Update ai-routes.js

ai-routes.js

@@ -456,7 +456,20 @@ router.post('/chat', async (req, res) => {
     // Build Current Message Parts
     const currentParts = [];
     
-    // Add Images first if any
+    // 1. Text Instruction (Priority)
+    // FIX: Put text FIRST to ensure model pays attention to instructions before processing media
+    if (text && text.trim()) {
+        currentParts.push({ text: text });
+    } else if (audio) {
+        // Fix: Explicit instruction for audio, placed before audio data if possible
+        currentParts.push({ text: "用户发送了语音消息，请听录音并回答。" });
+    } else {
+        // Fallback
+        currentParts.push({ text: "." });
+    }
+
+    // 2. Media (Images/Audio)
+    // Add Images
     if (images && Array.isArray(images)) {
         images.forEach(base64 => {
             if (base64) {
@@ -470,7 +483,7 @@ router.post('/chat', async (req, res) => {
         });
     }
 
-    // Add Audio if any
+    // Add Audio
     if (audio) {
         currentParts.push({
             inlineData: {
@@ -480,17 +493,6 @@ router.post('/chat', async (req, res) => {
         });
     }
 
-    // Add Text if any (must have text if no media, or can accompany media)
-    if (text && text.trim()) {
-        currentParts.push({ text: text });
-    } else if (audio) {
-        // Fix: If sending audio without text, add a prompt so the model knows what to do
-        currentParts.push({ text: "请听这段语音并回答。" });
-    } else if (currentParts.length === 0) {
-        // Fallback if empty
-        currentParts.push({ text: "." });
-    }
-
     // Prepare Response Stream
     res.setHeader('Content-Type', 'text/event-stream');
     res.setHeader('Cache-Control', 'no-cache');