[vinismarques]

Summary

Python 3.10 reached end of life in October 2026; 3.14 and 3.15 are now supported, and new workers default to 3.14.

 pyproject.toml
-  requires-python = ">=3.10,<3.14"
+  requires-python = ">=3.11,<3.16"
-  version = "5.2.3"
+  version = "5.3.0"

 scaffold/
   pyproject.toml
-    requires-python = ">=3.10,<3.14"
+    requires-python = ">=3.11,<3.15"
   Dockerfile
-    ARG BASE_IMAGE=python:3.12-slim-trixie
+    ARG BASE_IMAGE=python:3.14-slim-trixie

 .github/workflows/quality.yml
   quality   # lint, format, type check on 3.12
-    test
+  test      # matrix: 3.11, 3.12, 3.13, 3.14, 3.15

The scaffold allows up to 3.14 rather than 3.15, matching what major data-science libraries publish wheels for. TensorFlow is the exception: its latest stable release has no 3.14 wheels, so TensorFlow workers should override BASE_IMAGE and narrow requires-python.

Evidence

• Before: uv sync --python 3.14 → rejected by requires-python = ">=3.10,<3.14"
  After: uv run --locked --all-groups --python <v> pytest passes on 3.11, 3.12, 3.13, 3.14 and 3.15

Merge Danger

Door: two-way

Revert restores 3.10 support; anyone who already pinned 5.3.0 on 3.10 would need the next release.

Blast Radius: consumers

Workers still on Python 3.10 cannot upgrade to 5.3.0. Workers scaffolded from 5.3.0 only build once the 5.3.0 tag exists, since the scaffold pins the generating version.