Update app.py

app.py

@@ -182,6 +182,51 @@ def decode_base64_file(data_url: str) -> tuple[str, str, str]:
         logger.error(f"Failed to parse data URL: {e}")
         return None, None, None
 
+async def download_image_from_url(url: str) -> str:
+    """
+    从URL下载图片并转换为base64
+    返回base64编码的图片数据
+    """
+    try:
+        logger.info(f"Downloading image from URL: {url}")
+        
+        async with aiohttp.ClientSession() as session:
+            async with session.get(url, timeout=30) as response:
+                if response.status == 200:
+                    image_bytes = await response.read()
+                    
+                    # 检测图片格式
+                    content_type = response.headers.get('content-type', '')
+                    if not content_type.startswith('image/'):
+                        # 尝试从文件扩展名推断
+                        if url.lower().endswith(('.jpg', '.jpeg')):
+                            content_type = 'image/jpeg'
+                        elif url.lower().endswith('.png'):
+                            content_type = 'image/png'
+                        elif url.lower().endswith('.gif'):
+                            content_type = 'image/gif'
+                        elif url.lower().endswith('.webp'):
+                            content_type = 'image/webp'
+                        else:
+                            content_type = 'image/jpeg'  # 默认
+                    
+                    # 转换为base64
+                    base64_data = base64.b64encode(image_bytes).decode('utf-8')
+                    data_url = f"data:{content_type};base64,{base64_data}"
+                    
+                    logger.info(f"Successfully downloaded image, size: {len(image_bytes)} bytes, base64 size: {len(base64_data)} chars")
+                    return data_url
+                else:
+                    logger.error(f"Failed to download image: HTTP {response.status}")
+                    return None
+                    
+    except asyncio.TimeoutError:
+        logger.error(f"Timeout downloading image from {url}")
+        return None
+    except Exception as e:
+        logger.error(f"Error downloading image from {url}: {e}")
+        return None
+
 async def upload_image_to_imgbb(base64_data: str) -> str:
     """
     将 base64 图片上传到 imgbb
@@ -280,6 +325,25 @@ def format_image_as_data_url(base64_data: str) -> str:
         logger.warning(f"Failed to detect image format: {e}, using JPEG as default")
         return f"data:image/jpeg;base64,{base64_data}"
 
+def extract_images_from_context(content: str) -> List[str]:
+    """
+    从系统上下文中提取图片URL
+    """
+    images = []
+    try:
+        # 查找类似 <image name="..." url="..."></image> 的标签
+        import re
+        pattern = r'<image[^>]+url="([^"]+)"[^>]*></image>'
+        matches = re.findall(pattern, content)
+        for url in matches:
+            if url.startswith('http'):
+                images.append(url)
+                logger.info(f"Found image URL in context: {url}")
+    except Exception as e:
+        logger.error(f"Error extracting images from context: {e}")
+    
+    return images
+
 def extract_content_from_message(message: Dict[str, Any]) -> tuple[str, List[str], List[Dict[str, str]]]:
     """
     从消息中提取文本内容、图片和文件
@@ -290,7 +354,10 @@ def extract_content_from_message(message: Dict[str, Any]) -> tuple[str, List[str
     files = []
     
     if isinstance(content, str):
-        # 简单文本消息
+        # 检查文本内容中是否包含系统上下文中的图片
+        context_images = extract_images_from_context(content)
+        if context_images:
+            images.extend(context_images)
         return content, images, files
     elif isinstance(content, list):
         # 复合消息（文本 + 图片 + 文件）
@@ -301,7 +368,12 @@ def extract_content_from_message(message: Dict[str, Any]) -> tuple[str, List[str
                 item_type = item.get("type", "")
                 
                 if item_type == "text":
-                    text_parts.append(item.get("text", ""))
+                    text_content = item.get("text", "")
+                    text_parts.append(text_content)
+                    # 检查文本中的上下文图片
+                    context_images = extract_images_from_context(text_content)
+                    if context_images:
+                        images.extend(context_images)
                 
                 elif item_type == "image_url":
                     image_url = item.get("image_url", {})
@@ -319,8 +391,12 @@ def extract_content_from_message(message: Dict[str, Any]) -> tuple[str, List[str
                                 logger.warning(f"Image URL format not supported: {url[:100]}...")
                         except Exception as e:
                             logger.error(f"Error processing image: {e}")
+                    elif url.startswith("http"):
+                        # 外部图片URL
+                        images.append(url)
+                        logger.info(f"Found external image URL: {url}")
                     else:
-                        logger.warning(f"External image URLs not supported: {url}")
+                        logger.warning(f"Unsupported image URL format: {url}")
                 
                 elif item_type == "file" or (item_type == "image_url" and not item.get("image_url", {}).get("url", "").startswith("data:image/")):
                     # 处理文件上传
@@ -349,6 +425,10 @@ def extract_content_from_message(message: Dict[str, Any]) -> tuple[str, List[str
             
             elif isinstance(item, str):
                 text_parts.append(item)
+                # 检查文本中的上下文图片
+                context_images = extract_images_from_context(item)
+                if context_images:
+                    images.extend(context_images)
         
         return " ".join(text_parts), images, files
     
@@ -450,11 +530,24 @@ async def transform_openai_to_replicate(openai_request: Dict[str, Any], model_ov
         formatted_image = None
         if has_images and primary_image:
             logger.info(f"Processing image for model {model} with format {model_config.get('image_format')}")
-            formatted_image = await format_image_for_model(primary_image, model_config)
             
-            if not formatted_image:
-                logger.error("Failed to format image for model")
-                raise HTTPException(status_code=500, detail="Failed to process image")
+            # 如果是外部URL，先下载转换为base64
+            if primary_image.startswith("http"):
+                logger.info(f"Downloading external image: {primary_image}")
+                downloaded_image = await download_image_from_url(primary_image)
+                if downloaded_image:
+                    primary_image = downloaded_image
+                    logger.info("External image downloaded and converted to base64")
+                else:
+                    logger.error("Failed to download external image")
+                    primary_image = None
+            
+            if primary_image:
+                formatted_image = await format_image_for_model(primary_image, model_config)
+                
+                if not formatted_image:
+                    logger.error("Failed to format image for model")
+                    raise HTTPException(status_code=500, detail="Failed to process image")
         
         # 构建 Replicate 格式的输入
         replicate_input = {}
@@ -661,10 +754,11 @@ async def root():
         "status": "running",
         "replicate_token_configured": bool(REPLICATE_API_TOKEN),
         "imgbb_token_configured": bool(IMGBB_API_KEY),
-        "version": "1.2.2",
+        "version": "1.3.0",
         "supported_models": list(MODEL_CONFIGS.keys()),
         "vision_support": True,
         "file_support": True,
+        "external_image_support": True,
         "supported_text_files": list(SUPPORTED_TEXT_EXTENSIONS),
         "supported_image_files": list(SUPPORTED_IMAGE_EXTENSIONS),
         "claude4_vision_support": "Full support via imgbb image hosting"