fix: omit unpopulated max_completion_tokens to prevent 1-token truncation

main.go

@@ -114,8 +114,8 @@ type ChatCompletionRequest struct {
   Tools               []Tool        `json:"tools,omitempty"`
   ToolChoice          interface{}   `json:"tool_choice,omitempty"`
   Stream              bool          `json:"stream"`
-  MaxTokens           int           `json:"max_tokens"`
-  MaxCompletionTokens int           `json:"max_completion_tokens"`
+  MaxTokens           int           `json:"max_tokens,omitempty"`
+  MaxCompletionTokens int           `json:"max_completion_tokens,omitempty"`
   Temperature         *float64      `json:"temperature,omitempty"`
   TopP                *float64      `json:"top_p,omitempty"`
   ReasoningEffort     string        `json:"reasoning_effort,omitempty"`
@@ -1508,9 +1508,14 @@ func (s *QwenService) chatDirectOpenAI(endpointURL string, w http.ResponseWriter
 
   reqCopy := req
   reqCopy.Model = resolvedModel
-  if reqCopy.MaxTokens == 0 && maxTokens > 0 {
-    reqCopy.MaxTokens = maxTokens
+  effTokens := maxTokens
+  if reqCopy.MaxCompletionTokens > 0 {
+    effTokens = reqCopy.MaxCompletionTokens
+  } else if reqCopy.MaxTokens > 0 {
+    effTokens = reqCopy.MaxTokens
   }
+  reqCopy.MaxTokens = effTokens
+  reqCopy.MaxCompletionTokens = 0
 
   reqJSON, err := json.Marshal(reqCopy)
   if err != nil {
@@ -1541,6 +1546,10 @@ func (s *QwenService) chatDirectOpenAI(endpointURL string, w http.ResponseWriter
   }
 
   for k, vv := range resp.Header {
+    lowerKey := strings.ToLower(k)
+    if lowerKey == "content-length" || lowerKey == "transfer-encoding" || lowerKey == "connection" || lowerKey == "keep-alive" {
+      continue
+    }
     for _, v := range vv {
       w.Header().Add(k, v)
     }