fix(responses): drop redundant SSE prefix strip after SSEDecoder switch

cursoragent · yassin-berriai · cursoragent · commit 0b353ab2e6f2 · 2026-05-22T09:19:11.000Z
SSEDecoder already strips the 'data:' field prefix from each event, so the
extra call to _strip_sse_data_from_chunk on sse.data was redundant and could
incorrectly mangle payloads whose actual content starts with 'data:'.

Co-authored-by: Yassin Kortam &lt;yassin@berri.ai&gt;
diff --git a/litellm/responses/streaming_iterator.py b/litellm/responses/streaming_iterator.py
@@ -28,7 +28,7 @@
 from litellm.responses.utils import ResponsesAPIRequestUtils
 from litellm.types.llms.openai import ResponsesAPIStreamEvents
 from litellm.types.utils import CallTypes
-from litellm.utils import CustomStreamWrapper, async_post_call_success_deployment_hook
+from litellm.utils import async_post_call_success_deployment_hook
 
 
 @lru_cache(maxsize=1)
@@ -121,10 +121,10 @@ def _process_chunk(self, chunk) -> Optional[Any]:
         if not chunk:
             return None
 
-        # Handle SSE format (data: {...})
-        chunk = CustomStreamWrapper._strip_sse_data_from_chunk(chunk)
-        if chunk is None:
-            return None
+        # NOTE: ``SSEDecoder`` already strips the SSE ``data:`` field prefix, so
+        # the value passed in here is the raw field content. Do not re-run
+        # ``_strip_sse_data_from_chunk`` on it — doing so would incorrectly mangle
+        # payloads whose actual JSON value happens to start with ``data:``.
 
         # Handle "[DONE]" marker
         if chunk == STREAM_SSE_DONE_STRING:
