spray-json | A lightweight , clean and simple JSON implementation in Scala | JSON Processing library

You are not allowed to use context outside of an actor. And you do it in callback of your future.

You should address this problem using a json encoding/decoding library. Here is an example using circe and it's semi-automatic mode.

Since you mentionned in the comments that you are struggling with generic types in your case classes, I'm also including that. Basically, to derive an encoder or decoder for a class Foo[T] that contains a T, you have to prove that there is a way to encode and decode T. This is done by asking for an implicit Encoder[T] and Decoder[T] where you derive Encoder[Foo[T]] and Decoder[Foo[T]]. You can generalize this reasoning to work with more than one generic type of course, you just have to have an encoder/decoder pair implicitly available where you derive the encoder/decoder for the corresponding case class.

Your userFormat isn't in scope for ElasticsearchSource.typed, so import it, e.g.:

It seems that I needed to use "com.lightbend.akka" %% "akka-stream-alpakka-csv" % "2.0.2" instead of "com.lightbend.akka" %% "akka-stream-alpakka-reference" % "2.0.2".

Your problem is related to the Mongo Reactive Streams Driver. The version that Alpakka uses is a little bit old (Aug 13, 2019). See in the Alpakka doc in the "direct dependencies" section:

org.mongodb mongodb-driver-reactivestreams 1.12.0

In that jar you can find the Success class:

Can you try replacing "org.reactivemongo" %% "reactivemongo" % "1.0" with "org.reactivemongo" %% "reactivemongo" % "1.0.0" % "provided"?

I copy it from Maven Repository https://mvnrepository.com/artifact/org.reactivemongo/reactivemongo_2.13/1.0.0

The sbt-assembly plugins shade rule ShadeRule.keep documentation states

The ShadeRule.keep rule marks all matched classes as "roots". If any keep rules are defined all classes which are not reachable from the roots via dependency analysis are discarded when writing the output jar.

https://github.com/sbt/sbt-assembly#shading

So in this case all the classes of the class path x.* and FooBar.* are retained while creating the fat jar. Rest all other classes including the classes in scala-library are discarded.

To fix this remove all the ShadeRule.keep rules and instead try ShadeRule.zap to selectively discard classes not required.

For example, the following shade rules removes all the HDFS classes from the far jar:

Please note that when you are doing

A couple of things to check:

1. Is your model registered in the workspace? AZUREML_MODEL_DIR only works for registered models. See this link for information about registering a model
2. Are you specifying the same version of pyspark.ml.recommendation in your InferenceConfig as you use locally? This kind of error might be due to a difference in versions
3. Have you looked at the output of print(service.get_logs())? Check out our troubleshoot and debugging documentation here for other things you can try