调用llama3.2-vision,自动识别图片，输出prompt

app/service/prompt_generation/chatgpt_for_translation.py

@@ -9,6 +9,7 @@ from retry import retry
 
 from app.core.config import QWEN_API_KEY
 from app.service.chat_robot.script.service.CallQWen import get_language
+from app.service.prompt_generation.util import minio_util
 
 logger = logging.getLogger(__name__)
 
@@ -143,6 +144,38 @@ def get_translation_from_llama3(text):
 #     response = requests.post(url, data=json.dumps(payload), headers=headers)
 
 
+def get_prompt_from_image(image_path, text):
+    start_time = time.time()
+    # url = "http://localhost:11434/api/generate"
+    url = "http://10.1.1.243:11434/api/generate"
+
+    image_base64 = minio_util.minio_url_to_base64(image_path.img)
+    # image_base64 = minio_url_to_base64(image_path)
+
+    # 创建请求的负载 translator是自定义的翻译模型
+    payload = {
+        "model": "llama3.2-vision",
+        "images": [image_base64],
+        "prompt": f"{text}",
+        "stream": False
+    }
+    # 将负载转换为 JSON 格式
+    headers = {'Content-Type': 'application/json'}
+    response = requests.post(url, data=json.dumps(payload), headers=headers)
+    # 处理响应
+    if response.status_code == 200:
+        # print("Response from server:")
+        # print(response.json())
+        resp = json.loads(response.content).get("response")
+        logger.info(f"sketch re-generate server runtime is {time.time() - start_time} \n, response is {resp}")
+        # print("input : {}, sketch re-generate result : {}".format(text, resp))
+        return resp
+    else:
+        logger.info(f"sketch re-generate server runtime is {time.time() - start_time} , response is {response.content}")
+        print(f"Request failed with status code {response.status_code}")
+        print(response.text)
+
+
 def main():
     """Main function"""
     text = get_translation_from_llama3("[火焰]")