feat: 聊天增强图像转述支持自定义 Provider id

astrbot/core/config/default.py

@@ -57,6 +57,7 @@ DEFAULT_CONFIG = {
         "group_icl_enable": False,
         "group_message_max_cnt": 300,
         "image_caption": False,
+        "image_caption_provider_id": "",
         "image_caption_prompt": "Please describe the image using Chinese.",
         "active_reply": {
             "enable": False,
@@ -723,6 +724,12 @@ CONFIG_METADATA_2 = {
                         "obvious_hint": True,
                         "hint": "启用后，当接收到图片消息时，会使用模型先将图片转述为文字再进行后续处理。推荐使用 gpt-4o-mini 模型。",
                     },
+                    "image_caption_provider_id": {
+                        "description": "图像转述提供商 ID",
+                        "type": "string",
+                        "obvious_hint": True,
+                        "hint": "可选。图像转述提供商 ID。如为空将选择聊天使用的提供商。",
+                    },
                     "image_caption_prompt": {
                         "description": "图像转述提示词",
                         "type": "string"

packages/astrbot/long_term_memory.py

@@ -25,6 +25,7 @@ class LongTermMemory:
             self.max_cnt = 300
         self.image_caption = self.config["image_caption"]
         self.image_caption_prompt = self.config["image_caption_prompt"]
+        self.image_caption_provider_id = self.config["image_caption_provider_id"]
         
         self.active_reply = self.config["active_reply"]
         self.enable_active_reply = self.active_reply.get("enable", False)
@@ -42,7 +43,13 @@ class LongTermMemory:
         return cnt
 
     async def get_image_caption(self, image_url: str) -> str:
-        provider = self.context.get_using_provider()
+        
+        if not self.image_caption_provider_id:
+            provider = self.context.get_using_provider()
+        else:
+            provider = self.context.get_provider_by_id(self.image_caption_provider_id)
+            if not provider:
+                raise Exception(f"没有找到 ID 为 {self.image_caption_provider_id} 的提供商")
         response = await provider.text_chat(
             prompt=self.image_caption_prompt,
             session_id=uuid.uuid4().hex,