Merge branch 'master' of https://github.com/MaiBot-Plus/MaiMbot-Pro-Max

src/chat/emoji_system/emoji_manager.py

@@ -925,7 +925,7 @@ class EmojiManager:
                         raise RuntimeError("GIF表情包转换失败")
                     prompt = "这是一个动态图表情包，每一张图代表了动态图的某一帧，黑色背景代表透明，描述一下表情包表达的情感和内容，描述细节，从互联网梗,meme的角度去分析"
                     description, _ = await self.vlm.generate_response_for_image(
-                        prompt, image_base64, "jpg", temperature=0.3, max_tokens=1000
+                        prompt, image_base64, "jpeg", temperature=0.3, max_tokens=1000
                     )
                 else:
                     prompt = (

src/chat/utils/utils_image.py

@@ -158,7 +158,7 @@ class ImageManager:
                     return "[表情包(GIF处理失败)]"
                 vlm_prompt = "这是一个动态图表情包，每一张图代表了动态图的某一帧，黑色背景代表透明，描述一下表情包表达的情感和内容，描述细节，从互联网梗,meme的角度去分析"
                 detailed_description, _ = await self.vlm.generate_response_for_image(
-                    vlm_prompt, image_base64_processed, "jpg", temperature=0.4, max_tokens=300
+                    vlm_prompt, image_base64_processed, "jpeg", temperature=0.4, max_tokens=300
                 )
             else:
                 vlm_prompt = (

src/llm_models/model_client/aiohttp_gemini_client.py

@@ -20,6 +20,29 @@ from ..payload_content.tool_option import ToolOption, ToolParam, ToolCall
 logger = get_logger("AioHTTP-Gemini客户端")
 
 
+def _format_to_mime_type(image_format: str) -> str:
+    """
+    将图片格式转换为正确的MIME类型
+    
+    Args:
+        image_format (str): 图片格式 (如 'jpg', 'png' 等)
+        
+    Returns:
+        str: 对应的MIME类型
+    """
+    format_mapping = {
+        "jpg": "image/jpeg",
+        "jpeg": "image/jpeg", 
+        "png": "image/png",
+        "webp": "image/webp",
+        "gif": "image/gif",
+        "heic": "image/heic",
+        "heif": "image/heif"
+    }
+    
+    return format_mapping.get(image_format.lower(), f"image/{image_format.lower()}")
+
+
 def _convert_messages(messages: list[Message]) -> tuple[list[dict], list[str] | None]:
     """
     转换消息格式 - 将消息转换为Gemini REST API所需的格式
@@ -46,7 +69,7 @@ def _convert_messages(messages: list[Message]) -> tuple[list[dict], list[str] |
                 if isinstance(item, tuple):  # (format, base64_data)
                     parts.append({
                         "inline_data": {
-                            "mime_type": f"image/{item[0].lower()}",
+                            "mime_type": _format_to_mime_type(item[0]),
                             "data": item[1]
                         }
                     })

src/llm_models/model_client/gemini_client.py

@@ -62,6 +62,25 @@ def _convert_messages(
     :return: 转换后的消息列表(和可能存在的system消息)
     """
     
+    def _get_correct_mime_type(image_format: str) -> str:
+        """
+        获取正确的MIME类型，修复jpg到jpeg的映射问题
+        :param image_format: 图片格式
+        :return: 正确的MIME类型
+        """
+        # 标准化格式名称，解决jpg/jpeg兼容性问题
+        format_mapping = {
+            "jpg": "jpeg",
+            "jpeg": "jpeg", 
+            "png": "png",
+            "webp": "webp",
+            "heic": "heic",
+            "heif": "heif",
+            "gif": "gif"
+        }
+        normalized_format = format_mapping.get(image_format.lower(), image_format.lower())
+        return f"image/{normalized_format}"
+
     def _convert_message_item(message: Message) -> Content:
         """
         转换单个消息格式，除了system和tool类型的消息
@@ -83,7 +102,7 @@ def _convert_messages(
             for item in message.content:
                 if isinstance(item, tuple):
                     content.append(
-                        Part.from_bytes(data=base64.b64decode(item[1]), mime_type=f"image/{item[0].lower()}")
+                        Part.from_bytes(data=base64.b64decode(item[1]), mime_type=_get_correct_mime_type(item[0]))
                     )
                 elif isinstance(item, str):
                     content.append(Part.from_text(text=item))

src/llm_models/utils_model.py

@@ -33,6 +33,38 @@ error_code_mapping = {
 }
 
 
+def _normalize_image_format(image_format: str) -> str:
+    """
+    标准化图片格式名称，确保与各种API的兼容性
+    
+    Args:
+        image_format (str): 原始图片格式
+        
+    Returns:
+        str: 标准化后的图片格式
+    """
+    format_mapping = {
+        "jpg": "jpeg",
+        "JPG": "jpeg", 
+        "JPEG": "jpeg",
+        "jpeg": "jpeg",
+        "png": "png",
+        "PNG": "png",
+        "webp": "webp",
+        "WEBP": "webp",
+        "gif": "gif",
+        "GIF": "gif",
+        "heic": "heic",
+        "HEIC": "heic",
+        "heif": "heif",
+        "HEIF": "heif"
+    }
+    
+    normalized = format_mapping.get(image_format, image_format.lower())
+    logger.debug(f"图片格式标准化: {image_format} -> {normalized}")
+    return normalized
+
+
 class RequestType(Enum):
     """请求类型枚举"""
 
@@ -78,6 +110,9 @@ class LLMRequest:
         Returns:
             (Tuple[str, str, str, Optional[List[ToolCall]]]): 响应内容、推理内容、模型名称、工具调用列表
         """
+        # 标准化图片格式以确保API兼容性
+        normalized_format = _normalize_image_format(image_format)
+        
         # 模型选择
         model_info, api_provider, client = self._select_model()
 
@@ -85,7 +120,7 @@ class LLMRequest:
         message_builder = MessageBuilder()
         message_builder.add_text_content(prompt)
         message_builder.add_image_content(
-            image_base64=image_base64, image_format=image_format, support_formats=client.get_support_image_formats()
+            image_base64=image_base64, image_format=normalized_format, support_formats=client.get_support_image_formats()
         )
         messages = [message_builder.build()]