Release 0.5.0 of gordo-components

Introduces the following:

• Watchman service: (gordo_components.watchman) which provides updated status of expected endpoint status. To be found at http://<ambassador host>/gordo/v0/<project name>

Release 0.4.0 of gordo-components

0.4.0 brings the following exciting changes:

• Modelbuilder now stores various metadata by default, which is used by model-server on the metadata endpoint.
• Bump pyyaml to 4.2b4 to fix vulnerability.

Release 0.3.1 of gordo-components

Version 0.3.1 changes the docker images to be based on python:3.6.8 instead of python:3.6.6, and uses a slim image for the server. Should not have any visible effects except quicker start-time on a fresh machine.

Release 0.3.0 of gordo-components

• Drop Seldon dependency for owned Gordo ML Server

  • Adds functionally similar routing as Seldon, base route of /gordo/v0/{project_name}/{sanitized machine name}
  • Routes: /metadata, /predictions, /healthcheck
  • /predictions accepts POST request with data fields of X and y; allowing X to be a nested array, which was previously disallowed. ie. now this is ok for X: [ [1, 2, 3], [4, 5, 6] ]

• Additional docker based testing.

• Refactoring of the InfluxBackedDataset

Release 0.2.0 of gordo-components

Added the following

• Support of FunctionTransformer for config files; for functions which exist in the codebase
• Convert KerasAutoEncoder into a Transformer
• SeldonModel first calls .predict() and fails over to .transform() for calling of a model, incase the final out is a transformer.
• Optionally save metadata in serializer.dump and add serializer.load_metadata for retrieval.
• SeldonModel now returns any additional metadata in the tags key in returned predictions
• pipeline_from_definition() no longer mutates the passed config
• Default log level is set to DEBUG in SeldonModel optionally overriden with env var LOG_LEVEL

Release 0.1.1 of gordo components

Bugfix release fixing tensorflow graph bug in model-server.

Release 0.1.0 of gordo components

Release 0.1.0 of gordo components. This contains support for specifying start and end dates for training periods, extra testing, and proper SCM controlled versions.

Release 0.0.4 of gordo components

Release 0.0.4 contains a parser for scikit-learn pipelines, with KerasModel and KerasAutoEncoder. KerasAutoEncoder is the default model if not specified otherwise.

Release 0.0.3 of gordo components

Release 0.0.3 contains a updated FeedForward auto-encoder which confirms to the scikit-learn API, and some helper functions to create other Keras based models with scikit-learn API support.

Release 0.0.2 of gordo components

Version 0.0.2 actually uses influxdb for model building, and writes its models to /gordo