| 1 | import os |
| 2 | import sys |
| 3 | |
| 4 | models_dir = os.path.dirname(os.path.abspath(__file__)) |
| 5 | backend_dir = os.path.dirname(models_dir) |
| 6 | if backend_dir not in sys.path: |
| 7 | sys.path.insert(0, backend_dir) |
| 8 | |
| 9 | import json |
| 10 | import logging |
| 11 | import time |
| 12 | import uuid |
| 13 | import dashscope |
| 14 | from dashscope import MultiModalConversation |
| 15 | from dashscope.aigc.image_generation import ImageGeneration |
| 16 | from config import Config |
| 17 | try: |
| 18 | from models.image_processor import ImageProcessor |
| 19 | except ImportError: |
| 20 | from image_processor import ImageProcessor |
| 21 | |
| 22 | class DashScopeClient: |
| 23 | def __init__(self, api_key=None, base_url=None): |
| 24 | self.api_key = api_key or Config.DASHSCOPE_API_KEY |
| 25 | # 默认使用中国（北京）地域 API，如果参数或 config.yaml 未设置则使用默认地址 |
| 26 | self.base_url = base_url or Config.DASHSCOPE_BASE_URL |
| 27 | dashscope.api_key = self.api_key |
| 28 | dashscope.base_http_api_url = self.base_url |
| 29 | self.image_processor = ImageProcessor() |
| 30 | |
| 31 | def generate_image(self, prompt, model="wan2.7-image", size="1024*1024", n=1, session_id=None, save_dir=None): |
| 32 | """ |
| 33 | Text to Image generation using DashScope |
| 34 | """ |
| 35 | try: |
| 36 | messages = [{"role": "user", "content": [{"text": prompt}]}] |
| 37 | response = ImageGeneration.call( |
| 38 | model=model, |
| 39 | api_key=self.api_key, |
| 40 | messages=messages, |
| 41 | n=n, |
| 42 | size=size, |
| 43 | watermark=False, |
| 44 | ) |
| 45 | |
| 46 | if response.status_code == 200: |
| 47 | results = [] |
| 48 | try: |
| 49 | # 标准 ImageGeneration 返回结果解析 |
| 50 | if response.output and response.output.choices: |
| 51 | for item in response.output.choices: |
| 52 | if 'message' in item and 'content' in item['message']: |
| 53 | results.append(item['message']['content'][0]['image']) |
| 54 | except Exception as e: |
| 55 | logging.error(f"Failed to parse ImageGeneration outputs: {e}") |
| 56 | |
| 57 | # Check if we should download |
| 58 | if save_dir: |
| 59 | os.makedirs(save_dir, exist_ok=True) |
| 60 | local_files = [] |
| 61 | for i, url in enumerate(results): |
| 62 | file_name = f"ds_{session_id if session_id else 'nosess'}_{int(time.time())}_{i}_{uuid.uuid4().hex[:6]}.png" |
| 63 | file_path = os.path.join(save_dir, file_name) |
| 64 | if self.image_processor.download_image(url, file_path, proxies=Config.requests_proxies("dashscope")): |
| 65 | local_files.append(file_path) |
| 66 | return local_files |
| 67 | |
| 68 | return results |
| 69 | else: |
| 70 | error_msg = f"{response.code}, {response.message}, status={response.status_code}" |
| 71 | logging.error(f"Image generation failed: {error_msg}") |
| 72 | raise RuntimeError(error_msg) |
| 73 | except Exception as e: |
| 74 | logging.error(f"Error in generate_image (DashScope): {e}") |
| 75 | raise |
| 76 | |
| 77 | def edit_image(self, prompt, image_urls, model="wan2.7-image", size="1920*1080", n=1, session_id=None, save_dir=None): |
| 78 | """ |
| 79 | Image editing/compositing using DashScope ImageGeneration |
| 80 | """ |
| 81 | # Prepare content |
| 82 | content_list = [] |
| 83 | for img_url in image_urls: |
| 84 | content_list.append({"image": img_url}) |
| 85 | content_list.append({"text": prompt}) |
| 86 | |
| 87 | messages = [ |
| 88 | { |
| 89 | "role": "user", |
| 90 | "content": content_list |
| 91 | } |
| 92 | ] |
| 93 | |
| 94 | try: |
| 95 | # Use ImageGeneration.call with messages, same as generate_image |
| 96 | response = ImageGeneration.call( |
| 97 | model=model, |
| 98 | api_key=self.api_key, |
| 99 | messages=messages, |
| 100 | n=n, |
| 101 | size=size, |
| 102 | watermark=False, |
| 103 | ) |
| 104 | |
| 105 | if response.status_code == 200: |
| 106 | results = [] |
| 107 | try: |
| 108 | # 标准 ImageGeneration 返回结果解析 |
| 109 | if response.output and response.output.choices: |
| 110 | for item in response.output.choices: |
| 111 | # 简化解析逻辑以处理多张图片的返回结构 |
| 112 | if isinstance(item, dict): |
| 113 | if 'image' in item: # 部分新模型直接返回 {'image': 'url', 'finish_reason': ...} |
| 114 | results.append(item['image']) |
| 115 | elif 'url' in item: |
| 116 | results.append(item['url']) |
| 117 | elif 'message' in item and 'content' in item['message']: # 兼容 Message 结构 |
| 118 | content = item['message']['content'] |
| 119 | if isinstance(content, list): |
| 120 | for c in content: |
| 121 | if isinstance(c, dict) and 'image' in c: |
| 122 | results.append(c['image']) |
| 123 | except Exception as e: |
| 124 | logging.error(f"Failed to parse ImageGeneration outputs: {e}") |
| 125 | |
| 126 | # Check if we should download |
| 127 | if save_dir: |
| 128 | os.makedirs(save_dir, exist_ok=True) |
| 129 | local_files = [] |
| 130 | for i, url in enumerate(results): |
| 131 | file_name = f"ds_{session_id if session_id else 'nosess'}_{int(time.time())}_{i}_{uuid.uuid4().hex[:6]}.png" |
| 132 | file_path = os.path.join(save_dir, file_name) |
| 133 | if self.image_processor.download_image(url, file_path, proxies=Config.requests_proxies("dashscope")): |
| 134 | local_files.append(file_path) |
| 135 | return local_files |
| 136 | |
| 137 | return results |
| 138 | else: |
| 139 | error_msg = f"{response.code}, {response.message}, status={response.status_code}" |
| 140 | logging.error(f"Image edit failed: {error_msg}") |
| 141 | raise RuntimeError(error_msg) |
| 142 | except Exception as e: |
| 143 | logging.error(f"Error in edit_image: {e}") |
| 144 | raise |
| 145 | |
| 146 | |
| 147 | if __name__ == "__main__": |
| 148 | import sys |
| 149 | sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) |
| 150 | from config import Config |
| 151 | |
| 152 | print("=== DashScope 图片生成可用性测试 ===") |
| 153 | MODELS=["wan2.6-t2i", "wan2.7-image", "wan2.7-image-pro"] |
| 154 | save_dir = "code/result/image/test_avail" |
| 155 | api_key = Config.DASHSCOPE_API_KEY |
| 156 | base_url = Config.DASHSCOPE_BASE_URL |
| 157 | if not api_key: |
| 158 | print("✗ DASHSCOPE_API_KEY 未设置，跳过") |
| 159 | sys.exit(1) |
| 160 | print(f" API Key: {api_key[:6]}***{api_key[-4:]}") |
| 161 | print(f" Base URL: {base_url}") |
| 162 | client = DashScopeClient(api_key=api_key, base_url=base_url) |
| 163 | |
| 164 | # 文生图 |
| 165 | print("\n=== 文生图测试 ===") |
| 166 | prompt = "一只橘猫躺在阳光下的窗台上，水彩画风格" |
| 167 | for model in MODELS: |
| 168 | print(f"\nPrompt: {prompt}") |
| 169 | print(f"model: {model}") |
| 170 | os.makedirs(save_dir, exist_ok=True) |
| 171 | t0 = time.time() |
| 172 | try: |
| 173 | paths = client.generate_image( |
| 174 | prompt=prompt, model=model, |
| 175 | size="1024*1024", save_dir=save_dir, |
| 176 | ) |
| 177 | elapsed = time.time() - t0 |
| 178 | if paths: |
| 179 | print(f"✓ 生成 {len(paths)} 张图片 ({elapsed:.1f}s): {paths}") |
| 180 | else: |
| 181 | print(f"✗ 返回空列表 ({elapsed:.1f}s)") |
| 182 | except Exception as e: |
| 183 | print(f"✗ 失败: {e}") |
| 184 | sys.exit(1) |
| 185 | |
| 186 | # 图生图 |
| 187 | print("\n=== 图生图测试 ===") |
| 188 | img_path = "code/result/image/test_avail/test_input.png" |
| 189 | prompt = "在这张图片的基础上，添加一些飞舞的樱花花瓣，绘制为水彩画风格" |
| 190 | for model in MODELS: |
| 191 | print(f"\nPrompt: {prompt}") |
| 192 | print(f"model: {model}") |
| 193 | os.makedirs(save_dir, exist_ok=True) |
| 194 | t0 = time.time() |
| 195 | try: |
| 196 | paths = client.edit_image( |
| 197 | prompt=prompt, image_urls=[img_path], model=model, |
| 198 | size="1024*1024", save_dir=save_dir, |
| 199 | ) |
| 200 | elapsed = time.time() - t0 |
| 201 | if paths: |
| 202 | print(f"✓ 生成 {len(paths)} 张图片 ({elapsed:.1f}s): {paths}") |
| 203 | else: |
| 204 | print(f"✗ 返回空列表 ({elapsed:.1f}s)") |
| 205 | except Exception as e: |
| 206 | print(f"✗ 失败: {e}") |
| 207 | sys.exit(1) |
| 208 |