What language and architecture should I use for a weighing-scale inventory system?

If you use FTP instead of TCP, you could automate the process with a script: download the scale's file, transform it into a reliable CSV or another database-friendly format, and import it into the database. The exact transformation depends on whether the scale produces plain text, CSV, JSON, or a proprietary format.

Python would be a practical choice for this kind of project. It has built-in support for TCP connections and FTP, SQLite is easy to use for local storage, and libraries are available for writing Excel files. That said, the best language is usually the one you already know well enough to maintain. C++ and MySQL can absolutely handle the job if you are more comfortable with them.

A reasonable design would be a small service that listens for readings from the scale, validates and converts the incoming data, then writes each accepted measurement to MySQL. Keep the raw readings and inventory calculations separate if possible, so you can audit or reprocess data later. Generate the Excel files as an export step rather than using spreadsheets as the primary database.

First clarify exactly how the scale sends its data. If TCP provides live readings, that is probably a better fit for real-time inventory updates than periodically downloading files over FTP. FTP generally means the scale creates a file, your program retrieves it, and then processes it, which is more suitable for batch imports or daily measurement logs.