BentoML

Make sure the client meets <https://www.rfc-editor.org/rfc/rfc1867>

<@UQF2L5GQ0> I am using Postman to send the POST API request. I am using form-data (Content Type: application/x-www-form-urlencoded) and I have a file included in the arguments.

This may be helpful to you <https://stackoverflow.com/a/41435972>

I am sending the POST request appropriately. It still sends the same exception.
Here is how it looks like in the Postman:

Is the curl example on the swagger page working?

When I send the request from the Swagger page it works. I will try to implement that API definition and see if it works.
I wanted to ask also if it is possible to have as input multiple files as input (without a defined amount)

It's not supported by MultipartIO. Could you share us the case that requires variable io parameters?

Is there a way for me to accept multiple files? You had this in v0.13
I need it for one of my endpoints

Accepting multiple files (fixed amount) is supported

This is how it is solved in the swagger page:
```curl -X 'POST' \
  '<http://172.16.4.174:5000/predict_example5>' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'files_stream=@4.jpg;type=image/jpeg' \
  -F 'form_stream={"sid": "S0RjcTI2RjNQYjBZMDkxMDIwMjIwNTQ2NTA=", "x": "y", "z": "r"}'```

<@UQF2L5GQ0> Is there a way of doing it for an undefined amount of images?

It's not officially supported for the sake of for the sake of a clearer API.

Can you share an actual usage scenario where you need this feature? This could help to decide whether to include such a feature.

I have an api endpoint that I allow users upload multiple files and perform my model's predictions on all of the files to give them a "result" to all of the files they uploaded

your client upload multiple files through form-data?

Then you will generate a random key for each file?

I will return a json that includes the prediction from each file

Can you share more specific examples of the input and output of the API?

The input is a json and a list of files. The output is a json that holds the prediction of each image as a key-value json object

So the keys are the keys of the form data?

Actually I want to correct myself.
I accept list of files as input and the output is a JSON that holds the predictions of all the images all together. Just a JSON output

Then how does the API user know which result represent which image?

The API user wants, needs, and gets the result for all images together and not for each image

This is a feature that we intend to implement, but don't have any concrete plans to immediately. Is this something that's blocking you? <@U01GMPKSG0H> maybe we can consider doing this?