1.0.7

• Switched data loading from S3 to GCS
• Refactored data loading to be usable without Redis dependency
• Refactored configuration to minimize dependencies for the package
• Fixed packaging to be usable from Ensemble Spark job

0.7.9

• fixed results logging for related addons recommender

0.7.8

• fixed randomization edge case
• improved logging

0.7.7

• enable randomization
• fix randomization to work with negative weights

0.7.6

• hotfix: bug in experiments inclusion

0.0.16

Encapsulate distance function in similarity recommender

0.0.15

Remove JSON file caching when downloading from S3

0.0.14

Pin miniconda version

0.0.13

Soft-fail if SimilarityRecommender files cannot be downloaded

0.0.12: Extensively use Numpy for similarity based recommendations

This PR will extract Numpy arrays/matrices out of the donor
and LR curves files. All the heavy lifting computations
were changed to make use of the vectorized Numpy facilities.