fix(ollama): avoid leading newline in multimodal text content (#37481)

Fixes #37480 

---

Avoid prepending a leading newline when converting multimodal message
text content in `ChatOllama`. Previously, the first text segment in
list-format message content was always prefixed with `\n`, which could
break formatting-sensitive vision/OCR models.

Also adds a regression test to ensure multimodal message content does
not start with an unintended leading newline.

Verified by running:

- make format
- make lint
- make test

libs/partners/ollama/langchain_ollama/chat_models.py

@@ -1024,9 +1024,13 @@ class ChatOllama(BaseChatModel):
             else:  # List
                 for content_part in message.content:
                     if isinstance(content_part, str):
-                        content += f"\n{content_part}"
+                        if content:
+                            content += "\n"
+                        content += content_part
                     elif content_part.get("type") == "text":
-                        content += f"\n{content_part['text']}"
+                        if content:
+                            content += "\n"
+                        content += content_part["text"]
                     elif content_part.get("type") == "tool_use":
                         continue
                     elif content_part.get("type") == "image_url":

libs/partners/ollama/tests/unit_tests/test_chat_models.py

@@ -1137,3 +1137,30 @@ def test_non_ai_message_reasoning_content_ignored() -> None:
     ]
     ollama_messages = llm._convert_messages_to_ollama_messages(messages)
     assert "thinking" not in ollama_messages[0]
+
+
+def test_multimodal_message_content_has_no_leading_newline() -> None:
+    """Test that multimodal text content does not start with a newline."""
+    message = HumanMessage(
+        content=[
+            {
+                "type": "text",
+                "text": "Extract all text from this image.",
+            },
+            {
+                "type": "image_url",
+                "image_url": {
+                    "url": (
+                        "data:image/png;base64,"
+                        "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="
+                    )
+                },
+            },
+        ]
+    )
+
+    model = ChatOllama(model="any")
+
+    ollama_messages = model._convert_messages_to_ollama_messages([message])
+
+    assert ollama_messages[0]["content"] == "Extract all text from this image."