fix(openai): handle unrecognized server tool names in _consolidate_calls

[Yifei Leng (Senna1960321)]

Summary

Fixes #36140

_consolidate_calls crashes with UnboundLocalError when a server_tool_call has a name not in the known set (web_search, file_search, code_interpreter, remote_mcp, mcp_list_tools). This affects any new or unrecognized server tool types (e.g. computer, image_generation) during v1 → Responses API format conversion.

Changes:

• Add fallback else branch for unrecognized tool names that emits both items unchanged (matching the existing "not a matching pair" behavior)
• Change second if to elif for file_search to avoid redundant condition evaluation on every web_search call

Repro before fix:

from langchain_openai.chat_models._compat import _consolidate_calls

items = [
    {"type": "server_tool_call", "name": "computer", "id": "call_1"},
    {"type": "server_tool_result", "tool_call_id": "call_1", "output": "result"},
]
list(_consolidate_calls(items))
# UnboundLocalError: cannot access local variable 'collapsed' where it is not associated with a value

Review notes

The fix is intentionally minimal: unrecognized tools pass through unchanged rather than attempting a best-effort collapse, since we don't know the target schema for unknown tool types.

AI Disclosure

This bug was identified through code review with AI assistance.

[codspeed-hq]

Merging this PR will degrade performance by 12.75%

⚠️ 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
✅ 2 untouched benchmarks
⏩ 37 skipped benchmarks¹

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	test_init_time	121.5 ms	139.3 ms	-12.75%

---

_{Comparing Senna1960321:fix/consolidate-calls-unbound-local (fb09dc2) with master (1778b08)}

Footnotes

1. 37 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. ↩