# Command line Sometimes the question is not in your code at all. You are staring at a bucket listing, or a log line, or a colleague's message, and you want to know what `report.parq` is going to be served as without opening a REPL. The package ships a small console script for exactly that moment. # Install Install it as a tool so it is on your path everywhere. ``` bash uv tool install content-types # or pipx install content-types ``` Inside a project that already depends on the package the script is on the virtualenv's path, so `uv run content-types ...` works there without a separate install. # Use One argument in, one line out. ``` bash content-types example.jpg # image/jpeg content-types .webp # image/webp content-types data.parquet # application/vnd.apache.parquet ``` The argument goes straight through [get_content_type()](../reference/get_content_type.md#content_types.get_content_type) with its defaults, so the tool inherits the library's manners: case does not matter, a bare extension works with or without the dot, and an unknown extension prints `application/octet-stream` and exits 0. It does the forward lookup only. For the reverse direction, open Python. # Flags and exit codes The script is deliberately tiny, so this table is the whole contract. | Invocation | Output | Exit code | |----|----|----| | `content-types NAME` | the content type, or `application/octet-stream` for an unknown extension | 0 | | `content-types` | usage text on stdout | 1 | | `content-types -h` or `--help` | usage text | 0 | | `content-types --version` | the installed package version | 0 | Extra arguments after the first are ignored. There are no environment variables and no configuration file to discover. Under the hood the script is the function `content_types.cli()` wired up in `pyproject.toml`, which means you can call it yourself if you ever want the same behaviour from inside another tool.