refactor(responses): make web_search its own elif branch in tool executor

Move web_search config injection from nested inside the else catch-all
to a proper elif branch, consistent with the MCP and file_search tool
handling.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Sébastien Han <seb@redhat.com>

Author: leseb

src/ogx/providers/inline/responses/builtin/responses/tool_executor.py

@@ -378,9 +378,8 @@ async def _execute_tool(
                             query=query,
                             response_file_search_tool=response_file_search_tool,
                         )
-            else:
-                # For web_search, inject config from response_tools into kwargs
-                if function_name == "web_search" and ctx.response_tools:
+            elif function_name == "web_search":
+                if ctx.response_tools:
                     response_web_search_tool = next(
                         (t for t in ctx.response_tools if isinstance(t, OpenAIResponseInputToolWebSearch)),
                         None,
@@ -395,12 +394,13 @@ async def _execute_tool(
                         if response_web_search_tool.search_context_size:
                             tool_kwargs["search_context_size"] = response_web_search_tool.search_context_size
 
-                attributes = {
-                    "tool_name": function_name,
-                }
-                # TODO: follow semantic conventions for Open Telemetry tool spans
-                # https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-spans/#execute-tool-span
-                with tracer.start_as_current_span("invoke_tool", attributes=attributes):
+                with tracer.start_as_current_span("invoke_tool", attributes={"tool_name": function_name}):
+                    result = await self.tool_runtime_api.invoke_tool(
+                        tool_name=function_name,
+                        kwargs=tool_kwargs,
+                    )
+            else:
+                with tracer.start_as_current_span("invoke_tool", attributes={"tool_name": function_name}):
                     result = await self.tool_runtime_api.invoke_tool(
                         tool_name=function_name,
                         kwargs=tool_kwargs,