According to documentation:

https://kafka-python.readthedocs.io/en/master/apidoc/KafkaConsumer.html
https://kafka-python.readthedocs.io/en/master/apidoc/ClusterMetadata.html

partitions_for_topic method return type is set. However, if the topic does not exists, ClusterMetadata's method returns None (and thus KafkaConsumer's, which calls ClusterMetadata's partitions_for_topic for that topic).

Notice that the solution should either update the documentation (explictly stating that the method may return None, its type is Optional[set]) or even better, just returning an empty set if the topic is not found.