I was wondering if there were any plans or ideas for the documentation of function arguments within TalonDoc. This would just be needed within the Python function and not in the Talon script obviously. I don't think this is particularly difficult technically but more so a logistical question of getting people to align on a single documentation format for their Talon functions and the associated arguments.
For context, I work on the https://github.com/C-Loftus/talon-ai-tools repo and one of the things I am very interested in is the OpenAI function calling API. You can pass a list of functions with doc strings, argument names, and argument descriptions, and then the model will try to intuit which function you want to call based on a command. Thus I have a core group of a few functions as a test that can be called through natural language (ie you dont have to match the exact talon command but rather just the general semantic idea) but the issue is scaling this to a point that is maintainable and doesn't duplicate code.
Since we don't have argument descriptions, I generally need to manually annotate them which is probably not a good solution.
Not entirely sure if this is more appropriate for community since there hasn't been much general argument documentation discussion but figured I would bring it up here since it is related to documentation standards.
I was wondering if there were any plans or ideas for the documentation of function arguments within TalonDoc. This would just be needed within the Python function and not in the Talon script obviously. I don't think this is particularly difficult technically but more so a logistical question of getting people to align on a single documentation format for their Talon functions and the associated arguments.
For context, I work on the https://github.com/C-Loftus/talon-ai-tools repo and one of the things I am very interested in is the OpenAI function calling API. You can pass a list of functions with doc strings, argument names, and argument descriptions, and then the model will try to intuit which function you want to call based on a command. Thus I have a core group of a few functions as a test that can be called through natural language (ie you dont have to match the exact talon command but rather just the general semantic idea) but the issue is scaling this to a point that is maintainable and doesn't duplicate code.
Since we don't have argument descriptions, I generally need to manually annotate them which is probably not a good solution.
Not entirely sure if this is more appropriate for community since there hasn't been much general argument documentation discussion but figured I would bring it up here since it is related to documentation standards.