Integrate Qwen3.8 model intake

src/server/request.rs

@@ -603,8 +603,10 @@ fn request_reasoning(
     let explicit_effort = match request.reasoning_effort.as_deref() {
         Some("none") => Some(ReasoningMode::Direct),
         Some("low") => Some(ReasoningMode::Low),
+        Some("medium") => Some(ReasoningMode::Medium),
         Some("high") => Some(ReasoningMode::High),
         Some("max") => Some(ReasoningMode::Max),
+        Some("xhigh") => Some(ReasoningMode::XHigh),
         None => None,
         Some(value) => return Err((400, format!("unsupported reasoning_effort: {value}"))),
     };
@@ -679,6 +681,10 @@ mod tests {
             request_reasoning(&default, ModelChoice::Glm53Flash, "").unwrap(),
             ReasoningMode::Max
         );
+        assert_eq!(
+            request_reasoning(&default, ModelChoice::Qwen38FlashNext, "").unwrap(),
+            ReasoningMode::XHigh
+        );
 
         for (effort, mode) in [
             ("none", ReasoningMode::Direct),
@@ -698,6 +704,32 @@ mod tests {
         }
 
         assert!(request_reasoning(&request(Some("medium"), None), ModelChoice::Glm52, "").is_err());
+        assert_eq!(
+            request_reasoning(
+                &request(Some("medium"), None),
+                ModelChoice::Qwen38FlashNext,
+                ""
+            )
+            .unwrap(),
+            ReasoningMode::Medium
+        );
+        assert_eq!(
+            request_reasoning(
+                &request(Some("xhigh"), None),
+                ModelChoice::Qwen38FlashNext,
+                ""
+            )
+            .unwrap(),
+            ReasoningMode::XHigh
+        );
+        assert!(
+            request_reasoning(
+                &request(Some("high"), None),
+                ModelChoice::Qwen38FlashNext,
+                ""
+            )
+            .is_err()
+        );
         assert!(
             request_reasoning(&request(None, Some(false)), ModelChoice::Glm53Flash, "").is_err()
         );