BentoML

1. you can use any IODescriptors’ `from_sample` method to provide an example input! This way it shows up on the swagger UI nicely

2. Yes! simply add a docstring to your API function, e.g.:
```@svc.api(..)
def predict(..):
  """
  Description for the predict endpoint goes here..
  """```

We are also considering allow advanced users to customize the generated API docs via a extra_fields json blob, does that sound like what you’re looking for?

1. I _believe_ that if you use `from_sample` in your IO descriptor, it should display that in the UI.
2. 
```svc.api(..., doc="documentation")
```
should do the trick!

Oh shoot, I totally missed that Chaoyu had already responded.

Thank you both, <@UKB4CLKP1> and <@U02HFFXSZ9D>! I had been using .from_sample() when my io descriptors were DataFrames. I’ve switched to JSON with Pydantic because of the cleanliness and ease of specifying nullables, requireds, defaults, and validators. But I did see there was a JSON .from_sample(). Do you happen to know if I can use .from_sample() for the example data but then defer to the Pydantic models for everything else?

&gt; We are also considering allow advanced users to customize the generated API docs via a extra_fields json blob, does that sound like what you’re looking for?
<@UKB4CLKP1> yes, that does sound like what i’m looking for as a nice-to-have

<@U052700AXPT> yes, you can do something like this:
```input=JSON.from_sample({'sample': [1,2,3]}, pydantic_model=MyModel)```

I’ll keep a lookout for the extra_fields json blob in future versions, but this will be helpful in the meantime. Thanks again. :bento: