When using the VertexAIGeminiGenerator, I believe it this piece of code here should run without issues, assuming that the GCP_PROJECT_ID and GOOGLE_APPLICATION_CREDENTIALS are also set:

from haystack_integrations.components.generators.google_vertex import VertexAIGeminiGenerator

gemini = VertexAIGeminiGenerator()
result = gemini.run(parts = ["What is the most interesting thing you know?"])
for answer in result["replies"]:
    print(answer)

However, requiring project_id as a parameter hinders our ability to create a multi-LLM service abstraction that leverages various Haystack generators. We aim for generator-agnostic implementation to test and use the best performers.

As you can see in my console the variables are indeed set