# ProDevs

> ProDevs makes free news and tools for software developers, and takes on a few hands-on projects for one Docker host you already run.
> Guides are bylined ProDevs. Each guide says it was drafted with AI assistance, and opens with a dated how-we-test note: the host, the Docker Engine version, and what was measured. Talk2Term is a Zsh plugin on this site. It is not the company name.

## How we test

- [How we test](https://prodevs.in/tools/docker-volume-vs-bind-mount/): There is no separate methodology page. The note sits at the top of each guide. On the volume guide, the run was 29 September 2026 on Ubuntu 24.04.3 LTS with Docker Engine 28.5.1 (linux/amd64): one write per mount, then the containers were removed.

## Guides

- [Docker volume or bind mount: who has to see the files](https://prodevs.in/tools/docker-volume-vs-bind-mount/): Use a named volume when Docker should own the files, and a bind mount when a process on the host must read or edit that same path.
- [Idle Postgres in Docker: read the stats line, not shared_buffers](https://prodevs.in/tools/postgres-docker-idle-ram/): Cap the container, leave shared_buffers at the server default until you have a real workload, and size the host from docker stats under that workload. The idle line is only the empty-server floor.
- [Docker memory limits: set the cap, then prove the OOM](https://prodevs.in/tools/docker-memory-limit-oom/): Set --memory and --memory-swap to the same value, about 1.5× the steady docker stats reading for ordinary work (a starting cap, not a measurement of your app). If inspect shows OOMKilled true, raise both numbers together.
- [Docker or a VPS: they are not the same choice](https://prodevs.in/tools/docker-or-vps/): Install Docker Engine on the Linux VPS when the app should run in a container, and leave it off when packages on the host are enough.
- [Cloudflare vs a VPS: which bill you are actually choosing](https://prodevs.in/tools/cloudflare-vs-vps/): Cloudflare in front of the machine is a network layer. Workers is short request code. Buy a VPS when a process must stay running between requests.
- [VPS for AI coding agents: the RAM floor and the bill you are not paying](https://prodevs.in/tools/vps-for-ai-coding-agents/): Claude Code’s setup docs say 4 GB of RAM or more. Codex’s install doc says 4 GB minimum and 8 GB recommended. OpenClaw’s FAQ says 1 GB is the absolute minimum and 2 GB or more is recommended. The model bill is separate from the VPS invoice.
- [Self-Hosted GitHub Runners in 2026: What They Really Cost](https://prodevs.in/tutorials/self-hosted-github-runners-cost-2026/): Self-hosted GitHub Actions minutes cost $0 from GitHub as of 27 September 2026. A $0.002-per-minute fee announced on 16 December 2025 was postponed and is not in effect. You pay for the machine, its power, and your time.

## Products

- [Talk2Term](https://prodevs.in/talk2term/): Zsh plugin that turns plain English into shell commands. The page offers a free try at 20 uses a day.
- [Services](https://prodevs.in/services/): Three fixed jobs on one Linux host: a Docker memory-limit fix, Postgres-in-Docker right-sizing, or moving one app with Compose. Not for Kubernetes, choosing or buying a server, or a multi-app setup. Launch prices on that page run until 20 October 2026.