core[patch]: Propagate module name to create model (#26267)

* This allows pydantic to correctly resolve annotations necessary for
building pydantic models dynamically.
* Makes a small fix for RunnableWithMessageHistory which was fetching
the OutputType from the RunnableLambda that was yielding another
RunnableLambda. This doesn't propagate the output of the RunnableAssign
fully (i.e., with concrete type information etc.)

Resolves issue: https://github.com/langchain-ai/langchain/issues/26250

libs/core/tests/unit_tests/runnables/test_history.py

@@ -454,6 +454,41 @@ def test_get_input_schema_input_dict() -> None:
     )
 
 
+def test_get_output_schema() -> None:
+    """Test get output schema."""
+    runnable = RunnableLambda(
+        lambda input: {
+            "output": [
+                AIMessage(
+                    content="you said: "
+                    + "\n".join(
+                        [
+                            str(m.content)
+                            for m in input["history"]
+                            if isinstance(m, HumanMessage)
+                        ]
+                        + [input["input"]]
+                    )
+                )
+            ]
+        }
+    )
+    get_session_history = _get_get_session_history()
+    with_history = RunnableWithMessageHistory(
+        runnable,
+        get_session_history,
+        input_messages_key="input",
+        history_messages_key="history",
+        output_messages_key="output",
+    )
+    output_type = with_history.get_output_schema()
+
+    assert _schema(output_type) == {
+        "title": "RunnableWithChatHistoryOutput",
+        "type": "object",
+    }
+
+
 def test_get_input_schema_input_messages() -> None:
     from pydantic import RootModel