introduce Prediction API - #989

@jgeewax

Basic usage here LGTM. How about bulk-loading via the CSV trainedmodels.update method ? (https://cloud.google.com/prediction/docs/reference/v1.6/trainedmodels/update)

I'll ask someone from the Prediction team to take a look.

I could be misunderstanding the API, but I think the only place it accepts a CSV file is when you create the model and point it to a GCS file (which is supported here):

prediction.createModel('new-model-name', {
  data: gcs.file('input-data.csv')
}, function() {});

For the update API, we have three ways of hitting that endpoint:

model.setMetadata({
  output: 'label',
  csvInstance: ['data']
}, function() {});

model.train('label', 'data', function() {});

var writeStream = model.createWriteStream('label');
writeStream.write('data');
// ... any number of writes
writeStream.end();

Added todos for better docs that show how to do these things

@callmehiphop feel free to pick this one apart now :)

Still reading through a lot of the documentation, but are we not providing support for hostedmodels?

Changes made!

are we not providing support for hostedmodels?

I intended to originally. But when I read more about them, I found there are only 3, and they have a pretty serious limit:

Hosted models have a usage limit of 100 predictions/day/user across all demonstration models. Developers can set higher paid usage limits for models they create.

So, it sounded more like a "getting to know Prediction API" learning tool to me at this point. If and when it handles more queries, and the gallery grows larger, supporting this would be great.

@jgeewax any thoughts on supported hostedmodels?

@jgeewax should we wait for a review from someone from the Prediction team?

⌚ This PR will self-merge in 24 hours