Merge remote-tracking branch 'origin/develop' into develop

ext/video_agent_deploy.py

@@ -57,7 +57,7 @@ vol = modal.Volume.from_name("comfy_model", environment_name="dev", create_if_mi
 )
 @modal.web_server(8000, startup_timeout=120, label='image-video-agent-1')
 def ui_1():
-    process = subprocess.Popen("comfy launch -- --cpu --listen 0.0.0.0 --port 8000", shell=True)
+    process = subprocess.Popen("comfy launch -- --cpu --listen 0.0.0.0 --port 8001", shell=True)
     process.wait()
 
 

nodes/img_agent.py

@@ -1,4 +1,4 @@
-# -*- coding:utf-8 -*-
+# -*- coding: utf-8 -*-
 """
 File                   img_agent.py
 Author                 silence
@@ -22,7 +22,7 @@ except ImportError:
     import logging
 
     logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
-    logger = logging.getLogger("ImgSubmitNode_Final")
+    logger = logging.getLogger("ImgSubmitNode")
     print("提示: loguru 未安装，使用内置logging。建议安装以获得更好的日志体验: pip install loguru")
 
 
@@ -102,6 +102,12 @@ class ImgSubmitNode:
 
     @classmethod
     def INPUT_TYPES(cls):
+        # 从用户提供的截图中提取的尺寸选项
+        size_options = [
+            '2048x2048 (1:1)', '2394x1728 (4:3)', '1728x2394 (3:4)', '2560x1440 (16:9)',
+            '1440x2560 (9:16)', '2496x1664 (3:2)', '1664x2496 (2:3)', '3024x1296 (21:9)',
+            '1K', '2K', '4K'
+        ]
         return {
             "required": {
                 "model_name_display": (cls.MODEL_DATA["full_display_list"],),
@@ -112,6 +118,10 @@ class ImgSubmitNode:
             "optional": {
                 "image": ("IMAGE",),
                 "image_filename": ("STRING", {"multiline": False, "default": ""}),
+                "image_urls": (
+                "STRING", {"multiline": True, "default": "", "placeholder": "单个或多个图片URL，用英文逗号隔开..."}),
+                "num_images": ("INT", {"default": 1, "min": 1, "max": 9, "step": 1}),
+                "image_size": (size_options, {"default": "2K"}),
             }
         }
 
@@ -134,59 +144,97 @@ class ImgSubmitNode:
         logger.info(f"环境: [{environment}], 模型: '{model_name_display}' -> '{tech_name}'")
         return base_url, tech_name
 
-    def submit_task(self, model_name_display, prompt, aspect_ratio, environment, image_filename=None, image=None):
+    def submit_task(self, model_name_display, prompt, aspect_ratio, environment,
+                    image_filename=None, image=None, image_urls="", num_images=1, image_size="2K"):
+        file_obj = None
         try:
             base_url, tech_name = self._get_base_url_and_tech_name(environment, model_name_display)
-            model_config = self.MODEL_DATA["configs"].get(tech_name)
-            if not model_config:
-                raise ValueError(f"无法找到模型 '{tech_name}' 的配置。")
-
-            def validate_and_correct_parameter(param_name, user_value, supported_values):
-                if not supported_values: return user_value
-                if user_value in supported_values: return user_value
-
-                default_value = supported_values[0]
-                logger.warning(
-                    f"参数警告！模型 '{tech_name}' 不支持 '{param_name}': '{user_value}'。"
-                    f"已自动替换为支持的默认值: '{default_value}'。支持的选项: {supported_values}"
-                )
-                return default_value
-
-            final_ar = validate_and_correct_parameter("宽高比", aspect_ratio, model_config.get("supported_ar", []))
-
-            headers = {'accept': 'application/json'}
-            payload = {'prompt': prompt, 'model_name': tech_name, 'aspect_ratio': final_ar,
-                       'mode': 'turbo', 'webhook_flag': 'false'}
-            files_to_send = {}
-            file_obj = None
-
-            if image is not None:
-                logger.info(f"检测到 IMAGE (Tensor) 输入，优先处理。")
-                img_tensor = image[0]
-                img_np = np.clip(255. * img_tensor.cpu().numpy(), 0, 255).astype(np.uint8)
-                pil_image = Image.fromarray(img_np)
-                buffer = io.BytesIO()
-                pil_image.save(buffer, format="PNG")
-                buffer.seek(0)
-                files_to_send['img_file'] = ('image_from_workflow.png', buffer, 'image/png')
-            elif image_filename and image_filename.strip():
-                logger.info(f"处理文件名: {image_filename}")
-                full_path = folder_paths.get_full_path("input", image_filename.strip())
-                if not (full_path and os.path.exists(full_path)):
-                    return (f"错误: 在ComfyUI的input文件夹中未找到文件 '{image_filename}'",)
-                filename = os.path.basename(full_path)
-                mime_type, _ = mimetypes.guess_type(full_path) or ('application/octet-stream', None)
-                file_obj = open(full_path, 'rb')
-                files_to_send['img_file'] = (filename, file_obj, mime_type)
-            else:
-                logger.info("未提供任何图像输入，以纯文本模式运行。")
-
             api_endpoint = f'{base_url}/api/custom/image/submit/task'
-            logger.info(f"向端点 {api_endpoint} 发送请求...")
+            headers = {'accept': 'application/json'}
+            if "doubao-seedream-4-0" in tech_name:
+                logger.info(f"检测到豆包模型 '{tech_name}'，启用特定提交逻辑。")
 
-            response = requests.post(
-                api_endpoint, headers=headers, data=payload, files=files_to_send, timeout=60
-            )
+                # 从尺寸选项中提取实际值（例如 '2048x2048 (1:1)' -> '2048x2048'）
+                actual_size = image_size.split(' ')[0]
+
+                extra_params = {
+                    "sequential_image_generation": "auto",
+                    "size": actual_size,
+                    "max_images": num_images
+                }
+
+                payload = {
+                    'prompt': prompt,
+                    'model_name': tech_name,
+                    'aspect_ratio': aspect_ratio,
+                    'mode': 'turbo',
+                    'webhook_flag': 'false',
+                    'watermark': 'false',
+                    'extra': json.dumps(extra_params)
+                }
+
+                if image_urls and image_urls.strip():
+                    logger.info(f"使用提供的图片URL: {image_urls.strip()}")
+                    image_urls = image_urls.strip()
+                    image_urls = image_urls.strip().replace('\n', ',')
+                    payload['img_list'] = image_urls
+                else:
+                    logger.info("未提供图片URL，将以文生图模式运行。")
+
+                logger.info(f"向豆包模型端点 {api_endpoint} 发送请求...")
+                response = requests.post(
+                    api_endpoint, headers=headers, data=payload, timeout=60
+                )
+
+            else:
+                # 原始逻辑，适用于其他所有模型
+                logger.info(f"使用标准逻辑提交任务到模型 '{tech_name}'。")
+                model_config = self.MODEL_DATA["configs"].get(tech_name)
+                if not model_config:
+                    raise ValueError(f"无法找到模型 '{tech_name}' 的配置。")
+
+                def validate_and_correct_parameter(param_name, user_value, supported_values):
+                    if not supported_values: return user_value
+                    if user_value in supported_values: return user_value
+
+                    default_value = supported_values[0]
+                    logger.warning(
+                        f"参数警告！模型 '{tech_name}' 不支持 '{param_name}': '{user_value}'。"
+                        f"已自动替换为支持的默认值: '{default_value}'。支持的选项: {supported_values}"
+                    )
+                    return default_value
+
+                final_ar = validate_and_correct_parameter("宽高比", aspect_ratio, model_config.get("supported_ar", []))
+
+                payload = {'prompt': prompt, 'model_name': tech_name, 'aspect_ratio': final_ar,
+                           'mode': 'turbo', 'webhook_flag': 'false'}
+                files_to_send = {}
+
+                if image is not None:
+                    logger.info(f"检测到 IMAGE (Tensor) 输入，优先处理。")
+                    img_tensor = image[0]
+                    img_np = np.clip(255. * img_tensor.cpu().numpy(), 0, 255).astype(np.uint8)
+                    pil_image = Image.fromarray(img_np)
+                    buffer = io.BytesIO()
+                    pil_image.save(buffer, format="PNG")
+                    buffer.seek(0)
+                    files_to_send['img_file'] = ('image_from_workflow.png', buffer, 'image/png')
+                elif image_filename and image_filename.strip():
+                    logger.info(f"处理文件名: {image_filename}")
+                    full_path = folder_paths.get_full_path("input", image_filename.strip())
+                    if not (full_path and os.path.exists(full_path)):
+                        return (f"错误: 在ComfyUI的input文件夹中未找到文件 '{image_filename}'",)
+                    filename = os.path.basename(full_path)
+                    mime_type, _ = mimetypes.guess_type(full_path) or ('application/octet-stream', None)
+                    file_obj = open(full_path, 'rb')
+                    files_to_send['img_file'] = (filename, file_obj, mime_type)
+                else:
+                    logger.info("未提供任何图像输入，以纯文本模式运行。")
+
+                logger.info(f"向标准端点 {api_endpoint} 发送请求...")
+                response = requests.post(
+                    api_endpoint, headers=headers, data=payload, files=files_to_send, timeout=60
+                )
             response.raise_for_status()
             response_json = response.json()
             logger.info(f"任务提交成功，完整响应: {json.dumps(response_json, indent=2, ensure_ascii=False)}")
@@ -203,10 +251,11 @@ class ImgSubmitNode:
             if file_obj:
                 file_obj.close()
 
-# NODE_CLASS_MAPPINGS = {
-#     "ImgSubmitNode": ImgSubmitNode
-# }
-#
-# NODE_DISPLAY_NAME_MAPPINGS = {
-#     "ImgSubmitNode": "统一生图任务节点"
-# }
+
+NODE_CLASS_MAPPINGS = {
+    "ImgSubmitNode": ImgSubmitNode
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ImgSubmitNode": "提交图片生成"
+}