Fix: Python Enum types generate proper dropdown schemas with descriptions

[AvigailRoyzenberg]

Problem

Python scripts using Enum types don't generate proper UI dropdowns. Instead:

• Enum parameters show as generic text inputs (treated as object type)
• Docstring descriptions are ignored (all fields show empty "")
• Users must manually configure dropdowns for every enum parameter

class Color(str, Enum):
    RED = 'red'
    GREEN = 'green'

def main(color: Color = Color.RED):
    """
    Args:
        color (Color): Select a color
    """
    pass

Current schema (broken):

{
  "color": {
    "type": "object",
    "description": ""
  }
}

Solution

Parse Enum class definitions and docstrings to generate proper schemas:

New schema (fixed):

{
  "color": {
    "type": "string",
    "enum": ["red", "green", "blue"],
    "description": "Select a color"
  }
}

Changes

• Extract Enum class definitions and values from Python AST
• Parse docstring Args: sections for parameter descriptions
• Map Enum types to string enums instead of treating as resources
• Handle Enum.VALUE default values correctly

Testing

Added test_parse_python_sig_enum covering all functionality. All tests pass.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[AvigailRoyzenberg]

I have read the CLA Document and I hereby sign the CLA

[rubenfiszel]

This is very cool, thanks, but this parses the whole code as AST everytime which can be slow and worse it does it everytime, regardless of there is an enum or not.

3 things that can be used:

• do it lazily (but do it once) only parse metadata if we arrive to a situation where there is an unrecognized type
• only parse the code if a keyword is present in the code (Enum here but can be made more precise), no need to parse the ast if that's not present
• when parsing the ast, do a prepass to try remove everyhing that is not interesting (for instance, filter everything that is not class or within the class block)