Merge 636cf604f5 into f82e4c80ac

litellm/proxy/proxy_server.py

@@ -2545,7 +2545,9 @@ async def otel_unhandled_exception_handler(request: Request, exc: Exception):
     if isinstance(exc, (ProxyException, HTTPException, RequestValidationError)):
         raise exc
     response: Final = await _unhandled_exception_response(request, exc)
-    response.headers.update(trace_response_headers(getattr(request.state, "parent_otel_span", None)))
+    response.headers.update(
+        {**trace_response_headers(getattr(request.state, "parent_otel_span", None)), "connection": "close"}
+    )
     return response
 
 

tests/unit/proxy/proxy_server/test_exception_handlers.py

@@ -643,6 +643,31 @@ async def test_otel_unhandled_exception_handler_answers_a_db_outage_with_503_no_
     assert json.loads(response.body) == _DB_OUTAGE_503_BODY
 
 
+@pytest.mark.parametrize(
+    ("path", "exc", "status_code"),
+    [
+        ("/v1/messages/count_tokens", RuntimeError("kaboom"), 500),
+        ("/v1/messages/count_tokens", httpx.ConnectError("All connection attempts failed"), 503),
+        ("/v1/traces", RuntimeError("kaboom"), 500),
+    ],
+    ids=["unhandled_500", "db_outage_503", "otlp_500"],
+)
+def test_an_unhandled_exception_response_tells_the_client_to_drop_the_connection(
+    path: str, exc: Exception, status_code: int
+) -> None:
+    async def fail() -> None:
+        raise exc
+
+    app: Final = FastAPI()
+    app.exception_handler(Exception)(otel_unhandled_exception_handler)
+    app.add_api_route(path, fail, methods=["POST"])
+
+    response: Final = TestClient(app, raise_server_exceptions=False).post(path)
+
+    assert response.status_code == status_code
+    assert response.headers["connection"] == "close"
+
+
 @pytest.mark.asyncio
 @pytest.mark.parametrize("path", ("/v1/messages", "/claude_code_gateway/v1/messages", "/anthropic/v1/messages"))
 async def test_otel_unhandled_exception_handler_answers_the_anthropic_api_error_on_the_messages_routes(