mistral, openai: allow anthropic-style messages in message histories (#20565)

libs/partners/mistralai/langchain_mistralai/chat_models.py

@@ -283,9 +283,16 @@ def _convert_message_to_mistral_chat_message(
                 tool_calls.append(chunk)
         else:
             pass
+        if tool_calls and message.content:
+            # Assistant message must have either content or tool_calls, but not both.
+            # Some providers may not support tool_calls in the same message as content.
+            # This is done to ensure compatibility with messages from other providers.
+            content: Any = ""
+        else:
+            content = message.content
         return {
             "role": "assistant",
-            "content": message.content,
+            "content": content,
             "tool_calls": tool_calls,
         }
     elif isinstance(message, SystemMessage):

libs/partners/openai/langchain_openai/chat_models/base.py

@@ -148,6 +148,22 @@ def _convert_dict_to_message(_dict: Mapping[str, Any]) -> BaseMessage:
         return ChatMessage(content=_dict.get("content", ""), role=role, id=id_)
 
 
+def _format_message_content(content: Any) -> Any:
+    """Format message content."""
+    if content and isinstance(content, list):
+        # Remove unexpected block types
+        formatted_content = []
+        for block in content:
+            if isinstance(block, dict) and "type" in block and block["type"] != "text":
+                continue
+            else:
+                formatted_content.append(block)
+    else:
+        formatted_content = content
+
+    return formatted_content
+
+
 def _convert_message_to_dict(message: BaseMessage) -> dict:
     """Convert a LangChain message to a dictionary.
 
@@ -158,7 +174,7 @@ def _convert_message_to_dict(message: BaseMessage) -> dict:
         The dictionary.
     """
     message_dict: Dict[str, Any] = {
-        "content": message.content,
+        "content": _format_message_content(message.content),
     }
     if (name := message.name or message.additional_kwargs.get("name")) is not None:
         message_dict["name"] = name

libs/standard-tests/langchain_standard_tests/integration_tests/chat_models.py

@@ -117,12 +117,13 @@ class ChatModelIntegrationTests(ABC):
             assert isinstance(result.content, str)
             assert len(result.content) > 0
 
-    def test_tool_message(
+    def test_tool_message_histories(
         self,
         chat_model_class: Type[BaseChatModel],
         chat_model_params: dict,
         chat_model_has_tool_calling: bool,
     ) -> None:
+        """Test that message histories are compatible across providers."""
         if not chat_model_has_tool_calling:
             pytest.skip("Test requires tool calling.")
         model = chat_model_class(**chat_model_params)
@@ -130,9 +131,15 @@ class ChatModelIntegrationTests(ABC):
         function_name = "my_adder_tool"
         function_args = {"a": "1", "b": "2"}
 
-        messages = [
+        human_message = HumanMessage(content="What is 1 + 2")
-            HumanMessage(content="What is 1 + 2"),
+        tool_message = ToolMessage(
-            AIMessage(
+            name=function_name,
+            content=json.dumps({"result": 3}),
+            tool_call_id="abc123",
+        )
+
+        # String content (e.g., OpenAI)
+        string_content_msg = AIMessage(
             content="",
             tool_calls=[
                 {
@@ -141,13 +148,38 @@ class ChatModelIntegrationTests(ABC):
                     "id": "abc123",
                 },
             ],
-            ),
+        )
-            ToolMessage(
+        messages = [
-                name=function_name,
+            human_message,
-                content=json.dumps({"result": 3}),
+            string_content_msg,
-                tool_call_id="abc123",
+            tool_message,
-            ),
+        ]
+        result = model_with_tools.invoke(messages)
+        assert isinstance(result, AIMessage)
+
+        # List content (e.g., Anthropic)
+        list_content_msg = AIMessage(
+            content=[
+                {"type": "text", "text": "some text"},
+                {
+                    "type": "tool_use",
+                    "id": "abc123",
+                    "name": function_name,
+                    "input": function_args,
+                },
+            ],
+            tool_calls=[
+                {
+                    "name": function_name,
+                    "args": function_args,
+                    "id": "abc123",
+                },
+            ],
+        )
+        messages = [
+            human_message,
+            list_content_msg,
+            tool_message,
         ]
-
         result = model_with_tools.invoke(messages)
         assert isinstance(result, AIMessage)