fix(core): stringify empty list from tool return instead of passing as content blocks

[Giulio Leone (giulio-leone)]

Problem

BaseTool subclasses that return an empty list from _run/_arun get inconsistent behavior compared to non-empty lists:

class MyTool(BaseTool):
    name = "my_tool"
    description = "Returns a list"

    def _run(self, n: int) -> list:
        return [{"item": str(i)} for i in range(n)]

tool = MyTool()

# Non-empty list → correctly JSON-stringified
tool.run({"n": 1}, tool_call_id="1").content  # '[{"item": "0"}]'

# Empty list → passed as raw [] (0 content blocks, not an empty list value)
tool.run({"n": 0}, tool_call_id="2").content  # []  ← BUG

This causes downstream failures because [] is interpreted as zero content blocks rather than an empty list result.

Root Cause

_is_message_content_type() uses all() to check if every item in a list is a valid content block. For empty lists, all([]) returns True (vacuous truth), so the empty list passes the content-type check and is used directly instead of being stringified.

Fix

Add bool(obj) check to short-circuit on empty lists:

def _is_message_content_type(obj: Any) -> bool:
    return isinstance(obj, str) or (
        isinstance(obj, list)
        and bool(obj)  # ← empty lists are not content blocks
        and all(_is_message_content_block(e) for e in obj)
    )

Tests

• Extended parametrized test__is_message_content_type with empty list case
• Added regression test test_tool_returning_empty_list_is_stringified that verifies both n=1 (non-empty → JSON string) and n=0 (empty → "[]") return consistent string content

Fixes #30578

[codspeed-hq]

Merging this PR will degrade performance by 11.58%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

❌ 1 regressed benchmark
✅ 12 untouched benchmarks
⏩ 23 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	test_import_time[RunnableLambda]	521.7 ms	590 ms	-11.58%

---

_{Comparing giulio-leone:fix/empty-list-tool-return-30578 (70a3fce) with master (527fc02)}

Footnotes

1. 23 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[Giulio Leone (giulio-leone)]

Friendly ping — CI is green, tests pass, rebased on latest. Ready for review whenever convenient. Happy to address any feedback. 🙏

[Giulio Leone (giulio-leone)]

Friendly ping — rebased on latest and ready for review. Happy to address any feedback!