IBM Watson Machine Learning手写数字示例笔记本身份验证错误

Question

When working through section 6.1 of the Recognition of hand-written digits with IBM Watson Machine Learning notebook, calling the Watson Machine Learning REST API to get the list of published models using:

GET /v3/wml_instances/{instance_id}/published_models

endpoint, there is a authorization (401) error with the following response text:

{"trace":"-1xmid60eblabs","errors":[{"code":"invalid_token_format","message":"Invalid token format. Bearer token format should be used.","target":{"type":"header","name":"Authorization"}}]}

The previous REST API call in the notebook to get the instance details using the authorization token returned from the /v3/identity/token endpoint worked ok.

Looks like an error in the notebook. Did the API authorization scheme change breaking the notebook?

Answer 1

This endpoint in the REST API appears to be expecting authentication of the form Authorization: <type> <credentials> per W3C in HTTP 1.0

I updated the notebook code to add the type in the Authorization field of the header:

header = {'Content-Type': 'application/json', 'Authorization': 'Bearer ' + mltoken}

The rest of the Watson Machine Learning API calls to create the online deployment and score new data use the same header in the notebook and all execute correctly.