[serve] Forward tool_calls/tool_call_id in processor inputs

src/transformers/cli/serving/utils.py

@@ -930,6 +930,12 @@ def get_processor_inputs_from_messages(messages: list[dict], modality: Modality)
         for message in messages:
             parsed = {"role": message["role"], "content": []}
 
+            # Forward tool-use fields so apply_chat_template can handle multi-turn tool conversations
+            if "tool_calls" in message:
+                parsed["tool_calls"] = message["tool_calls"]
+            if "tool_call_id" in message:
+                parsed["tool_call_id"] = message["tool_call_id"]
+
             content = message.get("content")
             if modality == Modality.LLM:
                 if isinstance(content, str):

tests/cli/test_serve.py

@@ -209,6 +209,42 @@ def test_vlm_multi_turn(self):
             self.assertIsInstance(msg["content"], list)
             self.assertEqual(msg["content"][0]["type"], "text")
 
+    def test_llm_tool_use_fields_forwarded(self):
+        """Tool-use fields (tool_calls, tool_call_id) should be forwarded to processor inputs."""
+
+        get_processor_inputs_from_messages = BaseHandler.get_processor_inputs_from_messages
+
+        tool_calls = [
+            {"id": "call_1", "type": "function", "function": {"name": "get_weather", "arguments": '{"city": "Paris"}'}}
+        ]
+        messages = [
+            {"role": "user", "content": "What's the weather in Paris?"},
+            {"role": "assistant", "tool_calls": tool_calls},
+            {"role": "tool", "content": "22°C, sunny", "tool_call_id": "call_1"},
+        ]
+        result = get_processor_inputs_from_messages(messages, Modality.LLM)
+        self.assertEqual(len(result), 3)
+        self.assertEqual(result[1]["tool_calls"], tool_calls)
+        self.assertNotIn("tool_calls", result[0])
+        self.assertEqual(result[2]["tool_call_id"], "call_1")
+        self.assertNotIn("tool_call_id", result[0])
+
+    def test_vlm_tool_use_fields_forwarded(self):
+        """Tool-use fields should be forwarded for VLM modality as well."""
+
+        get_processor_inputs_from_messages = BaseHandler.get_processor_inputs_from_messages
+
+        tool_calls = [{"id": "call_1", "type": "function", "function": {"name": "describe", "arguments": "{}"}}]
+        messages = [
+            {"role": "user", "content": "Describe this"},
+            {"role": "assistant", "tool_calls": tool_calls},
+            {"role": "tool", "content": "A landscape photo", "tool_call_id": "call_1"},
+        ]
+        result = get_processor_inputs_from_messages(messages, Modality.VLM)
+        self.assertEqual(len(result), 3)
+        self.assertEqual(result[1]["tool_calls"], tool_calls)
+        self.assertEqual(result[2]["tool_call_id"], "call_1")
+
 
 class TestGenerativeModelList(unittest.TestCase):
     def test_lists_only_generative_models(self):