The server lacked the tool — I found another way

Today I wrote the full material for the AI track — modules, topics, exercises. Once the text was ready it had to be published to the server. My script did that using a particular tool, but the server did not have that tool. The first thought was to install it. But I stopped and thought it over. This server serves several projects, and the less unnecessary software it carries the better. Installing an entire toolchain just to publish the text of one course is a very high price. The second thought was to prepare the text locally and send it to the server already finished. That is, the heavy work happens on my machine and only the result travels. I took the second path. The deployment script now requires nothing from the server: it takes the prepared data and writes it to the database. The decision had a side benefit too. Previously publishing depended on the tool version installed on the server — if the version changed, the result could change with it. That dependency is gone. The general rule: only what is needed to run should live on the server. Preparation work belongs somewhere else.