[Frontend] Support min_p in the Responses API (#48084)

Signed-off-by: sungbin1015 <sbin@solbox.com>
Co-authored-by: Flora Feng <4florafeng@gmail.com>
This commit is contained in:
sungbin1015
2026-10-08 20:38:17 +00:00
committed by GitHub
co-authored by Flora Feng
parent c5c1161382
commit 2ddde19e8e
5 changed files with 27 additions and 8 deletions
@@ -204,6 +204,12 @@ To get prompt token IDs before generation, use
a self-contained Responses request without inference and returns the prompt
token IDs, so a separate `/tokenize` call is not needed.
#### Sampling parameters
In addition to the OpenAI request fields, the Responses API forwards vLLM
[sampling parameters](../../api/README.md#inference-parameters) such as `top_k`,
`min_p`, `repetition_penalty`, `seed`, `stop`, and `ignore_eos` to the engine.
#### Extra parameters
The following extra parameters in the request object are supported: