There is a dependency on LangChain which makes porting this away from Python difficult. Also may couple to Open AI.
Options:
- extract the multi prompt processing from LangChain
- try a simpler “menu prompt” approach to just ask the LLM to pick the best command or else the menu summary