Computed value with deeply nested foreign keys fails with AttributeError

For our application we need to calculate an aggregated value in a relation (Batch in the example) from a field in a relation that is four levels of foreign keys removed. Aggregates for a field three levels removed work fine, but with an additional level it fails with `AttributeError: 'QuerySet' object has no attribute 'value'.

I have a minimal example that demonstrates the problem. With line 18 commented-out, the result is as expected, but with line 18 active, the AttributeError is raised.

from tortoise import fields, run_async, Tortoise
from tortoise.contrib.pydantic import pydantic_model_creator
from tortoise.models import Model


class Batch(Model):
    actions: fields.ReverseRelation["Action"]
    name: str = fields.TextField()

    def names(self) -> list[str]:
        return [action.task.job.name for action in self.actions]

    def values(self) -> list[int]:
        return [action.task.job.node.value for action in self.actions]

    class PydanticMeta:
        computed = (
            # "values",   # Uncomment this to trigger the error
            "names",
        )


class Action(Model):
    batch: fields.ForeignKeyRelation["Batch"] = fields.ForeignKeyField("models.Batch", related_name="actions")
    task: fields.ForeignKeyRelation["Task"] = fields.ForeignKeyField("models.Task", related_name="actions")


class Task(Model):
    job: fields.ForeignKeyRelation["Job"] = fields.ForeignKeyField("models.Job", related_name="tasks")
    actions: fields.ReverseRelation["Action"]


class Job(Model):
    node: fields.ForeignKeyRelation["Node"] = fields.ForeignKeyField("models.Node", related_name="jobs")
    tasks: fields.ReverseRelation["Task"]
    name: str | None = fields.TextField(null=True)


class Node(Model):
    jobs: fields.ReverseRelation["Job"]
    value: int = fields.IntField()


async def run():
    await Tortoise.init(db_url="sqlite://:memory:", modules={"models": ["__main__"]})
    await Tortoise.generate_schemas()

    Batch_Pydantic = pydantic_model_creator(Batch)

    node_1 = await Node.create(value=4)
    node_2 = await Node.create(value=5)
    job_1 = await Job.create(node=node_1, name="First job")
    job_2 = await Job.create(node=node_2, name="Second job")
    task_1 = await Task.create(job=job_1)
    task_2 = await Task.create(job=job_2)
    batch = await Batch.create(name="batch 1")
    await Action.create(task=task_1, batch=batch)
    await Action.create(task=task_2, batch=batch)

    p = await Batch_Pydantic.from_tortoise_orm(batch)
    print(p.json(indent=4))


if __name__ == "__main__":
    run_async(run())

Probably related to this is the fact that the json print on line 61 does not include the referenced Node relation for the Job table rows:

{
    "id": 1,
    "name": "batch 1",
    "actions": [
        {
            "id": 1,
            "task": {
                "id": 1,
                "job": {
                    "id": 1,
                    "name": "First job"
                }
            }
        },
        {
            "id": 2,
            "task": {
                "id": 2,
                "job": {
                    "id": 2,
                    "name": "Second job"
                }
            }
        }
    ],
    "names": [
        "First job",
        "Second job"
    ]
}

Start with the minimal example, especially pydantic_model_creator(Batch) and Batch_Pydantic.from_tortoise_orm(batch), then trace how the computed values field accesses the four-level relation chain. Reproduce the AttributeError with the values entry enabled and verify that serialization completes and includes the referenced Node data.