myagent-ai 1.16.14 → 1.16.16

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "myagent-ai",
-  "version": "1.16.14",
+  "version": "1.16.16",
   "description": "本地桌面端执行型AI助手 - Open Interpreter 风格 | Local Desktop Execution-Oriented AI Assistant",
   "main": "main.py",
   "bin": {

package/web/api_server.py

@@ -150,7 +150,7 @@ def _agent_color(name: str) -> str:
 class ApiServer:
     def __init__(self, app_core):
         self.core = app_core
-        self.app = web.Application()
+        self.app = web.Application(client_max_size=50 * 1024 * 1024)  # 50MB，支持大文件上传（PDF/Word等）
         self._exec_progress: dict = {}
         # Wrap the executor to track progress in real-time (deferred to initialize())
         self._executor_wrapped = False
@@ -827,6 +827,8 @@ class ApiServer:
             try:
                 agent_cfg = self._read_agent_config(agent_path)
                 model_chain = self._build_model_chain(agent_cfg, agent_path)
+                # [v1.16.14] 有图片时，优先使用支持 vision 的模型
+                model_chain = self._reorder_model_chain_for_images(model_chain, bool(user_images))
                 logger.info(f"[{session_id}] model_chain={'有' if model_chain else '无'}, llm={'有' if self.core.llm else '无'}")
 
                 task_plan_context = self._build_task_plan_context(agent_path, chat_mode, message, session_id=session_id)
@@ -3572,6 +3574,38 @@ class ApiServer:
 
         return chain
 
+    def _reorder_model_chain_for_images(self, model_chain: list[dict], has_images: bool) -> list[dict]:
+        """[v1.16.14] 当消息包含图片时，将支持 vision 的模型优先排列
+
+        检查每个模型的 input_modes 字段，如果包含 "image"，则优先使用。
+        如果都没有 vision 能力，则保持原序（会触发 main_agent 的纯文本降级）。
+        """
+        if not has_images or not model_chain:
+            return model_chain
+
+        # 从 models_library 获取每个模型的 input_modes
+        vision_models = []
+        text_only_models = []
+        for mc in model_chain:
+            mc_id = mc.get("id", "")
+            input_modes = ["text"]  # 默认
+            if mc_id:
+                for me in self.core.config.models_library:
+                    if me.id == mc_id:
+                        input_modes = me.input_modes or ["text"]
+                        break
+            if "image" in input_modes:
+                vision_models.append(mc)
+            else:
+                text_only_models.append(mc)
+
+        if vision_models:
+            logger.info(f"消息含图片，优先使用 vision 模型: {[m.get('name', m.get('model')) for m in vision_models]}")
+            return vision_models + text_only_models
+        else:
+            logger.info("消息含图片，但模型链中无 vision 模型，将尝试当前主模型（失败后自动降级纯文本）")
+            return model_chain
+
     async def _try_model_chain(self, model_chain: list[dict], message: str, session_id: str,
                                  agent_path: str = None, agent_system_prompt: str = None,
                                  chat_mode: str = "") -> str: