Merge bbf9021189 into e9b1c11098

litellm/llms/cloudflare/chat/transformation.py

@@ -1,8 +1,13 @@
-from typing import Final
+from collections.abc import Coroutine
+from typing import Final, Literal, overload
 
 import httpx
 
 from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.prompt_templates.common_utils import (
+    audio_or_image_in_message_content,
+    handle_messages_with_content_list_to_str_conversion,
+)
 from litellm.litellm_core_utils.url_utils import join_api_base
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
@@ -87,6 +92,35 @@ class CloudflareChatConfig(OpenAIGPTConfig):
             api_base=api_base,
         )
 
+    @overload
+    def _transform_messages(
+        self,
+        messages: list[AllMessageValues],  # mutable-ok: base signature
+        model: str,
+        is_async: Literal[True],
+    ) -> Coroutine[object, object, list[AllMessageValues]]: ...  # mutable-ok: base signature
+
+    @overload
+    def _transform_messages(
+        self,
+        messages: list[AllMessageValues],  # mutable-ok: base signature
+        model: str,
+        is_async: Literal[False] = False,
+    ) -> list[AllMessageValues]: ...  # mutable-ok: base signature
+
+    def _transform_messages(
+        self,
+        messages: list[AllMessageValues],  # mutable-ok: base signature
+        model: str,
+        is_async: bool = False,
+    ) -> list[AllMessageValues] | Coroutine[object, object, list[AllMessageValues]]:  # mutable-ok: base signature
+        # Flattening keeps only the text, so skip it when any message carries an image or audio part
+        has_media: Final = any(audio_or_image_in_message_content(message) for message in messages)
+        flattened: Final = messages if has_media else handle_messages_with_content_list_to_str_conversion(messages)
+        if is_async:
+            return super()._transform_messages(messages=flattened, model=model, is_async=True)
+        return super()._transform_messages(messages=flattened, model=model, is_async=False)
+
     def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
         return CloudflareError(
             status_code=status_code,

tests/unit/llms/cloudflare/test_cloudflare_transformation.py

@@ -205,6 +205,90 @@ def test_validate_environment_sets_bearer_and_content_type():
     assert headers["Content-Type"] == "application/json"
 
 
+def test_transform_request_flattens_content_part_arrays_to_string():
+    config = CloudflareChatConfig()
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {"type": "text", "text": "Hello"},
+                {"type": "text", "text": "More text"},
+            ],
+        }
+    ]
+
+    body = config.transform_request(
+        model="@cf/meta/llama-2-7b-chat-int8",
+        messages=messages,
+        optional_params={},
+        litellm_params={},
+        headers={},
+    )
+
+    assert body["messages"][0]["content"] == "HelloMore text"
+
+
+def test_transform_request_leaves_string_content_untouched():
+    config = CloudflareChatConfig()
+    messages = [
+        {"role": "system", "content": "you are terse"},
+        {"role": "user", "content": "weather in nyc?"},
+    ]
+
+    body = config.transform_request(
+        model="@cf/meta/llama-2-7b-chat-int8",
+        messages=messages,
+        optional_params={},
+        litellm_params={},
+        headers={},
+    )
+
+    assert [m["content"] for m in body["messages"]] == ["you are terse", "weather in nyc?"]
+
+
+def test_transform_request_does_not_flatten_when_a_part_is_not_text():
+    config = CloudflareChatConfig()
+    parts = [
+        {"type": "text", "text": "What is in this picture?"},
+        {"type": "image_url", "image_url": {"url": "https://example.com/cat.png"}},
+    ]
+
+    body = config.transform_request(
+        model="@cf/meta/llama-3.2-11b-vision-instruct",
+        messages=[{"role": "user", "content": parts}],
+        optional_params={},
+        litellm_params={},
+        headers={},
+    )
+
+    assert body["messages"][0]["content"] == parts
+
+
+def test_async_transform_request_flattens_content_part_arrays_to_string():
+    config = CloudflareChatConfig()
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {"type": "text", "text": "Hello"},
+                {"type": "text", "text": "More text"},
+            ],
+        }
+    ]
+
+    body = asyncio.run(
+        config.async_transform_request(
+            model="@cf/meta/llama-2-7b-chat-int8",
+            messages=messages,
+            optional_params={},
+            litellm_params={},
+            headers={},
+        )
+    )
+
+    assert body["messages"][0]["content"] == "HelloMore text"
+
+
 def _chat_response() -> dict[str, object]:
     return {
         "id": "chatcmpl-cf",