago and never looked back. alldjango includes all the hard-won experience I've gained over the years building production-scale Django websites.