""" glm-4.5v 多模态功能测试 """ import requests import json # 创建一个1x1像素的红色图片作为测试 tiny_red_image = ( "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8DwHwAFBQIAX8jx0gAAAABJRU5ErkJggg==" ) # API配置 api_url = "http://localhost:8080/v1/chat/completions" api_key = "sk-your-api-key" # 构建正确的多模态请求 request_data = { "model": "glm-4.5v", # 使用多模态模型 "messages": [ { "role": "user", "content": [ # content必须是数组 { "type": "text", "text": "这是什么颜色的图片?" }, { "type": "image_url", "image_url": { "url": tiny_red_image } } ] } ], "stream": False } print("发送的请求:") print(json.dumps(request_data, indent=2, ensure_ascii=False)) print("\n" + "="*60) # 发送请求 headers = { "Authorization": f"Bearer {api_key}", "Content-Type": "application/json" } try: response = requests.post(api_url, json=request_data, headers=headers) print(f"响应状态码: {response.status_code}") if response.status_code == 200: result = response.json() print("\n模型回复:") print(result["choices"][0]["message"]["content"]) else: print("\n错误响应:") print(response.text) except Exception as e: print(f"\n发生错误: {e}")