This is an automated email from the ASF dual-hosted git repository.
kaxil pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/airflow.git
The following commit(s) were added to refs/heads/main by this push:
new ef1a92bd803 Remove unused [datalib] extra from OpenAI provider base
install (#69408)
ef1a92bd803 is described below
commit ef1a92bd8034bd052772594ed3f07d0903d51d5c
Author: Yash jain <[email protected]>
AuthorDate: Tue Jul 14 01:01:24 2026 +0530
Remove unused [datalib] extra from OpenAI provider base install (#69408)
---
providers/openai/README.rst | 2 +-
providers/openai/docs/changelog.rst | 17 +++++++++++++++++
providers/openai/docs/index.rst | 2 +-
providers/openai/pyproject.toml | 8 +++++++-
4 files changed, 26 insertions(+), 3 deletions(-)
diff --git a/providers/openai/README.rst b/providers/openai/README.rst
index 9faea5840c8..82cf7a72047 100644
--- a/providers/openai/README.rst
+++ b/providers/openai/README.rst
@@ -57,7 +57,7 @@ PIP package Version required
========================================== ==================
``apache-airflow`` ``>=2.11.0``
``apache-airflow-providers-common-compat`` ``>=1.12.0``
-``openai[datalib]`` ``>=2.37.0``
+``openai`` ``>=2.37.0``
========================================== ==================
The changelog for the provider package can be found in the
diff --git a/providers/openai/docs/changelog.rst
b/providers/openai/docs/changelog.rst
index ad42f7740ac..2dfae17aec0 100644
--- a/providers/openai/docs/changelog.rst
+++ b/providers/openai/docs/changelog.rst
@@ -20,6 +20,23 @@
Changelog
---------
+.. Behavior note
+
+The ``[datalib]`` extra of the ``openai`` SDK is no longer part of this
+provider's base install. That extra existed to support the SDK's legacy
+``openai tools fine_tunes.prepare_data`` CLI (removed in openai 1.x) and
+transitively pulled in ``numpy``, ``pandas``, and ``pandas-stubs`` for every
+provider user. None of these packages are imported by the provider's
+source, tests, or example DAGs, so the extra was pure transitive bloat.
+
+Users whose DAG code relied on ``numpy`` or ``pandas`` being installed as a
+side-effect of installing this provider should declare those packages
+explicitly, or install the SDK extra directly:
+
+.. code-block:: bash
+
+ pip install 'openai[datalib]'
+
1.8.0
.....
diff --git a/providers/openai/docs/index.rst b/providers/openai/docs/index.rst
index 678075f53e3..b39317fc48a 100644
--- a/providers/openai/docs/index.rst
+++ b/providers/openai/docs/index.rst
@@ -139,7 +139,7 @@ PIP package Version required
========================================== ==================
``apache-airflow`` ``>=2.11.0``
``apache-airflow-providers-common-compat`` ``>=1.12.0``
-``openai[datalib]`` ``>=2.37.0``
+``openai`` ``>=2.37.0``
========================================== ==================
Downloading official packages
diff --git a/providers/openai/pyproject.toml b/providers/openai/pyproject.toml
index e2421f0aea5..cf7f5a28799 100644
--- a/providers/openai/pyproject.toml
+++ b/providers/openai/pyproject.toml
@@ -61,7 +61,13 @@ requires-python = ">=3.10"
dependencies = [
"apache-airflow>=2.11.0",
"apache-airflow-providers-common-compat>=1.12.0",
- "openai[datalib]>=2.37.0",
+ # The ``[datalib]`` extra of the ``openai`` SDK pulls in numpy / pandas /
+ # pandas-stubs. It was historically used by the SDK's legacy
+ # ``openai tools fine_tunes.prepare_data`` CLI, which was removed in
+ # openai 1.x. Nothing in this provider (source, tests, example DAGs, or
+ # cross-provider example DAGs) uses pandas or numpy, so ``[datalib]`` is
+ # pure transitive bloat here and is not required at runtime.
+ "openai>=2.37.0",
]
[dependency-groups]
