Achieve seamless connection between program code and LLM prompt

[klb3713]

Is your feature request related to a problem? Please describe.
My Feature request has nothing to do with the current issue

Describe the solution you'd like
The current capability is one-way, i.e. formatting the output of LLM into program objects. The upgrade I would like to suggest is two-way. For example, you can execute the LLM call as a function, pass in the program object as a parameter, automatically convert the program object into a prompt internally, then execute the LLM call, and then format the return result into a program. object.
This feature will enable seamless execution of program code and LLM prompts. This is great!

Describe alternatives you've considered

Additional context

[AriMKatz]

Can't you do manual unpacking pretty easily?

[Mr-Ruben]

If this is not what you are after, could you show us some code?

class Output(BaseModel):
    name: str
    age: int

Let's imagine we just have the object.

# the object
person=Output(age=33,name="Sam")

# call only with object 
r=call_llm_with_class(   
                  prompt=str(person),
                  response_model=type(person),
                  )

print(r.output)
# {'name': 'Sam', 'age': 33}

# But that is similar to 
type(person)(**dict(r.output)) 

# which internally is
# type(person)(        **dict(r.output)                             )
# Output(                 **dict(r.output)                             )  
# Output(                 **dict({'name': 'Sam', 'age': 33})  )
# Output(                 **{'name': 'Sam', 'age': 33}          )
# Output(                 'name'= 'Sam', 'age'= 33             )

# Which produces
Output(name='Sam', age=33)

This is actual output, not pseudo-code.

call_llm_with_class is just a wrapper

[ivanleomk]

Closing this issue for now due to a lack of activity.