Background

We're on Pydantic ≥2.11. Many models map snake_case Python fields to camelCase external keys (storage/state/API JSON), and today every field spells its alias out by hand. For example in src/crawlee/statistics/_models.py:

stats_id: Annotated[int | None, Field(alias='statsId')] = None
requests_finished: Annotated[int, Field(alias='requestsFinished')] = 0
crawler_started_at: Annotated[datetime | None, Field(alias='crawlerStartedAt')] = None

Proposal

Set an alias generator once in the model config so the camelCase aliases are derived automatically:

from pydantic.alias_generators import to_camel

model_config = ConfigDict(validate_by_name=True, validate_by_alias=True, alias_generator=to_camel)

Fields then drop the manual alias (stats_id: int | None = None) and still (de)serialize as statsId. We already standardize on validate_by_name=True, validate_by_alias=True (the 2.11 successors of populate_by_name), so the only new piece is alias_generator. Best placed on a shared base model, since subclasses currently redeclare the config.

Where it makes sense

• Apply to models that already use camelCase aliases: Request, SessionModel, StatisticsState, the storage_clients/models.py set, event data classes, etc.
• Keep an explicit Field(alias=...) wherever to_camel produces the wrong key (acronyms / irregular casing). Each field's generated alias must equal the current one.
• Skip models whose serialized form is snake_case, e.g. RequestQueueState in _file_system/_request_queue_client.py, and Configuration (BaseSettings with env-var AliasChoices).

Scope

Refactor only. The wire format must not change; generated aliases have to match the existing ones, guarded by the current (de)serialization tests.