AI_Watch_Platform/backend/app/services/llm_agent.py

import base64
import json
from openai import AsyncOpenAI
import logging
from typing import List, Any

logger = logging.getLogger(__name__)

async def analyze_image(
    base_url: str,
    api_key: str,
    model_name: str,
    rules: List[Any],
    image_bytes: bytes
) -> str:
    try:
        client = AsyncOpenAI(
            base_url=base_url,
            api_key=api_key,
        )
        
        # Prepare rules text
        # Handle both dicts and Pydantic models
        rules_data = []
        for r in rules:
            if hasattr(r, 'model_dump'):
                rules_data.append(r.model_dump())
            elif hasattr(r, 'dict'):
                rules_data.append(r.dict())
            else:
                rules_data.append(r)
                
        rules_text = json.dumps(rules_data, ensure_ascii=False, indent=2)
        
        system_prompt = """你是一个智能安防助手。
请根据提供的监控画面和告警规则列表，逐项检查是否触发告警。
请严格按照 JSON 格式返回结果，返回一个对象列表。
每个对象包含以下字段：
- "alarm_name": 告警规则名称 (String)
- "alarm_content": 详细描述发现的情况，如果没有异常请填 "正常" (String)
- "is_alarm": 是否触发告警 (Boolean, true/false)
- "area": 异常发生的区域描述，如果无异常可填 "全局" (String)

不要包含 markdown 格式标记（如 ```json），直接返回 JSON 字符串。
"""

        user_content = f"请根据以下规则检查画面：\n{rules_text}"

        b64_image = base64.b64encode(image_bytes).decode('utf-8')
        
        response = await client.chat.completions.create(
            model=model_name,
            messages=[
                {
                    "role": "system",
                    "content": system_prompt
                },
                {
                    "role": "user",
                    "content": [
                        {"type": "text", "text": user_content},
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpeg;base64,{b64_image}"
                            },
                        },
                    ],
                }
            ],
            max_tokens=1000
        )
        
        return response.choices[0].message.content
    except Exception as e:
        logger.error(f"LLM Error: {e}")
        return json.dumps([{"alarm_name": "System Error", "alarm_content": str(e), "is_alarm": True, "time": "", "area": ""}])

async def test_connection(
    base_url: str,
    api_key: str,
    model_name: str
) -> str:
    try:
        client = AsyncOpenAI(
            base_url=base_url,
            api_key=api_key,
        )
        
        response = await client.chat.completions.create(
            model=model_name,
            messages=[
                {"role": "user", "content": "Hello, this is a connection test."}
            ],
            max_tokens=10
        )
        
        return response.choices[0].message.content
    except Exception as e:
        logger.error(f"LLM Test Error: {e}")
        raise e