feat: responses API compliance with upstream type alignment (#6089)

Signed-off-by: Matej Kosec <mkosec@nvidia.com>
Co-authored-by: Ishan Dhanani <ishandhanani@gmail.com>

tests/serve/common.py

@@ -156,6 +156,7 @@ def run_serve_deployment(
                     payload=payload.body,
                     timeout=payload.timeout,
                     method=payload.method,
+                    stream=payload.http_stream,
                 )
                 server_process.check_response(payload, response)
 

tests/serve/test_sglang.py

@@ -23,6 +23,8 @@ from tests.utils.payload_builder import (
     embedding_payload,
     embedding_payload_default,
     metric_payload_default,
+    responses_payload_default,
+    responses_stream_payload_default,
 )
 
 logger = logging.getLogger(__name__)
@@ -60,6 +62,8 @@ sglang_configs = {
         request_payloads=[
             chat_payload_default(),
             completion_payload_default(),
+            responses_payload_default(),
+            responses_stream_payload_default(),
             metric_payload_default(min_num_requests=6, backend="sglang"),
         ],
     ),

tests/utils/client.py

@@ -52,6 +52,7 @@ def send_request(
     timeout: float = 30.0,
     method: str = "POST",
     log_level: int = 20,
+    stream: bool = False,
 ) -> requests.Response:
     """
     Send an HTTP request to the engine with detailed logging.
@@ -97,7 +98,7 @@ def send_request(
         if method_upper == "GET":
             response = requests.get(url, params=payload, timeout=timeout)
         elif method_upper == "POST":
-            response = requests.post(url, json=payload, timeout=timeout)
+            response = requests.post(url, json=payload, timeout=timeout, stream=stream)
         else:
             # Fallback for other methods if needed
             response = requests.request(
@@ -117,8 +118,14 @@ def send_request(
         logger.debug("Response headers: %s", dict(response.headers))
 
         # Try to log response body (truncated if too long)
+        # Skip body logging for streaming responses to avoid consuming the stream
+        if stream:
+            logger.debug("Response body: <streaming, not logged>")
+        else:
             try:
-            if response.headers.get("content-type", "").startswith("application/json"):
+                if response.headers.get("content-type", "").startswith(
+                    "application/json"
+                ):
                     response_data = response.json()
                     response_str = json.dumps(response_data, indent=2)
                     if len(response_str) > 1000: