q3

src/lib/components/Piclets/PicletDetail.svelte

@@ -432,12 +432,6 @@
   
   
   /* Stats Tab */
-  .section-title {
-    margin: 0 0 16px;
-    font-size: 18px;
-    font-weight: bold;
-    color: #000;
-  }
   
   .stats-list {
     display: flex;

src/lib/services/qwen3Client.ts

@@ -68,15 +68,11 @@ export class Qwen3Client {
       throw new Error('Qwen3Client only supports "/chat" endpoint');
     }
 
+    // Note: Qwen3-Demo only uses these 3 parameters from the rwkv-compatible API
     const [
       message,
       chat_history = [],
-      system_prompt = "You are a helpful assistant.",
-      max_new_tokens = 2048,
-      temperature = 0.7,
-      top_p = 0.95,
-      top_k = 50,
-      repetition_penalty = 1.0
+      system_prompt = "You are a helpful assistant."
     ] = params;
 
     try {
@@ -84,14 +80,10 @@ export class Qwen3Client {
       await this.initializeGradioClient();
 
       // Use the proper Gradio Client API to call the add_message function
+      // Only pass parameters that actually exist in the Qwen3 Gradio app
       const response = await this.callQwen3API(message, {
         sys_prompt: system_prompt,
-        model: this.options.model,
-        max_new_tokens,
-        temperature,
-        top_p,
-        top_k,
-        repetition_penalty
+        model: this.options.model
       });
 
       // Return in the expected format: {data: [response_text]}
@@ -111,15 +103,12 @@ export class Qwen3Client {
         throw new Error('Gradio client not initialized');
       }
 
-      // Prepare settings for the Qwen3 space based on app.py structure
+      // Prepare settings for the Qwen3 space based on actual app.py structure
+      // Only use parameters that actually exist in the Gradio app
       const settingsFormValue = {
         model: options.model || this.options.model,
         sys_prompt: options.sys_prompt || "You are a helpful assistant.",
-        thinking_budget: Math.min(options.max_new_tokens || 20, 38), // Qwen3 has max 38k thinking budget
-        temperature: options.temperature || 0.7,
-        top_p: options.top_p || 0.95,
-        top_k: options.top_k || 50,
-        repetition_penalty: options.repetition_penalty || 1.0
+        thinking_budget: 38 // Use maximum thinking budget for best quality
       };
 
       // Thinking button state - disable for faster responses
@@ -242,37 +231,8 @@ export class Qwen3Client {
   }
 
   /**
-   * Test connection to Qwen3 service
+   * No connection testing - let natural failures trigger fallback to Zephyr-7B
    */
-  async testConnection(): Promise<boolean> {
-    try {
-      // Try to initialize the Gradio client first
-      await this.initializeGradioClient();
-      
-      // Test with a simple message
-      const result = await this.predict('/chat', [
-        'Hello, are you working? Please respond with just "Yes" if you can receive this message.',
-        [],
-        'You are a helpful assistant. Respond very briefly with just "Yes" if you can receive messages.',
-        50, // Small token limit for test
-        0.7,
-        0.95,
-        50,
-        1.0
-      ]);
-      
-      const response = result.data && result.data[0] && typeof result.data[0] === 'string' ? result.data[0] : '';
-      const isWorking = response.length > 0 && !response.includes('temporarily unavailable');
-      
-      console.log(`🔍 Qwen3 connection test result: ${isWorking ? 'PASS' : 'FAIL'}`);
-      console.log(`📝 Test response: "${response.substring(0, 50)}..."`);
-      
-      return isWorking;
-    } catch (error) {
-      console.error('Qwen3 connection test failed:', error);
-      return false;
-    }
-  }
 }
 
 // Export a default instance

src/lib/services/textGenerationClient.ts

@@ -8,7 +8,6 @@ import { qwen3Client } from './qwen3Client';
 
 interface TextGenerationClient {
   predict(endpoint: string, params: any[]): Promise<{data: any[]}>;
-  testConnection?(): Promise<boolean>;
 }
 
 class TextGenerationManager {