Is your feature already requested in another issue?
No
Is your feature a defect?
No
Is your feature request related to a problem? Please describe.
Yes. We want to use GPT based code generation for message handlers. We want RIDDL to do that
Describe the solution you'd like
Fine-tuning a GPT system to understand RIDDL is a lot of work. We'll get there eventually, but in the meantime, there's a faster path to code generation. Given that we have a competent chat-oriented GPT like ChatGPT or Microsoft Copilot, it should be simple enough to have RIDDL generate GPT AI prompts that would cause it to spit out the definition of code in a choice of languages by using a lot of context in the prompts. Since DDD is nicely compartmentalized, most of the context (other bounded contexts) is irrelevant.
RIDDL's hierarchical structure allows a useful AI Prompt to be generated to serve as the context for a more precise prompt to take an action such as generating an On Clause's code or modifying the RIDDL model. This should be a class, GeneratePrompt that takes a PassesResult which provides it all the quick lookup facilities to do its job quickly.
Describe the alternatives you've considered
Fine-tuning an AI to completely grok RIDDL and understand it well enough to generate code implementations from the handler specifications.
Additional Context
None
Is your feature already requested in another issue?
No
Is your feature a defect?
No
Is your feature request related to a problem? Please describe.
Yes. We want to use GPT based code generation for message handlers. We want RIDDL to do that
Describe the solution you'd like
Fine-tuning a GPT system to understand RIDDL is a lot of work. We'll get there eventually, but in the meantime, there's a faster path to code generation. Given that we have a competent chat-oriented GPT like ChatGPT or Microsoft Copilot, it should be simple enough to have RIDDL generate GPT AI prompts that would cause it to spit out the definition of code in a choice of languages by using a lot of context in the prompts. Since DDD is nicely compartmentalized, most of the context (other bounded contexts) is irrelevant.
RIDDL's hierarchical structure allows a useful AI Prompt to be generated to serve as the context for a more precise prompt to take an action such as generating an On Clause's code or modifying the RIDDL model. This should be a class, GeneratePrompt that takes a PassesResult which provides it all the quick lookup facilities to do its job quickly.
Describe the alternatives you've considered
Fine-tuning an AI to completely grok RIDDL and understand it well enough to generate code implementations from the handler specifications.
Additional Context
None