From 4d2f97397c4188cd90604f4629c17722dcd85fe5 Mon Sep 17 00:00:00 2001 From: Hao19911125 <99091644+Hao19911125@users.noreply.github.com> Date: Wed, 8 Apr 2026 16:08:57 +0800 Subject: [PATCH] =?UTF-8?q?fix(llm):=20=E6=94=B9=E7=94=A8=20max=5Ftokens?= =?UTF-8?q?=20=E9=81=BF=E5=85=8D=E4=B8=8E=20ST=20=E4=BB=A3=E7=90=86?= =?UTF-8?q?=E6=B3=A8=E5=85=A5=E7=9A=84=20max=5Ftokens=20=E5=86=B2=E7=AA=81?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ST 的 /api/backends/chat-completions/generate 端点会从预设配置注入 max_tokens，导致与扩展自行传递的 max_completion_tokens 冲突， 上游 API 返回 400。改为 max_tokens 后由 ST 统一处理，消除冲突。 Co-Authored-By: Claude Opus 4.6 --- llm/llm.js | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/llm/llm.js b/llm/llm.js index aac9219..dc4a555 100644 --- a/llm/llm.js +++ b/llm/llm.js @@ -1441,7 +1441,7 @@ async function callDedicatedOpenAICompatible( model: config.model, messages, temperature: filteredGeneration.temperature ?? 1, - max_completion_tokens: resolvedCompletionTokens, + max_tokens: resolvedCompletionTokens, stream: filteredGeneration.stream ?? false, frequency_penalty: filteredGeneration.frequency_penalty ?? 0, presence_penalty: filteredGeneration.presence_penalty ?? 0,