– Contextualise is a side project that scratches Brett’s own itch

– It uses a Topic map which is a type of graph model to model a lot of things

– A Topic map is a perfect fit for modeling any type of knowledge

– You can visualize a graph model by thinking about having circles connected by lines

– Graph databases can help you naturally model very complicated relationships

– Topic maps very much align with how you likely think about modeling data in real life

– Brett used a Topic map to build a CMS back in ~2006

– The current Topic map engine Brett built for this project came about in ~2016

– Both Topic DB and Contextualise are open source on GitHub

– Motivation for choosing Flask over Pyramid and Django

– Contextualise has been up and running since February 2020

– Google Analytics is used for very basic traffic metrics, not personalized tracking

– Handling file uploads using the code supplied in Flask’s documentation

– When files are written to disk, they go into a specific directory

– The application uses server side rendered Jinja templates with a bit of JavaScript

– Drawing out the graph map using the vis.js charting library

– There’s a few more visualization features coming to Contextualise soon

– Motivation for choosing Gunicorn and nginx over other app / web servers

– Brett doesn’t use Docker, but folks have contributed Docker support in his project

– It’s hosted on Linode using Ubuntu 18.04 with PostgreSQL and nginx w/ Let’s Encrypt

– Topic DB is an abstraction on top of PostgreSQL, not a dedicated graph database

– Topic DB uses the Psycopg2 library to connect to PostgreSQL

– The biggest problem with Psycopg2 was getting it running on MacOS

– On Windows you can use WSL, but ideally Docker solves this problem all together

– Let’s Encrypt’s Certbot tool was used to configure SSL certificates

– Everything is running on a single server with 2 GB of memory, 1 CPU core a 50 GB SSD

– Hearing stories about deploying everything to a single server is great

– About 200 users are signed up with 80+ active weekly users, the server is fine

– Linode makes it easy to vertically scale up if needed

– The server was set up by hand while following a bunch of DigitalOcean guides

– You can restart Gunicorn in about 500 milliseconds when not using Docker

– Brett breaks down a custom Python script he uses to deploy things

– Besides handling SSL certificates, nginx also serves static files

– Dealing with secrets using an INI file that’s not commit to version control

– An environment file would be pretty similar

– Flask-Security-Too is used because the original Flask-Security was abandoned

– His app sends email out but he uses his ISP’s SMTP servers to send it

– Logging is done with Flask’s logger, along with periodic manual lookups for errors

– Nick chimes in with a way to get notified by email if a Flask error occurs

– Topic DB is well tested but the Flask app has no tests at the moment

– Backups are handled with Linode’s automated server backups

– Brett really likes Linode

– Linode has alarms you can set up to get notified if your server gets too busy

– There’s currently nothing hooked up to do a healthcheck on the site itself

– Best tips? Use tools you can be productive with & be mindful of your dependencies

– There’s only a handful of external dependencies in Contextualise

– 21 Python files and about 4,200 lines of code make up the Flask app

– The Topic DB project has 27 Python files with 3,000+ lines of code

– Python allows you to write fewer lines of code compared to some other languages

– You can find Brett on GitHub, Twitter and he has his own site too