[Bugfix] Expose usage field in GenerateResponse for disaggregated serving (#42748)
Signed-off-by: AIvashov <ivashov.aleksey@proton.me> Signed-off-by: NickLucche <nicolo.lucchesi@mistral.ai> Co-authored-by: NickLucche <nicolo.lucchesi@mistral.ai>
Aleksei Ivashov committed
Jul 1, 2026 at 13:00 UTC
024b06b0dc0c0a6e5ec45bcdea21d35e43fcc23f
2 files changed
+4
-2
vllm/entrypoints/scale_out/token_in_token_out/protocol.py
+3
-1
@@ -221,8 +221,10 @@ class GenerateResponse(BaseModel):
221
"through out the inference process and return in response."
222
),
223
)
224
+ model: str | None = None
225
+ created: int | None = None
226
choices: list[GenerateResponseChoice]
225
-
227
+ usage: UsageInfo | None = Field(default=None)
228
prompt_logprobs: list[dict[int, Logprob] | None] | None = None
229
230
kv_transfer_params: dict[str, Any] | None = Field(
vllm/entrypoints/scale_out/token_in_token_out/serving.py
+1
-1
@@ -320,7 +320,7 @@ class ServingTokens(OpenAIServing):
320
request_metadata.final_usage_info = usage
321
322
response = GenerateResponse(
323
- id=request_id,
323
+ request_id=request_id,
324
created=created_time,
325
model=model_name,
326
choices=choices,