fix(environments): pin PythonEnvironment._write_to_file encoding to UTF-8 (#2827)

* fix(environments): pin PythonEnvironment._write_to_file encoding to UTF-8

PythonEnvironment._write_to_file is the shared helper every concrete
PythonEnvironment subclass uses to land an agent-generated script on
disk before invoking subprocess. The bare 'open(script_path, "w")'
inherits 'locale.getpreferredencoding(False)' — cp1252 on default
Windows installs — so the first non-cp1252 character (CJK string
literal, emoji, smart quote from copy-pasted documentation, PEP 3131
identifier) raises 'UnicodeEncodeError' mid-write. The script is then
partially written, the subprocess runs an invalid file, and the agent
sees a syntax error or truncated output instead of the intended
behavior.

Pin encoding="utf-8" on the open call. Adds source-level regression
coverage in test/environments/test_python_environment_utf8.py that
runs on every CI lane (no PythonEnvironment subclass extras required)
so the kwarg cannot silently regress.

* chore: add license header to test/environments/__init__.py

`check-license-headers` pre-commit hook flagged the new test package
init as missing the standard SPDX header; add the 2023-2026 ag2 header
matching the sibling test file to clear the pre-commit-check / pr-check
failures.

* Remove outdated UTF-8 regression comment

Remove regression comment regarding UTF-8 encoding handling in PythonEnvironment._write_to_file.

---------

Co-authored-by: Qingyun Wu <qingyun0327@gmail.com>
Co-authored-by: Semen Frolov <148821259+vvlrff@users.noreply.github.com>

Author: 3 people

autogen/environments/python_environment.py

@@ -87,7 +87,13 @@ def _write_to_file(self, script_path: str, content: str) -> None:
             script_path: Path to the file to write.
             content: Content to write to the file.
         """
-        with open(script_path, "w") as f:
+        # Pin UTF-8 so any non-ASCII glyph in agent-supplied script content
+        # (string literals, comments, docstrings, identifiers) round-trips on
+        # platforms whose locale.getpreferredencoding() is not UTF-8 (cp1252
+        # on default Windows installs). Without it `open(path, "w")` raises
+        # UnicodeEncodeError mid-write on the first non-cp1252 character,
+        # turning a successful agent code-gen into a runtime failure.
+        with open(script_path, "w", encoding="utf-8") as f:
             f.write(content)
 
     # Utility method for subclasses to wrap (for async support)

test/environments/__init__.py

@@ -0,0 +1,3 @@
+# Copyright (c) 2023 - 2026, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0

test/environments/test_python_environment_utf8.py

@@ -0,0 +1,16 @@
+# Copyright (c) 2023 - 2026, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+
+
+def test_python_environment_write_to_file_pins_utf8() -> None:
+    source = (REPO_ROOT / "autogen" / "environments" / "python_environment.py").read_text(encoding="utf-8")
+    assert 'open(script_path, "w", encoding="utf-8") as f' in source, (
+        "PythonEnvironment._write_to_file must pin encoding='utf-8' on its "
+        "open() call so non-cp1252 script content (CJK string literals, emoji, "
+        "smart quotes, PEP 3131 identifiers) does not crash on Windows."
+    )