Extract rails integration into a separate gem?

[sdhull]

What do you think about extracting the rails integration into a separate gem, much like how the MCP stuff lives in a separate gem?

I have a colleague who built a large POC (+7.6k lines) using this gem but not using the rails integration bits at all. I know there are pros and cons of rolling your own vs using a larger, community-supported project, but I'd feel a lot better about rolling our own if the large, community-supported project wasn't also pulled in as a result of require "ruby_llm".

It would also be nice to ensure that dependency on the rails integration doesn't sneak into the non-rails portions of RubyLLM gem

[crmne]

No. The rails integration is 532 lines of code, not loaded unless you're running RubyLLM inside Rails, and it adds no dependencies to the library.

Splitting it out would mean releasing, maintaining, and synchronizing two gems for no practical benefit.

It would also be nice to ensure that dependency on the rails integration doesn't sneak into the non-rails portions of RubyLLM gem

As your colleague’s 7.6k+ line POC shows, it doesn’t.

❯ cloc lib/ruby_llm/active_record/*_methods.rb lib/ruby_llm/active_record/acts_as.rb
       4 text files.
       4 unique files.
       0 files ignored.

github.com/AlDanial/cloc v 2.06  T=0.01 s (794.1 files/s, 132616.8 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Ruby                             4            125             11            532
-------------------------------------------------------------------------------
SUM:                             4            125             11            532
-------------------------------------------------------------------------------

[sdhull]

Copy that. I do see the extra work of "releasing, maintaining, and synchronizing two gems" would be a pain. I'm not entirely convinced that there would be "no practical benefit" longer-term, but I do see your point.

Anyway thanks for the quick reply and clear answer