[Hyacinth-of-Security (lxcxjxhx)]

What

Three text-mode open() calls in shipped code rely on the platform default encoding. On Windows that is the ANSI code page (GBK on a zh-CN machine, cp1252 on en-US), so:

• ChatHistory.store_chat_history_to_file() writes a file that is not UTF-8: a history containing héllo — 世界 lands on disk as h\xa8\xa6llo \xa1\xaa \xca\xc0\xbd\xe7 — GBK bytes for the same text. That file cannot be read on Linux/macOS or inside a container, where open() defaults to UTF-8: UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa8, and any character outside the ANSI code page raises UnicodeEncodeError when writing.
• ChatHistory.load_chat_history_from_file() is the matching half, so the pair only round-trips on the machine that wrote it.
• OnnxGenAICompletionBase._load_genai_settings reads genai_config.json from a model directory without an encoding — those configs come from the model package and are UTF-8.

Change

-        with open(file_path, "w") as local_file:
+        with open(file_path, "w", encoding="utf-8") as local_file:

-        with open(file_path) as file:
+        with open(file_path, encoding="utf-8") as file:

-            with open(json_gen_ai_config) as file:
+            with open(json_gen_ai_config, encoding="utf-8") as file:

Three lines, no behaviour change on POSIX (UTF-8 is already the default there).

Verification

Windows 11 / CPython 3.13, semantic_kernel installed editable from python/:

# before: file is not UTF-8
utf-8 decodable: False -> 'utf-8' codec can't decode byte 0xa8 in position 218

# after: file is UTF-8 and the content is byte-identical to the input
utf8 decode ok: True
text matches original: True
round trip matches: True

python/tests/unit/contents/test_chat_history.py: 76 passed before and after (the suite does not pin an encoding, so it does not notice the difference — the fix is about interoperability, not about a failing test).

Scope

Only python/semantic_kernel/. A scan of the tree finds 32 more text-mode opens without an encoding, but they are in tests, samples and scripts; I kept this PR to the three in the shipped package so it stays reviewable in one pass. Happy to send the test/sample sweep separately if you want it.

[Copilot]

🟡 Changes recommended

Existing tests would still pass if explicit UTF-8 encodings were removed from ONNX initialization and chat-history persistence, so targeted regression coverage is needed.

2 open issues

• Test explicit UTF-8 encoding for ONNX config reads · New
• Test UTF-8 persistence and decoding of non-ASCII chat history · New

What changed in this PR

This PR makes file encoding explicit in the Python package, improving chat-history portability and ensuring ONNX model configuration files are read as UTF-8.

Changes:

• Use UTF-8 when writing and reading chat-history files.
• Use UTF-8 when reading ONNX genai_config.json files.

File	Description
python/​semantic_kernel/​contents/​chat_history.py	Specifies UTF-8 for history file reads and writes.
python/​semantic_kernel/​connectors/​ai/​onnx/​services/​onnx_gen_ai_completion_base.py	Specifies UTF-8 for model configuration reads.

_{Review effort: Balanced}