fix(responses): fix subtle bugs in non-function tool calling (llamastack#3817)

We were generating "FunctionToolCall" items even for MCP (and
file-search, etc.) server-side calls. ID mismatches, etc. galore.

Author: ashwinb

llama_stack/providers/inline/agents/meta_reference/responses/streaming.py

@@ -44,8 +44,11 @@
     OpenAIResponseObjectStreamResponseRefusalDone,
     OpenAIResponseOutput,
     OpenAIResponseOutputMessageContentOutputText,
+    OpenAIResponseOutputMessageFileSearchToolCall,
     OpenAIResponseOutputMessageFunctionToolCall,
+    OpenAIResponseOutputMessageMCPCall,
     OpenAIResponseOutputMessageMCPListTools,
+    OpenAIResponseOutputMessageWebSearchToolCall,
     OpenAIResponseText,
     OpenAIResponseUsage,
     OpenAIResponseUsageInputTokensDetails,
@@ -177,6 +180,7 @@ async def create_response(self) -> AsyncIterator[OpenAIResponseObjectStream]:
                 # (some providers don't support non-empty response_format when tools are present)
                 response_format = None if self.ctx.response_format.type == "text" else self.ctx.response_format
                 logger.debug(f"calling openai_chat_completion with tools: {self.ctx.chat_tools}")
+
                 params = OpenAIChatCompletionRequestWithExtraBody(
                     model=self.ctx.model,
                     messages=messages,
@@ -613,19 +617,22 @@ async def _process_streaming_chunks(
 
                             # Emit output_item.added event for the new function call
                             self.sequence_number += 1
-                            function_call_item = OpenAIResponseOutputMessageFunctionToolCall(
-                                arguments="",  # Will be filled incrementally via delta events
-                                call_id=tool_call.id or "",
-                                name=tool_call.function.name if tool_call.function else "",
-                                id=tool_call_item_id,
-                                status="in_progress",
-                            )
-                            yield OpenAIResponseObjectStreamResponseOutputItemAdded(
-                                response_id=self.response_id,
-                                item=function_call_item,
-                                output_index=len(output_messages),
-                                sequence_number=self.sequence_number,
-                            )
+                            is_mcp_tool = tool_call.function.name and tool_call.function.name in self.mcp_tool_to_server
+                            if not is_mcp_tool and tool_call.function.name not in ["web_search", "knowledge_search"]:
+                                # for MCP tools (and even other non-function tools) we emit an output message item later
+                                function_call_item = OpenAIResponseOutputMessageFunctionToolCall(
+                                    arguments="",  # Will be filled incrementally via delta events
+                                    call_id=tool_call.id or "",
+                                    name=tool_call.function.name if tool_call.function else "",
+                                    id=tool_call_item_id,
+                                    status="in_progress",
+                                )
+                                yield OpenAIResponseObjectStreamResponseOutputItemAdded(
+                                    response_id=self.response_id,
+                                    item=function_call_item,
+                                    output_index=len(output_messages),
+                                    sequence_number=self.sequence_number,
+                                )
 
                         # Stream tool call arguments as they arrive (differentiate between MCP and function calls)
                         if tool_call.function and tool_call.function.arguments:
@@ -806,6 +813,35 @@ async def _coordinate_tool_execution(
             if not matching_item_id:
                 matching_item_id = f"tc_{uuid.uuid4()}"
 
+            self.sequence_number += 1
+            if tool_call.function.name and tool_call.function.name in self.mcp_tool_to_server:
+                item = OpenAIResponseOutputMessageMCPCall(
+                    arguments="",
+                    name=tool_call.function.name,
+                    id=matching_item_id,
+                    server_label=self.mcp_tool_to_server[tool_call.function.name].server_label,
+                    status="in_progress",
+                )
+            elif tool_call.function.name == "web_search":
+                item = OpenAIResponseOutputMessageWebSearchToolCall(
+                    id=matching_item_id,
+                    status="in_progress",
+                )
+            elif tool_call.function.name == "knowledge_search":
+                item = OpenAIResponseOutputMessageFileSearchToolCall(
+                    id=matching_item_id,
+                    status="in_progress",
+                )
+            else:
+                raise ValueError(f"Unsupported tool call: {tool_call.function.name}")
+
+            yield OpenAIResponseObjectStreamResponseOutputItemAdded(
+                response_id=self.response_id,
+                item=item,
+                output_index=len(output_messages),
+                sequence_number=self.sequence_number,
+            )
+
             # Execute tool call with streaming
             tool_call_log = None
             tool_response_message = None
@@ -1064,7 +1100,11 @@ async def _add_mcp_list_tools(
         self.sequence_number += 1
         yield OpenAIResponseObjectStreamResponseOutputItemAdded(
             response_id=self.response_id,
-            item=mcp_list_message,
+            item=OpenAIResponseOutputMessageMCPListTools(
+                id=mcp_list_message.id,
+                server_label=mcp_list_message.server_label,
+                tools=[],
+            ),
             output_index=len(output_messages) - 1,
             sequence_number=self.sequence_number,
         )

llama_stack/providers/inline/agents/meta_reference/responses/tool_executor.py

@@ -93,7 +93,7 @@ async def execute_tool_call(
 
         # Build result messages from tool execution
         output_message, input_message = await self._build_result_messages(
-            function, tool_call_id, tool_kwargs, ctx, error_exc, result, has_error, mcp_tool_to_server
+            function, tool_call_id, item_id, tool_kwargs, ctx, error_exc, result, has_error, mcp_tool_to_server
         )
 
         # Yield the final result
@@ -356,6 +356,7 @@ async def _build_result_messages(
         self,
         function,
         tool_call_id: str,
+        item_id: str,
         tool_kwargs: dict,
         ctx: ChatCompletionContext,
         error_exc: Exception | None,
@@ -375,7 +376,7 @@ async def _build_result_messages(
             )
 
             message = OpenAIResponseOutputMessageMCPCall(
-                id=tool_call_id,
+                id=item_id,
                 arguments=function.arguments,
                 name=function.name,
                 server_label=mcp_tool_to_server[function.name].server_label,
@@ -389,14 +390,14 @@ async def _build_result_messages(
         else:
             if function.name == "web_search":
                 message = OpenAIResponseOutputMessageWebSearchToolCall(
-                    id=tool_call_id,
+                    id=item_id,
                     status="completed",
                 )
                 if has_error:
                     message.status = "failed"
             elif function.name == "knowledge_search":
                 message = OpenAIResponseOutputMessageFileSearchToolCall(
-                    id=tool_call_id,
+                    id=item_id,
                     queries=[tool_kwargs.get("query", "")],
                     status="completed",
                 )