[Himansh Raj (iamthehimansh)]

Motivation and Context

Fixes #14685.

Since #14472, ChatHistory.add_tool_message("") keeps an empty tool result. But that result doesn't survive the XML prompt round trip that KernelFunctionFromPrompt performs on every chat completion call. to_element() renders it as a self-closing <function_result id="call_1" />. When from_element() parses that back, it reads element.text, which is None for an empty element. ChatMessageContent then stringifies the value, so the model receives the literal tool result "None" instead of an empty one.

Description

• FunctionResultContent.from_element now maps a missing element text to "". Non-empty results are unchanged.
• Added test_element_round_trip_keeps_result (empty and non-empty), which serialises the element to XML and parses it back.
• Added test_chat_history_prompt_round_trip_keeps_empty_tool_result, which reproduces the issue end to end through to_prompt() / from_rendered_prompt().

Before (main 4bb64ee, macOS arm64, Python 3.12):

{'role': 'tool', 'content': 'None', 'tool_call_id': 'call_1'}

After:

{'role': 'tool', 'content': '', 'tool_call_id': 'call_1'}

Tests, run from python/:

• pytest tests/unit/contents: 421 passed on main, 424 passed with this change.
• With the source change reverted, the two empty-result tests fail: 2 failed, 25 passed in test_function_result_content.py.
• ruff check / ruff format --check on both files are clean.

This doesn't overlap the open #14375 (the from_function_call_content_and_result factory) or #14480 (the add_tool_message path), and it doesn't touch the JSON serialize() / restore_chat_history() path, which already keeps "".

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines and the pre-submission formatting script raises no violations
• All unit tests pass, and I have added new tests where possible
• I didn't break anyone 😄

AI-assisted: yes (Cana)

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.