From be3193a29859a9d855e2535033a5703a647d467f Mon Sep 17 00:00:00 2001
From: goran-revops <46086727+goran-revops@users.noreply.github.com>
Date: Thu, 24 Sep 2026 13:23:17 +0200
Subject: [PATCH] fix(proxy): share the parallel-request slot with the SSE
 keepalive task

---
 litellm/proxy/common_request_processing.py | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/litellm/proxy/common_request_processing.py b/litellm/proxy/common_request_processing.py
index 2e43c6b0d22..83af9db99cd 100644
--- a/litellm/proxy/common_request_processing.py
+++ b/litellm/proxy/common_request_processing.py
@@ -1218,6 +1218,13 @@ async def open_sse_before_first_byte(
     if interval is None:
         return await produce_response
 
+    # The slot record must exist before this task is forked, or the release
+    # looks in the original task and never sees it.
+    from litellm.proxy.hooks.parallel_request_limiter_v3 import (
+        get_or_create_request_stash,
+    )
+
+    get_or_create_request_stash()
     produce_task: Final = asyncio.ensure_future(produce_response)
     await asyncio.wait((produce_task,), timeout=interval)
     if produce_task.done():