langchain-google-vertexai's pyarrow<24.0.0 conflicts with projects needing pyarrow 25
envgap__langchain-ai__langchain-google-1951
01 / FAILURE SIGNATURE
As reported upstream
### Error Message and Stack Trace (if applicable)
Not a benchmark task.
- The project already builds and runs before the fix, so there is nothing to repair.
02 / ENVIRONMENT RECIPE
- Base commit
75f75815e0fdb8b719a1cb576184cc4833a68a16- Manifest
libs/vertexai/pyproject.toml- Reproduce
Awaiting issue-specific recipe- Run under trace
Awaiting a meaningful runtime command
03 / ORIGINAL ISSUE TEXT
langchain-ai/langchain-google #1951 · read the original issue
### Package (Required)
- [ ] langchain-google-genai
- [x] langchain-google-vertexai
- [ ] langchain-google-community
- [ ] Other / not sure / general
### Checked other resources
- [x] I added a descriptive title to this issue
- [x] I searched the LangChain documentation and API reference (linked above)
- [x] I used the GitHub search to find a similar issue and didn't find it
- [x] I am sure this is a bug and not a question or request for help
### Example Code (Python)
```python
[project]
name = "repro"
version = "0.1.0"
requires-python = ">=3.13"
dependencies = [
"langchain-google-vertexai>=3.2.4",
"pyarrow>=25.0.1",
]
Then `uv lock`.
```
### Error Message and Stack Trace (if applicable)
```shell
× No solution found when resolving dependencies:
╰─▶ Because langchain-google-vertexai>=3.2.4 depends on
pyarrow>=19.0.1,<24.0.0 and your project depends on
langchain-google-vertexai>=3.2.4, we can conclude that your project
depends on pyarrow>=19.0.1,<24.0.0.
And because your project depends on pyarrow>=25.0.1, we can conclude
that your project's requirements are unsatisfiable.
```
### Description
We ran into this in a monorepo where several services share an internal library. One service uses `langchain-google-vertexai`; another reads Parquet and needs pyarrow 25. Because the shared library declares `pyarrow`, vertexai's `<24.0.0` cap holds it back and the two can't coexist.
As far as I can tell vertexai is the only package in the tree that requires pyarrow, and doesn't import it directly – I assume it's there for the optional pyarrow paths in `google-cloud-bigquery` / `google-cloud-aiplatform`, which don't set an upper bound themselves.
Would you consider bumping the ceiling to `<26.0.0`?
**Environment:** Python 3.13 · langchain-google-vertexai 3.2.4 · uv 0.12.504 / LABELS
Labels from the report text only; not yet run
No supported category has been assigned.
Label rules and the text that matched
[]