update

README.md

@@ -2,11 +2,11 @@
 
 ## 1. 项目介绍
 
-ComfyUI-DeepSeek-Toolkit 是一个 ComfyUI 的自定义节点包，旨在帮助用户利用来自中国供应商的强大 AI 大语言模型 (LLM)，例如 DeepSeek、Qwen 等。该工具包致力于让更广泛的受众能够访问这些先进的模型，即使是那些计算资源有限的用户。
+ComfyUI-DeepSeek-Toolkit 是一个 ComfyUI 的自定义节点包，旨在帮助用户利用来自中国公司的强大 AI 大语言模型 (LLM)，例如 DeepSeek、Qwen 等。该工具包致力于让更广泛的受众能够访问这些先进的模型，即使是那些计算资源有限的用户。
 
 该项目的灵感来自于 DeepSeek 和 Qwen 等令人印象深刻的 LLM 的出现。它旨在通过提供轻量级和 API 驱动的解决方案，将这些前沿 AI 技术与日常用户连接起来，从而将这些模型集成到 ComfyUI 工作流程中。
 
-ComfyUI-DeepSeek-Toolkit 的使命是使开发人员和创意人员能够在现成的硬件（如旧笔记本电脑）上完成通常需要大量计算能力的任务。该项目设想了一个 AI 民主化并为所有人所用的未来。
+ComfyUI-DeepSeek-Toolkit 的使命是使开发人员和创意人员能够在现成的硬件（如旧笔记本电脑）上完成通常需要大量计算能力的任务。
 
 ## 2. 功能特性
 

__init__.py

@@ -4,6 +4,7 @@ from .py.image_understanding import NODE_CLASS_MAPPINGS as IMAGE_UNDERSTANDING_M
 from .py.model_loader import NODE_CLASS_MAPPINGS as MODEL_LOADER_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS as MODEL_LOADER_DISPLAY_MAPPINGS
 from .py.openai_compatible_loader import NODE_CLASS_MAPPINGS as OPENAI_COMPATIBLE_LOADER_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS as OPENAI_COMPATIBLE_LOADER_DISPLAY_MAPPINGS
 from .py.image_preprocessor import NODE_CLASS_MAPPINGS as IMAGE_PREPROCESSOR_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS as IMAGE_PREPROCESSOR_DISPLAY_MAPPINGS
+from .py.VideoFileUploader import NODE_CLASS_MAPPINGS as VIDEO_FILE_UPLOADER_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS as VIDEO_FILE_UPLOADER_DISPLAY_MAPPINGS # 新增导入
 
 # 合并所有节点映射
 NODE_CLASS_MAPPINGS = {
@@ -12,7 +13,8 @@ NODE_CLASS_MAPPINGS = {
     **IMAGE_UNDERSTANDING_MAPPINGS,
     **MODEL_LOADER_MAPPINGS,
     **OPENAI_COMPATIBLE_LOADER_MAPPINGS,
-    **IMAGE_PREPROCESSOR_MAPPINGS
+    **IMAGE_PREPROCESSOR_MAPPINGS,
+    **VIDEO_FILE_UPLOADER_MAPPINGS # 新增合并
 }
 
 NODE_DISPLAY_NAME_MAPPINGS = {
@@ -21,7 +23,8 @@ NODE_DISPLAY_NAME_MAPPINGS = {
     **IMAGE_UNDERSTANDING_DISPLAY_MAPPINGS,
     **MODEL_LOADER_DISPLAY_MAPPINGS,
     **OPENAI_COMPATIBLE_LOADER_DISPLAY_MAPPINGS,
-    **IMAGE_PREPROCESSOR_DISPLAY_MAPPINGS
+    **IMAGE_PREPROCESSOR_DISPLAY_MAPPINGS,
+    **VIDEO_FILE_UPLOADER_DISPLAY_MAPPINGS # 新增合并
 }
 
 WEB_DIRECTORY = "./web"

google_file_api.py

@@ -0,0 +1,79 @@
+import os
+import io
+import base64
+from typing import Tuple, Optional
+import google.generativeai as genai  # 导入 Gemini API
+import time
+import folder_paths
+import json
+
+class GoogleDriveUpload:
+    """
+    A ComfyUI node for uploading video files to Gemini via File API.
+    """
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "api_key": ("STRING", {"multiline": False, "default": ""}),  # API 密钥字符串
+                "file_path": ("STRING", {"default": "", "label": "Video File Path"}),
+                "file_name": ("STRING", {"default": "", "label": "File Name"}),
+                # "mime_type": (["video/mp4", "video/avi", "video/quicktime", "video/x-matroska"], {"default": "video/mp4", "label": "MIME Type"}), # Gemini File API 不需要mime_type
+                # "folder_id": ("STRING", {"default": "", "label": "Google Drive Folder ID", "description": "Optional: If empty, uploads to root."}), # Gemini File API 不需要folder_id
+            },
+        }
+
+    RETURN_TYPES = ("STRING",)
+    RETURN_NAMES = ("file_uri",)
+    FUNCTION = "upload_file"
+    CATEGORY = "DeepSeek_Toolkit/Google Drive"
+
+    def upload_file(self, api_key: str, file_path: str, file_name: str) -> Tuple[str]:
+        """
+        Uploads a file to Gemini via File API.
+
+        Args:
+            api_key (str): The Gemini API key.
+            file_path (str): The path to the file to upload.
+            file_name (str): The desired name for the file on Gemini.
+
+        Returns:
+            str: The URI of the uploaded file.
+        """
+        try:
+            # 1. 验证 API 密钥
+            if not api_key:
+                raise ValueError("API 密钥不能为空")
+
+            # 2. 验证文件路径
+            if not os.path.exists(file_path):
+                raise FileNotFoundError(f"文件未找到: {file_path}")
+
+            # 3. 构建 Gemini API 客户端
+            genai.configure(api_key=api_key)
+            model = genai.GenerativeModel(model_name="gemini-1.5-pro") # 假设使用 gemini-1.5-pro 模型
+            client = model
+
+            # 4. 上传文件
+            print("Uploading file...")
+            video_file = client.upload(path=file_path, name=file_name)
+            file_uri = video_file.uri
+            print(f"Completed upload: {file_uri}")
+
+            return (file_uri,)
+
+        except Exception as e:
+            print(f"上传文件时发生错误: {e}")
+            return (f"Error: {str(e)}",)
+
+
+# 节点类映射
+NODE_CLASS_MAPPINGS = {
+    "GoogleDriveUpload": GoogleDriveUpload
+}
+
+# 节点显示名称映射
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "GoogleDriveUpload": "Google Drive Upload (Gemini File API)"
+}

py/VideoFileUploader.py

@@ -0,0 +1,44 @@
+import folder_paths
+import nodes
+import os
+from typing import Tuple
+
+class VideoFileUploader:
+    """
+    Custom node for uploading video files to Gemini API
+    """
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "video_path": ("STRING", {"default": "", "label": "Video File Path"}),
+                "api_key": ("STRING", {"default": "", "label": "Gemini API Key"}), # 模拟API Key输入
+            }
+        }
+
+    RETURN_TYPES = ("STRING",)
+    RETURN_NAMES = ("video_url",)
+    FUNCTION = "upload_video"
+    CATEGORY = "DeepSeek_Toolkit"
+
+    def upload_video(self, video_path: str, api_key: str) -> Tuple[str]:
+        """
+        Uploads a video file to the Gemini API using the File API.
+        """
+        # 模拟文件上传
+        if not os.path.exists(video_path):
+            raise FileNotFoundError(f"Video file not found: {video_path}")
+
+        # 模拟上传过程
+        print(f"[INFO] 上传视频文件: {video_path}")
+        # 模拟生成视频URL
+        video_url = f"gemini://file/{os.path.basename(video_path)}"  # 模拟视频URL
+        print(f"[INFO] 视频上传完成，URL: {video_url}")
+
+        return (video_url,)
+
+
+# 注册节点
+NODE_CLASS_MAPPINGS = {"VideoFileUploader": VideoFileUploader}
+NODE_DISPLAY_NAME_MAPPINGS = {"VideoFileUploader": "Video File Uploader"}

py/__init__.py

@@ -3,4 +3,5 @@ from . import image_generation
 from . import image_understanding
 from . import model_loader
 from . import openai_compatible_loader
-from . import image_preprocessor
+from . import image_preprocessor
+from . import VideoFileUploader # 新增导入

py/openai_compatible_loader.py

@@ -27,6 +27,7 @@ class OpenAICompatibleLoader:
             },
             "optional": {
                 "prep_img": ("STRING", {"default": "", "forceInput": True}),
+                "video_url": ("STRING", {"default": "", "forceInput": True, "label": "Video URL"}), # 新增视频URL输入
                 "system_prompt": ("STRING", {"default": "你是一个AI大模型", "multiline": True}),
                 "prompt": ("STRING", {"multiline": True}),
                 "temperature": ("FLOAT", {"default": 0.7, "min": 0.0, "max": 2.0}),
@@ -70,7 +71,7 @@ class OpenAICompatibleLoader:
 
     def generate(self, base_url: str, api_key: str, prompt: str,
                  model: str, temperature: float,
-                 max_tokens: int, system_prompt: Optional[str] = None, prep_img: Optional[str] = None, enable_memory: bool = False):
+                 max_tokens: int, system_prompt: Optional[str] = None, prep_img: Optional[str] = None, video_url: Optional[str] = None, enable_memory: bool = False): # 新增video_url参数
         content = []  # 将 content 初始化放在函数的最开始
 
         # 移除图像参数类型的日志
@@ -85,6 +86,9 @@ class OpenAICompatibleLoader:
             # 将 prep_img 添加到 content 中
             content.append({"type": "image_url", "image_url": {"url": prep_img}})
 
+        if video_url: # 如果提供了视频URL，则将其添加到content中
+            content.append({"type": "video_url", "video_url": {"url": video_url}})
+
         if prompt.strip():
             content.append({"type": "text", "text": prompt})
 
@@ -101,7 +105,7 @@ class OpenAICompatibleLoader:
                 "content": content
             })
             # 移除带有图像内容的消息日志
-        elif not prompt.strip() and not system_prompt and prep_img is None: # 更精确的判断用户是否提供了有效输入
+        elif not prompt.strip() and not system_prompt and prep_img is None and video_url is None: # 更精确的判断用户是否提供了有效输入
             raise ValueError("用户输入的 prompt 不能为空")
 
         # 模型选择逻辑 (保持不变)
@@ -216,6 +220,9 @@ class OpenAICompatibleLoader:
             else:
                 image_tokens = 0  # 未知类型，默认为 0
             input_tokens += image_tokens
+        if video_url:
+            video_tokens = len(video_url) // 1000 # 估算token
+            input_tokens += video_tokens
 
         try:
             time.sleep(1)