Hello At the moment llm.rb lets you spawn a tool with a number of different concurrency strategies: fiber, thread, async-task, and ractor. The first three are great for IO bound work, ractor is great for CPU bound work but is not always easy to use. Enter Kernel.fork The fifth strategy is Kernel.fork backed by an IPC layer that transports values between a parent and child. It is also usually easier to use than ractors - at least for now. With this change llm.rb makes it easy to execute your tools with all the major concurrency strategies that Ruby supports, so you can focus on solving problems and not implementing infrastructure that is usually not easy to get right. Check it out: https://github.com/llmrb/llm.rb#readme Robert