fix(anthropic): answer an unbound thread id on chat and Responses with the 404 and drop the stale beta-join tests

litellm/proxy/hooks/message_thread_ownership.py

@@ -20,7 +20,6 @@ if TYPE_CHECKING:
     from litellm.proxy.utils import InternalUsageCache
 
 _PASS_THROUGH_CALL_TYPE: Final = "pass_through_endpoint"
-_MESSAGES_CALL_TYPE: Final = "anthropic_messages"
 _ANTHROPIC_BETA_HEADER: Final = "anthropic-beta"
 _OWNER_KEY_PREFIX: Final = "anthropic_message_thread_owner:"
 _OWNER_TTL_SECONDS: Final = 3600
@@ -217,7 +216,7 @@ class MessageThreadOwnership(CustomLogger):
         if previous_message_id is None:
             return None
         binding: Final = await self._binding(previous_message_id, user_api_key_dict)
-        if binding is None and call_type != _MESSAGES_CALL_TYPE:
+        if binding is None and call_type == _PASS_THROUGH_CALL_TYPE:
             return None
         if binding is None or binding.owner != _owner(user_api_key_dict):
             raise HTTPException(status_code=404, detail=message_thread_not_found_body(_UNOWNED_THREAD_MESSAGE))

tests/unit/proxy/hooks/test_message_thread_ownership.py

@@ -131,15 +131,19 @@ async def test_without_a_key_only_the_signed_in_identity_that_created_a_thread_c
     assert not await allows(ownership, bob, "msg_signed_in")
 
 
+@pytest.mark.parametrize("call_type", ("anthropic_messages", "acompletion", "aresponses"))
 @pytest.mark.asyncio
-async def test_a_continue_naming_a_message_this_proxy_never_served_in_a_thread_is_thread_not_found() -> None:
+async def test_a_continue_naming_a_message_this_proxy_never_served_in_a_thread_is_thread_not_found(
+    call_type: CallTypesLiteral,
+) -> None:
     ownership: Final = hook()
     await ownership.async_post_call_success_hook(
         data={}, user_api_key_dict=OWNER, response=anthropic_message("msg_unthreaded")
     )
 
-    assert not await allows(ownership, OWNER, "msg_unthreaded")
-    assert not await allows(ownership, OWNER, "msg_unknown")
+    assert not await allows(ownership, OWNER, "msg_unthreaded", call_type)
+    assert not await allows(ownership, OWNER, "msg_unknown", call_type)
+    assert not await allows(ownership, INTRUDER, "msg_unknown", call_type)
 
 
 @pytest.mark.asyncio
@@ -182,13 +186,10 @@ async def test_a_thread_continue_on_another_route_stays_with_the_key_that_create
     assert not await allows(ownership, INTRUDER, "msg_cross_route", call_type)
 
 
-@pytest.mark.parametrize("call_type", ("pass_through_endpoint", "acompletion", "aresponses"))
 @pytest.mark.asyncio
-async def test_a_continue_outside_the_messages_route_of_a_message_the_proxy_never_bound_is_left_to_the_upstream(
-    call_type: CallTypesLiteral,
-) -> None:
-    assert await allows(hook(), OWNER, "msg_unknown", call_type)
-    assert await allows(hook(), INTRUDER, "msg_unknown", call_type)
+async def test_a_pass_through_continue_of_a_message_the_proxy_never_bound_is_left_to_the_upstream() -> None:
+    assert await allows(hook(), OWNER, "msg_unknown", "pass_through_endpoint")
+    assert await allows(hook(), INTRUDER, "msg_unknown", "pass_through_endpoint")
 
 
 @pytest.mark.asyncio

tests/unit/responses/test_responses_utils.py

@@ -968,23 +968,6 @@ class TestMergeProviderSpecificHeaders:
             extra_headers=None, provider_specific_header=None, custom_llm_provider="chatgpt"
         ) is None
 
-    def test_client_anthropic_betas_join_the_extra_header_betas(self):
-        merged = ResponsesAPIRequestUtils.merge_provider_specific_headers(
-            extra_headers={"anthropic-beta": "advisor-tool-2026-03-01", "x-tenant": "team-a"},
-            provider_specific_header={
-                "custom_llm_provider": "anthropic",
-                "extra_headers": {"anthropic-beta": "message-threads-2026-08-12"},
-            },
-            custom_llm_provider="anthropic",
-        )
-
-        assert merged is not None
-        assert merged["x-tenant"] == "team-a"
-        assert set(str(merged["anthropic-beta"]).split(",")) == {
-            "advisor-tool-2026-03-01",
-            "message-threads-2026-08-12",
-        }
-
 
 def test_responses_object_bills_an_explicit_cache_hit_at_the_explicit_rate():
     model = "dashscope/qwen-responses-explicit-cache-test"

tests/unit/test_main.py

@@ -5384,41 +5384,6 @@ def test_anthropic_oauth_credential_does_not_persist_into_next_provider_hop():
     assert "anthropic-version" not in shared_headers
 
 
-def test_anthropic_chat_sends_the_client_betas_beside_the_extra_header_betas():
-    from litellm.types.utils import ProviderSpecificHeader
-
-    def echo_sent_betas(request: httpx.Request) -> httpx.Response:
-        return httpx.Response(
-            200,
-            json={
-                "id": "msg_01EchoedBetas",
-                "type": "message",
-                "role": "assistant",
-                "model": "claude-haiku-5-5",
-                "content": [{"type": "text", "text": request.headers.get("anthropic-beta", "")}],
-                "stop_reason": "end_turn",
-                "usage": {"input_tokens": 1, "output_tokens": 1},
-            },
-        )
-
-    client: Final = HTTPHandler()
-    client.client = httpx.Client(transport=httpx.MockTransport(echo_sent_betas))
-
-    response: Final = litellm.completion(
-        model="anthropic/claude-haiku-5-5",
-        messages=[{"role": "user", "content": "Say OK"}],
-        extra_headers={"anthropic-beta": "advisor-tool-2026-03-01"},
-        provider_specific_header=ProviderSpecificHeader(
-            custom_llm_provider="anthropic", extra_headers={"anthropic-beta": "message-threads-2026-08-12"}
-        ),
-        api_key="sk-fake-anthropic-key",
-        client=client,
-    )
-
-    sent_betas: Final = frozenset(beta.strip() for beta in response.choices[0].message.content.split(","))
-    assert {"advisor-tool-2026-03-01", "message-threads-2026-08-12"} <= sent_betas
-
-
 STREAM_COST_MODEL = "gpt-4o"
 STREAMED_USAGE = {"prompt_tokens": 137, "completion_tokens": 42, "total_tokens": 179}