fix: generic type args should be serialized for de-serialization, clo…

[SylphiaWindy]

• fix(checkpoint): generic type args should be serialized for de-serialization

• PR message*:

  • Description: when jsonplus serialize a pydantic v2 object, it simply writes type name of the object and module of the type to graph. In most cases, if the object is created from a specialized generic pydantic model, the deserialization will fail and a fallback dict object returned. The original code will try to import class GenericModel[AnotherModel] from the module, this only work if the specialized type is defined at top level of a module, I assume it is pydantic's behaviour to add identifier GenericModel[AnotherModel] to module after calling a generic pydantic model's __class_getitem__ method.

The fix code will write all generic type arguments into graph and then create these types on de-serialization.
It is compatible with data serialized by the original code and maintains the same behavior as the original code (returning a dict). closing #6102.

• Issue: jsonplus cannot deserialize generic pydantic(v2) types correctly #6102
• Dependencies: None
• tests: added tests.test_jsonplus.test_serde_jsonplus_with_pydantic_generic

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Preview	Comments	Updated (UTC)
langgraph-docs-preview	Ignored	Preview		Sep 8, 2025 10:57am