Tested guides for running your own stack
Self-hosted AI, Docker, Nginx and databases on a small Linux server, with the commands, the real output and the machine they ran on.
First guide, in review
- Tested
- 2026-11-17
- Server
- Hetzner cx33, 4 vCPU, 7.6 GiB, no GPU
- OS
- Ubuntu 24.04.4 LTS
- Runtime
- Ollama 0.40.16
- Result
- llama3.2:1b at 15.87 tok/s
- Run on a real server first
- Date, OS, hardware and versions logged
- Untested steps are marked
Learning path
Nine guides in order, from a first container to a stack you can monitor and repair.
Start here
A first container, a local model, and how to read what your server is doing.
- Install Docker Compose and run your first stack Planned
- Run an LLM locally with Ollama on Linux In review
- Read top output on Linux Planned
Make it yours
Serve a chat UI through Nginx with TLS, and back up PostgreSQL.
- Install Open WebUI and connect it to Ollama Planned
- Use Nginx as a reverse proxy with TLS Planned
- Back up PostgreSQL with pg_dump and restore it Planned
Keep it running
Manage and monitor the stack, then harden the server.
- Install Portainer and secure it Planned
- Install Grafana and Prometheus with Docker Compose Planned
- Harden a fresh Linux server: SSH, firewall, updates Planned
Each guide has three levels: Naked eye to get it running, Telescope to understand it, Deep field to go further.
Six hubs, one stack
Each hub covers one layer of a server you run yourself.
- LinuxThe everyday server skills every guide assumes: reading system output, users, and updates.
- DockerContainers and Compose stacks that start cleanly, keep their data, and stay within their disk budget.
- NginxReverse proxies and TLS so your services answer on one domain with valid certificates.
- DatabasesInstall, back up, and check the databases your stack depends on, starting with PostgreSQL.
- Self-hosted AIRun language models, chat UIs and AI gateways on your own hardware, with measured speed and memory use.
- ObservabilitySee what your stack is doing: resource monitors, logs and dashboards, and what the numbers mean.
Latest observation logs
- In review: Run an LLM locally with Ollama on Linux (CPU only, 8 GB) Tested 2026-11-17 Server: Hetzner cx33, 4 vCPU, 7.6 GiB, no GPU
Older archive: programming articles restored from the Internet Archive. These are not tested guides and are being cut down.
JavaScript, PHP, C++, Java, Python, C#, Android, ASP.NET, Golang, MySQL, Server, Linux, JSP, Oracle, iOS, Redis, MongoDB, Swift, PostgreSQL, Nginx
- Summary of data interaction between Android client and server
- Implementation of response compression in ASP. NET Core
- The usage and conversion of Array and ArrayList in C
- Details of the exception handling mechanism in C++
- Go language implementation of genetic algorithm example code
- Summary of the Method of iOS Traversing Sets of NSArray NSDictionary NSSet
- About the Java collection framework interview questions (with answers)
- Jquery $. Get of $. Post of $. Ajax of $. GetJSON of usage summary
- The optimization techniques for loop statements in JavaScript are discussed in depth
- Servlet dynamic web technical explanation
- Detailed Explanation of top Command Output in Linux
- mysql View Creation View (CREATE VIEW) and Usage Restriction Example Explanation
- Configuration of Nginx servers against CC attacks
- Difference analysis between Oracle hard parsing and soft parsing
- Related Object Oriented Questions in php Interview
- PostgreSQL tutorial of viii: index in detail
- Use Python to get all kinds of information about the Linux system
- Swift hello world! Swift quick start tutorial
- Swift tutorial and the structure of such details