Is your feature request related to a problem? Please describe.
The current way to parse commands heavily relies on consistent output format generated by the LLM. If the LLM deviates from that structure, the application won't be able to reliably build applications any more.
Describe the solution you'd like
There should be a fallback when a command (e.g. dec-write) fails, and a good way to test if smaller models work reliably with the current system prompt.
Is your feature request related to a problem? Please describe.
The current way to parse commands heavily relies on consistent output format generated by the LLM. If the LLM deviates from that structure, the application won't be able to reliably build applications any more.
Describe the solution you'd like
There should be a fallback when a command (e.g. dec-write) fails, and a good way to test if smaller models work reliably with the current system prompt.