Mikhail Sveshnikov

@shcheklein @dmpetrov @omesser @aguschin

> no mlem ls I'd say if we dont have index, dont act like we have one. We can implement repo scanning as internal api used only by studio and...

We have `mlem types` now, but we should try and add this to help

you can run `mlem types env heroku` and it will print all available options for it in general, `mlem types ` The reason to use `-c` instead of `--` is...

related https://github.com/iterative/mlem/issues/332

So, I ran the example. Let's do this step by step. First, I ran the `create_message` only for one type like this ```python dt = interface.get_method_returns("predict") print("\n".join(create_message(Signature(name="aaa", args=[], returns=dt)))) ```...

Next, I did the same for dataframe type like this (also add `as_frame=True` to `load_iris`) ``` dt = interface.get_method_args("predict")['data'] print("\n".join(create_message(Signature(name="aaa", args=[], returns=dt)))) ``` and got ``` message DataFrame { repeated...

As for `ConstrainedListValue`, it's a result of ``` dt = interface.get_method_returns("predict_proba") print("\n".join(create_message(Signature(name="aaa", args=[], returns=dt)))) ``` if you do this `print(dt.get_serializer().get_model().schema())` you'll see that this type is actually a 2D array....

Also, can you add link to numpy protobuf impl here?

https://github.com/ilevkivskyi/typing_inspect