feat: Auto模式按用户输入意图自动路由视频/图像/语音生成模型

backend/python/app/api/chat.py

@@ -33,6 +33,30 @@ router = APIRouter()
 
 logger = logging.getLogger(__name__)
 
+
+def detect_media_intent(text: str) -> str | None:
+    """根据用户输入判断是否为 图像/视频/语音 生成请求，返回对应能力或 None。
+
+    用于 Auto 模式：用户未指定具体模型时，依据意图自动匹配平台各类能力的默认模型。
+    疑问句（含「吗/呢/怎么/能否」或结尾问号）不触发，避免「你能生成视频吗」被误判。
+    """
+    if not text:
+        return None
+    t = text.lower().strip()
+    # 疑问句不触发媒体生成
+    if t.endswith(("？", "?")) or any(q in t for q in ("吗", "呢", "怎么", "如何", "能否", "是否可以", "可以吗", "会不会", "why", "how")):
+        return None
+    video_kw = ("视频", "短片", "影片", "vide", "动画视频", "一段视频", "做个视频", "生成视频")
+    image_kw = ("图片", "图像", "插画", "海报", "照片", "imag", "画一张", "画个", "生成图片", "一张图", "一幅画", "生成一幅")
+    audio_kw = ("语音", "配音", "朗读", "播报", "念一段", "有声", "tts", "音频", "生成语音")
+    if any(k in t for k in video_kw):
+        return "video"
+    if any(k in t for k in image_kw):
+        return "image"
+    if any(k in t for k in audio_kw):
+        return "audio"
+    return None
+
 # CST timezone (UTC+8) for consistent time display
 CST_TZ = timezone(timedelta(hours=8))
 
@@ -986,8 +1010,10 @@ async def send_message(
     if model_id and model_id != "auto":
         ai_model = db.query(AIModel).filter(AIModel.id == model_id).first()
     if not ai_model:
-        # Auto / 未指定 → 使用平台默认对话模型（文本能力默认）
+        # Auto / 未指定 → 按用户输入意图匹配对应能力的默认模型，否则回退文本对话
-        ai_model = resolve_auto_model("text", db)
+        intent = detect_media_intent(req.content)
+        resolved = resolve_auto_model(intent, db) if intent else None
+        ai_model = resolved or resolve_auto_model("text", db)
 
     if ai_model:
         # ====== 图像/视频生成模型分支 ======