Command Line

pip install typeseg installs the PyPI package, including the typeseg command and its alias segcat. See the content types inside a file directly in your terminal, with colored text, a legend, and a table of segment ranges and confidence scores.

Recorded Mamba output from TypeSeg on PyPI. The historical screenshot’s banner reads textseg.precise; the current API call is typeseg.precise. Colors mark content types; the table shows ranges and confidence. Shell command text is redacted.

Inspect a file

Start with U-Net for fast local inspection:

typeseg --model fast file.txt

Use Mamba when full-file context matters more than throughput:

typeseg --model precise file.txt

typeseg file.txt also uses Mamba: precise is the CLI default. U-Net is selected explicitly with --model fast.

Read from standard input

Omit the filename to read piped text:

cat file.txt | typeseg --model fast

In PowerShell:

Get-Content -Raw file.txt | typeseg --model fast

The file argument is read as UTF-8, with replacement for invalid characters. For other encodings, decode the input first and retain the original byte mapping if you need byte-accurate annotations.

Demo, alias, and module invocation

typeseg --model fast --demo       # built-in mixed-text example
segcat --model fast file.txt      # alias of typeseg
python -m typeseg --model fast file.txt
typeseg --help

Understand the display

The output shows the active backend, input length, a content-type legend, highlighted source, and a table of character ranges, labels, and confidence. Range ends are excluded. Confidence describes content type, not maliciousness.

Text tinting is automatic in an interactive terminal. NO_COLOR disables automatic text tinting; TYPESEG_COLOR=always or TYPESEG_COLOR=never overrides it. The CLI also uses ANSI formatting for its legend and table.

For structured results and configurable post-processing, use the Python API. For the package reference, see TypeSeg on PyPI.