gettext-tools/doc/msgpre.texi - master - niceSOFT/gettext

gettext-tools/doc/msgpre.texi
245 строк7 KB
Bruno Haible
New program 'msgpre'.
29 дек 2025, 21:51
29 дек 2025, 21:51e549b33
@c This file is part of the GNU gettext manual.
@c Copyright (C) 1995-2025 Free Software Foundation, Inc.
@c See the file gettext.texi for copying conditions.

@pindex msgpre
@cindex @code{msgpre} program, usage
@example
msgpre [@var{option}...]
@end example

@cindex pretranslate a message catalog
@cindex translate through a Large Language Model
The @code{msgpre} program pretranslates a translation catalog.

@strong{Warning:} The pretranslations might not be what you expect.
They might be of the wrong form, be of poor quality, or reflect some biases.

@subsection Input file location

@table @samp
@item -i @var{inputfile}
@itemx --input=@var{inputfile}
@opindex -i@r{, @code{msgpre} option}
@opindex --input@r{, @code{msgpre} option}
Input PO file.

@item -D @var{directory}
@itemx --directory=@var{directory}
@opindex -D@r{, @code{msgpre} option}
@opindex --directory@r{, @code{msgpre} option}
Add @var{directory} to the list of directories.  Source files are
searched relative to this list of directories.  The resulting @file{.po}
file will be written relative to the current directory, though.

@end table

If no @var{inputfile} is given or if it is @samp{-}, standard input is read.

@subsection Output file location

@table @samp
@item -o @var{file}
@itemx --output-file=@var{file}
@opindex -o@r{, @code{msgpre} option}
@opindex --output-file@r{, @code{msgpre} option}
Write output to specified file.

@end table

The results are written to standard output if no output file is specified
or if it is @samp{-}.

@subsection Message selection

@table @samp
@item --keep-fuzzy
@opindex --keep-fuzzy@r{, @code{msgpre} option}
Keep fuzzy messages unmodified.
Pretranslate only untranslated messages.

@end table

@subsection Large Language Model (LLM) options

@table @samp
@item --species=@var{type}
@opindex --species@r{, @code{msgpre} option}
Specifies the type of Large Language Model execution engine.
The default and only valid value is @code{ollama}.

@item --url=@var{url}
@opindex --url@r{, @code{msgpre} option}
Specifies the URL of the server that runs Large Language Model execution engine.
For @code{ollama}, the default is @code{http://localhost:11434}.

@item -m @var{model}
@itemx --model=@var{model}
@opindex -m@r{, @code{msgpre} option}
@opindex --model@r{, @code{msgpre} option}
Specifies the model to use.
This option is mandatory; no default exists.
The specified model must
already be installed in the Large Language Model execution engine.

@item --prompt=@var{text}
@opindex --prompt@r{, @code{msgpre} option}
Specifies the prompt to use before each @code{msgid} from the PO file.
It allows you to specify extra instructions for the LLM.
The prompt should include an instruction like
"Translate into @var{target language}.".
Some hints for good prompts are described in the article
``How to write AI prompts for translation''
@url{https://poeditor.com/blog/ai-prompts-for-translation/}.

@item --postprocess=@var{command}
@opindex --postprocess@r{, @code{msgpre} option}
Specifies a command to post-process the output from the LLM.
This should be a Bourne shell command
that reads from standard input and writes to standard output.

For instance, the @code{ministral-3:14b} model
often emphasizes part of the output with @samp{**} characters.
To eliminate these markers,
you could use the command @samp{sed -e 's/[*][*]//g'}.

@end table

@subsection Input file syntax

@table @samp
@item -P
@itemx --properties-input
@opindex -P@r{, @code{msgpre} option}
@opindex --properties-input@r{, @code{msgpre} option}
Assume the input file is a Java ResourceBundle in Java @code{.properties}
syntax, not in PO file syntax.

@item --stringtable-input
@opindex --stringtable-input@r{, @code{msgpre} option}
Assume the input file is a NeXTstep/GNUstep localized resource file in
@code{.strings} syntax, not in PO file syntax.

@end table

@subsection Output details

@table @samp
@item --color
@itemx --color=@var{when}
@opindex --color@r{, @code{msgpre} option}
Specify whether or when to use colors and other text attributes.
See @ref{The --color option} for details.

@item --style=@var{style_file}
@opindex --style@r{, @code{msgpre} option}
Specify the CSS style rule file to use for @code{--color}.
See @ref{The --style option} for details.

@item --force-po
@opindex --force-po@r{, @code{msgpre} option}
Always write an output file even if it contains no message.

@item --indent
@opindex --indent@r{, @code{msgpre} option}
Write the .po file using indented style.

@item --no-location
@opindex --no-location@r{, @code{msgpre} option}
Do not write @samp{#: @var{filename}:@var{line}} lines.

@item -n
@itemx --add-location=@var{type}
@opindex --add-location@r{, @code{msgpre} option}
Generate @samp{#: @var{filename}:@var{line}} lines (default).

The optional @var{type} can be either @samp{full}, @samp{file}, or
@samp{never}.  If it is not given or @samp{full}, it generates the
lines with both file name and line number.  If it is @samp{file}, the
line number part is omitted.  If it is @samp{never}, it completely
suppresses the lines (same as @code{--no-location}).

@item --strict
@opindex --strict@r{, @code{msgpre} option}
Write out a strict Uniforum conforming PO file.  Note that this
Uniforum format should be avoided because it doesn't support the
GNU extensions.

@item -p
@itemx --properties-output
@opindex -p@r{, @code{msgpre} option}
@opindex --properties-output@r{, @code{msgpre} option}
Write out a Java ResourceBundle in Java @code{.properties} syntax.  Note
that this file format doesn't support plural forms and silently drops
obsolete messages.

@item --stringtable-output
@opindex --stringtable-output@r{, @code{msgpre} option}
Write out a NeXTstep/GNUstep localized resource file in @code{.strings} syntax.
Note that this file format doesn't support plural forms.

@item -w @var{number}
@itemx --width=@var{number}
@opindex -w@r{, @code{msgpre} option}
@opindex --width@r{, @code{msgpre} option}
Set the output page width.  Long strings in the output files will be
split across multiple lines in order to ensure that each line's width
(= number of screen columns) is less or equal to the given @var{number}.

@item --no-wrap
@opindex --no-wrap@r{, @code{msgpre} option}
Do not break long message lines.  Message lines whose width exceeds the
output page width will not be split into several lines.  Only file reference
lines which are wider than the output page width will be split.

@item -s
@itemx --sort-output
@opindex -s@r{, @code{msgpre} option}
@opindex --sort-output@r{, @code{msgpre} option}
Generate sorted output.  Note that using this option makes it much harder
for the translator to understand each message's context.

@item -F
@itemx --sort-by-file
@opindex -F@r{, @code{msgpre} option}
@opindex --sort-by-file@r{, @code{msgpre} option}
Sort output by file location.

@end table

@subsection Informative output

@table @samp
@item -h
@itemx --help
@opindex -h@r{, @code{msgpre} option}
@opindex --help@r{, @code{msgpre} option}
Display this help and exit.

@item -V
@itemx --version
@opindex -V@r{, @code{msgpre} option}
@opindex --version@r{, @code{msgpre} option}
Output version information and exit.

@item -q
@itemx --quiet
@itemx --silent
@opindex -q@r{, @code{msgpre} option}
@opindex --quiet@r{, @code{msgpre} option}
@opindex --silent@r{, @code{msgpre} option}
Suppress progress indicators.

@end table

@subsection Examples

To pretranslate the file @code{foo.po}:

@smallexample
msgpre --model=ministral-3:14b < foo.po > foo-pretranslated.po
@end smallexample

@noindent
Note that this command can take a long time,
depending on the model and the available hardware.