fix(python): emit Pydantic field aliases to fix nested field validation

[kafkas]

Description

Linear ticket: FER-8077

A previous fix, #11011 only helped when callers invoked model_validate() directly, but it did not apply to nested validation paths (e.g. when a parent model is validated via TypeAdapter), so wire keys like is_base64_encoded could still fail with “Field required” inside nested objects. This PR fixes nested decoding by emitting pydantic.Field(alias=...) for aliased fields so Pydantic v2 can resolve wire aliases during compiled nested validation.

Changes Made

• Ensures pydantic.Field(alias=...) is emitted for aliased fields
• Removes the UncheckedBaseModel.model_validate Pydantic v2 override
• Adds a test for validation via TypeAdapter

Testing

• Unit tests added/updated
• Manual testing completed

---

Note

Emit pydantic.Field(alias=...) for aliased fields and update parsing/config so aliases are honored during nested Pydantic v2 validation; remove the v2 model_validate override.

• Codegen / Models:
  • Emit pydantic.Field(alias=...) for aliased fields across generated models (including optional fields, unions, nested objects, and examples).
  • Update initializer to always pass alias to field definitions.
• Core utilities:
  • Refine parse_obj_as to detect real Pydantic aliases and only pre-dealias when needed; supports both v1 and v2 paths.
  • Remove UncheckedBaseModel.model_validate override for Pydantic v2.
  • Enable population by field name and alias (populate_by_name / allow_population_by_field_name) in UniversalBaseModel.
• Tests/Seeds:
  • Regenerate seed SDKs and test utilities to use pydantic.Field(alias=...) and new model config to ensure nested validation works with wire aliases.

^{Written by Cursor Bugbot for commit 5d2873c. This will update automatically on new commits. Configure here.}

[github-actions]

🌱 Seed Test Selector

Select languages to run seed tests for:

• Python
• TypeScript
• Java
• Go
• Ruby
• C#
• PHP
• Swift
• Rust
• OpenAPI
• Postman

---

How to use: Click the ⋯ menu above → "Edit" → check the boxes you want → click "Update comment". Tests will run automatically and snapshots will be committed to this PR.

[cursor]

This is the final PR Bugbot will review for you during this billing cycle

Your free Bugbot reviews will reset on January 18

Details

You are on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

[cursor]

Wrong base class in issubclass check for v1-on-v2 context

The alias detection logic checks issubclass(type_, pydantic.BaseModel) but UniversalBaseModel in this file extends pydantic.v1.BaseModel (line 56). In Pydantic v2, these are entirely separate class hierarchies - pydantic.v1.BaseModel is NOT a subclass of pydantic.BaseModel. This means the condition will always be False for any model in the with_pydantic_v1_on_v2 context, causing the alias detection code to never execute. The check needs to use pydantic.v1.BaseModel instead to match the base class actually used by models in this context.

 

[cursor]

Missing allow_population_by_field_name config in v1-on-v2 variant

The UniversalBaseModel.Config class in the with_pydantic_v1_on_v2 variant is missing allow_population_by_field_name = True, which this PR adds to all other variants (both generated seed files and the similar with_pydantic_v1_on_v2/with_aliases variant). This config option is needed to allow Pydantic v1 to accept both wire aliases and Python field names. Combined with the wrong issubclass check bug, when de-aliasing always occurs but models also have pydantic.Field(alias=...) set, validation will fail because Pydantic v1 without this config only accepts the alias.