AIGenPipeline Release 2.1.1

This is the next public release of the AIGenPipeline, a tool supporting the use of AI to systematically create text files - code, documentation, you name it. It allows using the OpenAI chat completion API either via a script or via a Java programm, depending on the complexity of your tasks. For each task you provide a prompt file, input files and specify an output file, and the tool checks whether the outputfile has to be generated for the first time or needs to be regenerated because the prompt / input files have changed. For more documentation please consult the project website.

Changes

• Release 2.1.0 adds an option -wvf that allows keeping the version in an external file for brain dead file formats that don't allow comments - e.g. for foo.json we generate / read foo.json.version for checking whether the file needs to be regenerated.
• If you generate many files (e.g., hundreds) the startup time might become an issue. So I've added the option to use command files: if you give just one file argument then from this file it reads an arbitrary number of command lines separated by a blank line - see the homepage for more details.
• Release 2.1.1 adds support for the o1-* models of OpenAI: they do not allow a system message, so that's moved into a first user message (if either the model is named like o1-9? or option -ns / --nosysmsg is given) and instead of the deprecated max_tokens we use max_completion_tokens.

Download

You can either

• build it from the source locally,
• grab the zip from the command line package (on maven central / maven search )
• use the framework jar (on maven central)
  directly with the class AIGenerationTask .

What's Changed

• A couple of dependency bumps by @dependabot
• Introduce Command files by @stoerr in #52
• Add version files for version markers by @stoerr in #53

Full Changelog: aigenpipeline-2.0.1...aigenpipeline-2.1.1