[Frontend] Forward cache salt for content parts (#54315)
Signed-off-by: Eli Gotts <78387377+eligotts@users.noreply.github.com> Co-authored-by: OpenAI Codex <noreply@openai.com>
eligotts committed
Aug 30, 2026 at 23:29 UTC
4ae172231c633f00be275133ad91094396d3280c
1 file changed
+2
vllm/entrypoints/scale_out/token_in_token_out/serving.py
+2
@@ -163,6 +163,8 @@ class ServingTokens(GenerateBaseServing):
163
mm_parser.parse_video(url, uuid)
164
mm_data, mm_uuids = await tracker.resolve_items()
165
prompt = TokensPrompt(prompt_token_ids=request.token_ids)
166
+ if request.cache_salt is not None:
167
+ prompt["cache_salt"] = request.cache_salt
168
if mm_data:
169
prompt["multi_modal_data"] = mm_data
170
if mm_uuids: