yadda yadda yadda

Document user-defined ranking objectives

1

For MVP I guess it will be enough to have click-focused ranking. In the future we may have something more user-configurable.

Postgres support for external storage

1

As this one should be more suitable for production usage.

It can be useful to locally play with the tool, when distributes setup is not needed.

So you put a number there, and it will emit a set of percentiles over a sample of last input data. Can be useful in cases like blog posts, when...

So you can directly map a numerical vector coming from outside (like low-dimensional doc2vec item description embedding) directly as a feature value.

Nothing is stopping Metarank from importing historical events from kafka/pulsar apart from CLI options: * expose the way to pull events from such storages * update docs

To simplify integration with existing backends.