The random tests are a joke: it reuses the same graph used by the fixed tests for the entire test suite and only chooses random cities between 16 possible cities. This can barely be called random.

It is not explained what the 3rd argument of findShortestWay does, and it is always supplied with 0. Is it the author solution's recursion base argument?

Dijkstra's algorithm is a duplicate to existing katas.

Since you expect an absolutely exact value, you need to specify the requirements further. For instance, when do you considere that points are "in" or "out" (inclusive/exclusive, I mean). Unless you're not refering to the same method, but then it's even worse... :/

Hi,

Isn't it too close to this one: https://www.codewars.com/kata/57ab4322cf1fa567f8000c4a ?
(as it's a part of it)

The solution setup is class Node { /*...*/ } but the tests want a class Graph.
"storage [...] contains [...] edges of this vertex as values of this vertex." I don't understand that.
"Should contain method add()" Why do you require this method when it is not called in the three main tests?
Why do you pass 0 as a third argument to findShortestWay?
The random tests just call findShortestWay twice and compare the results. Better compare with your own model solution.
A native speaker should go over the description.

Also: You start with cities and distances, then continue with a generic undirected graph (vertices, edges, weights). You should mention that all weights are non-negative or Dijkstra won't work.