Our tech stack

We self-host key components of our development environment and for use cases of our own and our clients’ applications, such as:

• VMs, containers, and deployment on PaaS alternatives.
• Securely encrypted cross-site backups.
• Large Language Models for LLM-enhanced features and code completion with most common models available, including mixtral-8x7b.
• Mesh networking.
• S3-compatible object storage, in a HA setup.
• Maps and (reverse) geocoding servers, in a HA setup.
• Collaborative task management using Kanban.
• Web-based collaboration on code using Git.
• Email testing.
• GDPR-compliant newsletter and mailing list management.
• GDPR-compliant analytics of websites and apps.
• File management and sharing within your team and with your customers.

We create UIs using the following libraries:

• Phoenix LiveView: our preferred can-do-it-all way of building rich web apps.
• htmx: for sprinkling interactivity on micro-frontends.
• NextJS: for complex web apps requiring the involvement of third parties in co-development.
• Tailwind and Bootstrap: either of those gets the job done.

We have experience in using the following pieces of software, preferably (whenever possible) paired with an SQLite database:

• NocoDB: for low-code setups with a ready-made REST API.
• Strapi: for Jamstack websites with Hugo.
• Wordpress with ACF: for complex web apps requiring a familiar UI for editors and admins.
• Grav CMS: when a database is unnecessary.
• Django with Django Admin: for medium-complexity web apps with a ready-baked admin UI.
• Django with Django Ninja and Django Admin: for REST APIs and CRUD with a ready-baked admin UI for data management.
• FastAPI: for rapid prototyping of REST APIs.
• Hugo: for Jamstack and static websites.
• Phoenix and Ecto with SQLite: for Elixir applications without a CMS.

We are competent in the following languages, depending on the purpose / use-case:

• Elixir: for everything .
• Go: for performant REST APIs.
• TypeScript: for UIs in NextJS, React Native or Sveltkit.
• JavaScript: for JS hooks in Phoenix LiveView web apps.
• Python: as a “glue” language, for REST APIs and mini-frontends with htmx , and for using packages for Machine Learning and I/O of specific file formats.
• PHP: for custom API routes when using Wordpress as a headless CMS.

Servers

OVERBRING S.M.P.C. operates its own hybrid cloud with:

• VPSs in a data center in Frankfurt,
• a mix of purpose-built, refurbished, and SBC servers, across different sites in Greece.

We make heavy use of Proxmox VE and its sister-project Proxmox Backup Server for incremental, cross-site backups and, sometimes, FreeBSD.

On bare metal or on VMs/containers we prefer Debian and Alpine Linux.

Everything is linked together over the yggdrasil IPv6 mesh network, for which we run multiple non-public peers across sites for redundancy.

We eschew using public, subscription-based PaaS offerings and instead prefer to use dokku to manage the lifecycle of applications from building to scaling.

Desktops and laptops

All of our infrastructure, including desktop PCs and laptops, exclusively runs Linux or BSD variants.

We abhor planned obsolescence and digital serfdom. Thus:

• We do not support or care to get into proprietary OSs.
• We have not paid the “Apple tax” in more than a decade.

LLMs and Generative AI

Some of our servers feature GPUs, APUs and CPUs with enough (V)RAM to run locally-hosted Large Language Models using Ollama .

We and our clients to benefit from rapid prototyping of LLM-enabled features at good-enough quality without “breaking the bank” through exploratory calls to proprietary subscription-based APIs.