fix(anthropic): return 400 instead of 500 when a content list holds a bare string (#42420)

* fix(anthropic): skip non-dict content items in beta-header and file-id helpers so malformed content lists return 400 instead of 500

Fixes #42094
Supersedes #42101

Co-authored-by: Pawan-Shahane <shahanepawan511@gmail.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(e2e): spawn the DB-less regression proxy with -P so the cwd cannot shadow the pinned checkout

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(e2e): launch the DB-less proxy via -I -c with an explicit sys.path so python 3.10 works, drop DIRECT_URL, remove restating docstrings

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(e2e): gate the self-booted DB-less proxy behind the owned_gateway opt-in the Buildkite container cannot satisfy

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* test(anthropic): move the bare string content item repro to tests/integration

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* style(tests): wrap the anthropic bare string wire test to the 120 column limit

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* style(tests): wrap anthropic common_utils test literals to the 120 column limit

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

* style: fix ruff findings in touched test files

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

---------

Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Pawan-Shahane <shahanepawan511@gmail.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/litellm_core_utils/prompt_templates/common_utils.py

@@ -1657,7 +1657,7 @@ def get_file_ids_from_messages(messages: list[AllMessageValues]) -> list[str]:
                 if isinstance(content, str):
                     continue
                 for c in content:
-                    if c["type"] == "file":
+                    if isinstance(c, dict) and c["type"] == "file":
                         file_object = cast(ChatCompletionFileObject, c)
                         file_object_file_field = file_object.get("file")
                         if not isinstance(file_object_file_field, dict):

litellm/llms/anthropic/common_utils.py

@@ -314,7 +314,7 @@ class AnthropicModelInfo(BaseLLMModelInfo):
             _message_content = message.get("content")
             if _message_content is not None and isinstance(_message_content, list):
                 for content in _message_content:
-                    if "cache_control" in content:
+                    if isinstance(content, dict) and "cache_control" in content:
                         return True
 
         return False
@@ -359,7 +359,7 @@ class AnthropicModelInfo(BaseLLMModelInfo):
         for message in messages:
             if "content" in message and message["content"] is not None and isinstance(message["content"], list):
                 for content in message["content"]:
-                    if "type" in content and content["type"] != "text":
+                    if isinstance(content, dict) and "type" in content and content["type"] != "text":
                         return True
         return False
 

tests/integration/contracts.json

@@ -159,6 +159,12 @@
     "tests/integration/routing/test_redis_recovery.py::test_owned_redis_outage_recovers_requests_and_real_response_cache": [
       "other.routing.redis.owned_outage_recovers_serving_and_response_cache"
     ],
+    "tests/integration/providers/test_anthropic_wire.py::test_anthropic_bare_string_content_item_is_rejected_as_client_error_before_the_wire[type_word]": [
+      "other.provider_wire.anthropic.bare_string_content_item_is_client_error"
+    ],
+    "tests/integration/providers/test_anthropic_wire.py::test_anthropic_bare_string_content_item_is_rejected_as_client_error_before_the_wire[plain]": [
+      "other.provider_wire.anthropic.bare_string_content_item_is_client_error"
+    ],
     "tests/integration/providers/test_anthropic_wire.py::test_anthropic_tool_history_and_cache_tokens_keep_wire_and_accounting_contracts": [
       "other.provider_wire.anthropic.tool_history_system_cache_and_internal_fields",
       "quota_management.spend_tracking.cache_tokens.disjoint_classes_use_explicit_rates"

tests/integration/providers/test_anthropic_wire.py

@@ -3,7 +3,6 @@ import uuid
 from typing import Final
 
 import pytest
-
 from integration._support.client import Gateway, eventually, object_value
 from integration._support.database import read_rows
 from integration._support.wire import Reply, Request, wire_server
@@ -59,3 +58,26 @@ def test_anthropic_tool_history_and_cache_tokens_keep_wire_and_accounting_contra
         parsed: Final = json.loads(metadata) if isinstance(metadata, str) else object_value(metadata)
         assert parsed["cost_breakdown"]["input_cost"] == pytest.approx(0.0245)
         assert parsed["cost_breakdown"]["output_cost"] == pytest.approx(0.008)
+
+
+@pytest.mark.covers("other.provider_wire.anthropic.bare_string_content_item_is_client_error")
+@pytest.mark.parametrize(
+    "text", [pytest.param("what type of file is this?", id="type_word"), pytest.param("hello", id="plain")]
+)
+def test_anthropic_bare_string_content_item_is_rejected_as_client_error_before_the_wire(
+    gateway: Gateway, text: str
+) -> None:
+    def respond(request: Request) -> Reply:
+        raise AssertionError(f"upstream must not be reached: {request.target}")
+
+    with wire_server(respond) as wire, gateway.scenario() as scenario:
+        model: Final = scenario.model(
+            model="anthropic/claude-sonnet-4-5-20250929", api_base=wire.url, api_key="synthetic-anthropic-key"
+        )
+        response: Final = gateway.request(
+            "POST",
+            "/v1/chat/completions",
+            {"model": model, "max_tokens": 16, "timeout": 5, "messages": [{"role": "system", "content": [text]}]},
+        )
+        assert response.status_code == 400, response.text
+        assert wire.drain() == ()

tests/test_litellm/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_common_utils.py

@@ -4,7 +4,6 @@ import json
 import os
 import sys
 from typing import Final
-from unittest.mock import MagicMock, patch
 
 import pytest
 
@@ -356,6 +355,20 @@ def test_get_file_ids_from_messages_file_field_not_dict():
     assert get_file_ids_from_messages(messages) == []
 
 
+def test_get_file_ids_from_messages_skips_bare_string_content_items():
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                "what type of file is this?",
+                {"type": "file", "file": {"file_id": "file-abc"}},
+            ],
+        }
+    ]
+
+    assert get_file_ids_from_messages(messages) == ["file-abc"]
+
+
 def test_update_messages_with_model_file_ids_skips_non_openai_file_blocks():
     """`update_messages_with_model_file_ids` is also called on user content
     before provider dispatch. It must tolerate non-OpenAI file blocks the same

tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py

@@ -14,6 +14,7 @@ import json
 import os
 import sys
 from types import SimpleNamespace
+from typing import Final
 from unittest.mock import patch
 
 import pytest
@@ -2275,3 +2276,72 @@ def test_create_anthropic_model_list_response_lists_ids_as_told():
     assert (gpt["id"], gpt["display_name"], gpt["max_input_tokens"]) == ("claude-router-gpt-4o[1m]", "GPT 4o", 1000000)
     assert (haiku["id"], haiku["display_name"]) == ("claude-haiku-4-5", "claude-haiku-4-5")
     assert (response["first_id"], response["last_id"]) == ("claude-router-gpt-4o[1m]", "claude-haiku-4-5")
+
+
+class TestMalformedContentListItems:
+    @pytest.mark.parametrize(
+        "content",
+        [
+            pytest.param(["what type of file is this?"], id="string_containing_type"),
+            pytest.param(["how do I set cache_control?"], id="string_containing_cache_control"),
+            pytest.param([None], id="none_item"),
+            pytest.param([5], id="int_item"),
+            pytest.param([["nested"]], id="list_item"),
+        ],
+    )
+    def test_beta_headers_resolve_for_non_dict_content_items(self, content: list[object]) -> None:
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        config: Final = AnthropicModelInfo()
+        messages: Final = [{"role": "user", "content": content}]
+
+        headers: Final = config.validate_environment(
+            headers={},
+            model="claude-sonnet-4-5",
+            messages=messages,
+            optional_params={},
+            litellm_params={},
+            api_key=FAKE_REGULAR_KEY,
+        )
+
+        assert headers["x-api-key"] == FAKE_REGULAR_KEY
+        assert config.is_cache_control_set(messages) is False
+        assert config.is_pdf_used(messages) is False
+
+    def test_real_content_parts_still_set_their_beta_headers(self) -> None:
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        config: Final = AnthropicModelInfo()
+
+        assert config.is_pdf_used([{"role": "user", "content": [{"type": "image", "source": {}}]}]) is True
+        assert config.is_pdf_used([{"role": "user", "content": [{"type": "text", "text": "hi"}]}]) is False
+        assert (
+            config.is_cache_control_set(
+                [
+                    {
+                        "role": "user",
+                        "content": [{"type": "text", "text": "hi", "cache_control": {"type": "ephemeral"}}],
+                    }
+                ]
+            )
+            is True
+        )
+
+    def test_mixed_list_keeps_detecting_the_valid_part(self) -> None:
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        config: Final = AnthropicModelInfo()
+        messages: Final = [{"role": "user", "content": ["what type of file is this?", {"type": "image", "source": {}}]}]
+
+        assert config.is_pdf_used(messages) is True
+
+    def test_litellm_completion_rejects_bare_string_content_item_as_bad_request(self) -> None:
+        import litellm
+
+        with pytest.raises(litellm.BadRequestError):
+            litellm.completion(
+                model="anthropic/claude-haiku-4-5-20251001",
+                messages=[{"role": "user", "content": ["what type of file is this?"]}],
+                api_key=FAKE_REGULAR_KEY,
+                max_tokens=5,
+            )