修复流式参数处理：尊重客户端明确指定的stream参数

- 修正transformers中强制添加stream=true的错误逻辑
- 只有客户端明确指定stream参数时才转发该参数
- 客户端未指定stream时不强制添加，保持原有意图
- 更新routes.js中相应的流式判断逻辑
- 确保非流式请求得到正确处理

routes.js

@@ -89,7 +89,7 @@ async function handleChatCompletions(req, res) {
 
     if (model.type === 'anthropic') {
       transformedRequest = transformToAnthropic(openaiRequest);
-      const isStreaming = openaiRequest.stream !== false;
+      const isStreaming = openaiRequest.stream === true;
       headers = getAnthropicHeaders(authHeader, clientHeaders, isStreaming, modelId);
     } else if (model.type === 'openai') {
       transformedRequest = transformToOpenAI(openaiRequest);
@@ -120,7 +120,7 @@ async function handleChatCompletions(req, res) {
       });
     }
 
-    const isStreaming = transformedRequest.stream !== false;
+    const isStreaming = transformedRequest.stream === true;
 
     if (isStreaming) {
       res.setHeader('Content-Type', 'text/event-stream');
@@ -268,7 +268,7 @@ async function handleDirectResponses(req, res) {
       });
     }
 
-    const isStreaming = openaiRequest.stream !== false;
+    const isStreaming = openaiRequest.stream === true;
 
     if (isStreaming) {
       // 直接转发流式响应，不做任何转换
@@ -350,7 +350,7 @@ async function handleDirectMessages(req, res) {
     const clientHeaders = req.headers;
     
     // 获取 headers
-    const isStreaming = anthropicRequest.stream !== false;
+    const isStreaming = anthropicRequest.stream === true;
     const headers = getAnthropicHeaders(authHeader, clientHeaders, isStreaming, modelId);
 
     // 注入系统提示到 system 字段

transformers/request-anthropic.js

@@ -6,10 +6,14 @@ export function transformToAnthropic(openaiRequest) {
   
   const anthropicRequest = {
     model: openaiRequest.model,
-    messages: [],
-    stream: openaiRequest.stream !== false
+    messages: []
   };
 
+  // Only add stream parameter if explicitly provided by client
+  if (openaiRequest.stream !== undefined) {
+    anthropicRequest.stream = openaiRequest.stream;
+  }
+
   // Handle max_tokens
   if (openaiRequest.max_tokens) {
     anthropicRequest.max_tokens = openaiRequest.max_tokens;

transformers/request-openai.js

@@ -7,10 +7,14 @@ export function transformToOpenAI(openaiRequest) {
   const targetRequest = {
     model: openaiRequest.model,
     input: [],
-    store: false,
-    stream: openaiRequest.stream !== false
+    store: false
   };
 
+  // Only add stream parameter if explicitly provided by client
+  if (openaiRequest.stream !== undefined) {
+    targetRequest.stream = openaiRequest.stream;
+  }
+
   // Transform max_tokens to max_output_tokens
   if (openaiRequest.max_tokens) {
     targetRequest.max_output_tokens = openaiRequest.max_tokens;