About the guidance usage

[Hanochhu]

Only local LLM can controll the structure of output?

    with assistant():
        lm += "```python \n"+gen(name="code", stop="```",
                                 temperature=1.0)

ConstraintException: The model attempted to generate b'To convert the' after the prompt `b'...        <|im_end|><|im_start|>assistant\n'`, but that does
not match the given grammar constraints! Since your model is a remote API that does not support full guidance
integration we cannot force the model to follow the grammar, only flag an error when it fails to match.
You can try to address this by improving the prompt, making your grammar more flexible, rerunning with
a non-zero temperature, or using a model that supports full guidance grammar constraints.

[hrusli]

I guess for now it is what it is..
you can try:

with assistant():
      lm += "```python \n"
with assistant():
        lm += gen(name="code", stop="```",
                                 temperature=0.2)

[Hanochhu]

tk, it's work