fix(python): preserve empty tool results

[mikemikimike]

Motivation and Context

Fixes #14359.

ChatHistory.add_tool_message() currently treats an empty string as absent content. This creates a tool message with no FunctionResultContent, loses the supplied tool-call correlation, and makes ChatMessageContent.to_dict() raise IndexError.

An empty string is a valid result for a successful command with no output or a no-op tool.

Description

Change the tool-message preparation path to distinguish None from an empty string. Empty string results now create the same FunctionResultContent as other string results, preserving the result, id, call_id, and optional function name. Non-tool messages retain their existing truthiness behavior.

Added a focused regression test that verifies the content object fields and successful serialized tool-message dictionary.

This is backward compatible: non-empty tool results and all existing message forms are unchanged, and the fix only makes the documented string input work for the valid empty-string boundary case.

Validation

• uv run pytest tests/unit/contents/test_chat_history.py -k empty_tool_message -q --basetemp D:\\project\\tscodex\\github_pr\\tmp\\semantic-kernel-14359-pytest — passed (1 test)
• uv run pytest tests/unit/contents/test_chat_history.py -q --basetemp D:\\project\\tscodex\\github_pr\\tmp\\semantic-kernel-14359 — passed (56 tests)
• uv run ruff check semantic_kernel/contents/chat_history.py tests/unit/contents/test_chat_history.py — passed
• uv run ruff format --check semantic_kernel/contents/chat_history.py tests/unit/contents/test_chat_history.py — passed
• uv run mypy semantic_kernel/contents/chat_history.py — passed
• uv run pre-commit run --files semantic_kernel/contents/chat_history.py tests/unit/contents/test_chat_history.py — passed
• git diff --check — passed

The full Python unit suite was not run because this is a localized content-model change; no model, provider, credentials, database, or external service is required.

Contribution Checklist

• The code builds clean without any errors or warnings for the touched Python module
• The PR follows the SK Contribution Guidelines and the pre-submission checks pass
• The focused unit tests pass, including a regression test
• The change is limited to the reported bug

AI Disclosure

This contribution was prepared with AI assistance and reviewed against the repository source, issue reproduction, and local test results.

[Copilot]

🟢 Approval recommended

The functional change is small, targeted, and covered by a regression test, with only a minor typing-annotation nit noted.

Pull request overview

Fixes a Python ChatHistory.add_tool_message() edge case where an empty-string tool result was treated as “no content”, producing an invalid tool message (no FunctionResultContent) and causing ChatMessageContent.to_dict() to fail.

Changes:

• Adjust tool-message preparation to treat "" as valid content by distinguishing None from empty strings.
• Preserve tool-call correlation (tool_call_id) and function metadata for empty tool results by always creating FunctionResultContent when content is not None.
• Add a focused regression test covering empty-string tool results and successful serialization via to_dict().

File summaries

File	Description
python/semantic_kernel/contents/chat_history.py	Updates tool-message preparation logic to build FunctionResultContent even when the tool result is an empty string.
python/tests/unit/contents/test_chat_history.py	Adds regression coverage ensuring empty tool results preserve correlation and serialize without errors.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

MAF Automated Review — Iteration 1

Result: No findings
Scope: full PR (1 commit(s)): 6de01064685f
Model: claude-opus-4.8

Overview

This PR changes a single guard in ChatHistory._prepare_for_add from a truthiness
test (content and not items) to an identity test (content is not None and not items) for AuthorRole.TOOL, so an empty-string tool result now builds a
FunctionResultContent(result="") instead of producing an item-less message that
crashed ChatMessageContent.to_dict() with IndexError (issue #14359). The change
is precisely scoped: None still skips the branch (behavior unchanged), non-empty
strings were already handled, the string overload's required tool_call_id guard
(chat_history.py:176-180) still runs first, and a focused regression test locks in
the content-object fields and the previously-crashing serialization path. Reviewers
empirically reproduced both the pre-fix crash and the post-fix success and could not
construct any PR-introduced defect; the residual concerns are Low-severity and
pre-existing (non-TOOL empty-content asymmetry and an XML round-trip fidelity gap).

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
No publishable findings remained after source verification for this scope.

[mikemikimike]

Addressed in commit f53a8be by changing _prepare_for_add return annotation to dict[str, Any]. This reflects the actual role and items values returned by the helper.