fix(proxy): keep the stored prompt refusal's first sentence inside the websocket close reason

The close reason carries the refusal's first sentence, and the stored prompt sentence ran past the 123 bytes a close frame allows, so it was cut mid-word. The sentence is shorter now and the rest of the explanation stays in the error event.

litellm/proxy/pass_through_endpoints/pass_through_endpoints.py

@@ -2508,8 +2508,8 @@ _TRANSCRIPTION_MODEL_PLACEMENTS: Final = (
 _STORED_PROMPT_PLACEMENTS: Final = (("prompt",), ("response", "prompt"), ("session", "prompt"))
 STORED_PROMPT_PROBE_MODEL: Final = "litellm-unverifiable-stored-prompt-model"
 _STORED_PROMPT_REFUSAL: Final = (
-    "This key is limited to specific models, and the gateway cannot check which model a stored prompt runs, "
-    "so frames that reference a stored prompt are refused for this key."
+    "This key is limited to specific models, so it cannot run a stored prompt. "
+    "The gateway cannot check which model a stored prompt runs."
 )
 _FIRST_SENTENCE: Final = re.compile(r".*?[.!?](?=\s|$)")
 

tests/unit/proxy/pass_through_endpoints/test_pass_through_endpoints.py

@@ -7774,10 +7774,13 @@ async def test_websocket_passthrough_refuses_a_stored_prompt_for_a_key_limited_t
 
     relayed = await _relay_client_frames([json.dumps(frame)], gate)
 
+    error_message: Final = json.loads(relayed.client.send_text.await_args.args[0])["error"]["message"]
+    close_reason: Final = relayed.client.close.await_args.kwargs["reason"]
     assert gate.checked == [STORED_PROMPT_PROBE_MODEL]
     assert relayed.upstream.send.await_args_list == []
-    assert "stored prompt" in json.loads(relayed.client.send_text.await_args.args[0])["error"]["message"]
     assert relayed.client.close.await_args.kwargs["code"] == 1008
+    assert "stored prompt" in close_reason
+    assert close_reason.endswith(".") and error_message.startswith(close_reason)
 
 
 @pytest.mark.asyncio