Compatibility with pydantic v2.10+

[yeesian]

Details in #610 (comment)

[Viicos]

I've opened googleapis/python-aiplatform#4701.

[ostefano]

Hi @yeesian , I see that python-aiplatform v1.76.0 has been released (thank you btw 🙏).

Does that mean we could remove the pydantic constraint and bump langchain-google-*?

[yeesian]

Does that mean we could remove the pydantic constraint and bump langchain-google-*?

I tried (in an adhoc fashion using pip to force install the latest version of pydantic) and it worked for me (with google-cloud-aiplatform==1.76, but not 1.75).

If you're willing, can you give it a shot and let me know if it worked for you too?

If so, I think the upper bound in

langchain-google/libs/vertexai/pyproject.toml

Line 23 in bd24d50

pydantic = ">=2.9,<2.10"

can be lifted @lkuligin

[ostefano]

@yeesian is there a specific test I can run to verify everything works as expected or instantiating VertexAiEmbedding (by invoking the API retrieving the embedding) would the best approach?

[ostefano]

@yeesian @lkuligin I am done running all our integration tests, and everything seems to be working all right when using google-cloud-aiplatform==1.76 and pydantic==2.10.4.

[tool.uv]
package = true
override-dependencies = [
    # Package 'google-cloud-aiplatform' is not compatible with modern pydantic >2.10
    #   see https://github.com/langchain-ai/langchain-google/issues/613
    #   see https://github.com/googleapis/python-aiplatform/pull/4701
    # However version '1.76' should be compatible even though the dependency has not been bumped yet
    # Here we force it so we can test it
    #   see https://github.com/langchain-ai/langchain-google/issues/613#issuecomment-2578383795
    "google-cloud-aiplatform==1.76",
    "pydantic==2.10.4",
]

Can we lift the upper bound here https://github.com/langchain-ai/langchain-google/blob/main/libs/vertexai/pyproject.toml and do a new release?

[NuniTelo]

Hi! We are having this issue even with pydantic v1. Is there any way to fix it? We are stuck at the moment.

[ostefano]

Hi! We are having this issue even with pydantic v1. Is there any way to fix it? We are stuck at the moment.

Unlikely the lower bound can be adjusted since pydantic v2 seems to be the minimum supported version ">=2.9,<2.10".

[ostefano]

@lkuligin added PR here to ease the process: #696

[ostefano]

Just noticed that #693 has also been opened.

Modulo the updated google-cloud-aiplatform>=1.76.0, both PRs do the same.