Place audio after text in chat messages

app.py

@@ -93,14 +93,20 @@ def _resolve_media_source(path: str) -> str:
 
 
 def _user_content(text: str, files: list[str]) -> list[dict]:
-    """Build a user message content list from text and uploaded file paths."""
-    content: list[dict] = []
+    """Build a user message content list from text and uploaded file paths.
+
+    Media placement follows the model card's guidance: image and video go before
+    the text, audio goes after it.
+    """
+    before: list[dict] = []
+    after: list[dict] = []
     for path in files:
         kind = _classify_file(path)
-        if kind:
-            content.append({"type": kind, "url": _resolve_media_source(path)})
-    content.append({"type": "text", "text": text})
-    return content
+        if kind == "audio":
+            after.append({"type": kind, "url": _resolve_media_source(path)})
+        elif kind:
+            before.append({"type": kind, "url": _resolve_media_source(path)})
+    return [*before, {"type": "text", "text": text}, *after]
 
 
 def process_history(history: list[dict]) -> list[dict]: