PydanticAI doesn't return None but raw VertexAI code does

[DataMonsterBoy]

So just using VertexAI libraries I can have the option to get back null values. However, when writing the equivalent code using pydanticAI I never get back null values:

Using vertex libraries:

from vertexai.generative_models import GenerativeModel, GenerationConfig

response_schema = {
    "type": "object",
    "properties": {
        "age": {
            "type": "INTEGER", "nullable": True
        }
    }
}

generation_config = GenerationConfig(
    response_mime_type="application/json",
    response_schema=response_schema
)

model = GenerativeModel("gemini-1.5-flash")
result = model.generate_content(
    "The man was very old. what is the age of the old man?",
        generation_config=generation_config
    )
print(result.candidates[0].content.parts[0].text)
# "{\"age\": null}"

Using pydanticAI:

from pydantic_ai import Agent
from pydantic import BaseModel
from pydantic_ai.models.vertexai import VertexAIModel

class AgeModel(BaseModel):
    age: int | None = None

gemini_model = VertexAIModel('gemini-1.5-flash')

prompt = "The man was very old. what is the age of the old man?"
agent = Agent(gemini_model, result_type=AgeModel)
age = agent.run_sync(prompt)
print(age.data)
# age=100

[samuelcolvin]

makes sense, we need to fix the JSON schema generated for gemini to include "nullable": True.