209 lines
7.3 KiB
Python
209 lines
7.3 KiB
Python
"""
|
||
场景推荐服务
|
||
"""
|
||
import time
|
||
from typing import List
|
||
from app.schemas.value import (
|
||
ScenarioRecommendationRequest,
|
||
ScenarioRecommendationResponse,
|
||
CompanyInfo,
|
||
DataAsset,
|
||
ExistingScenario,
|
||
RecommendedScenario,
|
||
)
|
||
from app.utils.llm_client import llm_client
|
||
from app.utils.logger import logger
|
||
from app.core.config import settings
|
||
from app.core.exceptions import LLMAPIException
|
||
|
||
|
||
# ==================== 提示词模板 ====================
|
||
|
||
SYSTEM_PROMPT = """你是一位专业的数据应用场景规划专家，擅长基于企业背景、数据资产清单和存量场景，智能推荐潜在的数据应用场景。
|
||
|
||
## 你的专业能力
|
||
- 深入理解各行业的数据应用场景和最佳实践
|
||
- 熟悉数据资产的价值评估和场景依赖分析
|
||
- 能够识别高价值的数据应用场景
|
||
- 具备优秀的场景规划和推荐能力
|
||
|
||
## 输出要求
|
||
1. **准确性**：场景推荐必须基于实际的数据资产
|
||
2. **价值性**：推荐场景必须具有明确的商业价值
|
||
3. **可行性**：场景实施难度评估必须合理
|
||
4. **专业性**：使用专业术语，符合行业标准
|
||
5. **结构化**：严格按照JSON格式输出
|
||
"""
|
||
|
||
|
||
def build_scenario_recommendation_prompt(
|
||
company_info: CompanyInfo,
|
||
data_assets: List[DataAsset],
|
||
existing_scenarios: List[ExistingScenario],
|
||
recommendation_count: int
|
||
) -> str:
|
||
"""构建场景推荐提示词"""
|
||
|
||
# 格式化企业信息
|
||
industry_str = "、".join(company_info.industry)
|
||
|
||
# 格式化数据资产
|
||
assets_info = "\n".join([
|
||
f"- {asset.name}：{asset.description}\n 核心表：{', '.join(asset.core_tables)}"
|
||
for asset in data_assets
|
||
])
|
||
|
||
# 格式化存量场景
|
||
scenarios_info = "\n".join([
|
||
f"- {scenario.name}：{scenario.description}"
|
||
for scenario in existing_scenarios
|
||
])
|
||
|
||
prompt = f"""请基于以下信息推荐潜在的数据应用场景：
|
||
|
||
## 企业信息
|
||
行业: {industry_str}
|
||
企业描述: {company_info.description}
|
||
数据规模: {company_info.data_scale}
|
||
数据来源: {', '.join(company_info.data_sources)}
|
||
|
||
## 可用数据资产
|
||
{assets_info}
|
||
|
||
## 存量场景（避免重复推荐）
|
||
{scenarios_info}
|
||
|
||
## 推荐要求
|
||
1. 推荐 {recommendation_count} 个潜在数据应用场景
|
||
2. 场景分类：降本增效、营销增长、金融服务、决策支持、风险控制等
|
||
3. 推荐指数评分：1-5星（综合考虑业务价值、实施难度、数据准备度）
|
||
4. 分析场景依赖的数据资产
|
||
5. 评估商业价值和实施难度
|
||
6. 避免与存量场景重复
|
||
|
||
## 输出格式（JSON）
|
||
{{
|
||
"recommended_scenarios": [
|
||
{{
|
||
"id": 1,
|
||
"name": "场景名称",
|
||
"type": "场景分类",
|
||
"recommendation_index": 5,
|
||
"desc": "场景详细描述",
|
||
"dependencies": ["依赖的资产1", "依赖的资产2"],
|
||
"business_value": "商业价值描述",
|
||
"implementation_difficulty": "实施难度（低/中/高）",
|
||
"estimated_roi": "预估ROI（低/中/高）",
|
||
"technical_requirements": ["技术要求1", "技术要求2"],
|
||
"data_requirements": ["数据要求1", "数据要求2"]
|
||
}}
|
||
]
|
||
}}
|
||
"""
|
||
return prompt
|
||
|
||
|
||
# ==================== 主要服务类 ====================
|
||
|
||
class ScenarioRecommendationService:
|
||
"""场景推荐服务"""
|
||
|
||
@staticmethod
|
||
async def recommend(request: ScenarioRecommendationRequest) -> dict:
|
||
"""
|
||
推荐潜在场景
|
||
|
||
Args:
|
||
request: 场景推荐请求
|
||
|
||
Returns:
|
||
推荐结果
|
||
"""
|
||
start_time = time.time()
|
||
|
||
logger.info(
|
||
f"开始场景推荐 - 项目ID: {request.project_id}, "
|
||
f"资产数: {len(request.data_assets)}, 存量场景数: {len(request.existing_scenarios)}"
|
||
)
|
||
|
||
try:
|
||
# 获取配置
|
||
model = request.options.model if request.options else settings.DEFAULT_LLM_MODEL
|
||
temperature = settings.DEFAULT_TEMPERATURE
|
||
count = request.options.recommendation_count if request.options else 10
|
||
exclude_types = request.options.exclude_types if request.options else []
|
||
|
||
logger.info(f"使用模型: {model}, 推荐数量: {count}")
|
||
|
||
# 构建提示词
|
||
prompt = build_scenario_recommendation_prompt(
|
||
company_info=request.company_info,
|
||
data_assets=request.data_assets,
|
||
existing_scenarios=request.existing_scenarios,
|
||
recommendation_count=count
|
||
)
|
||
|
||
logger.debug(f"提示词长度: {len(prompt)} 字符")
|
||
|
||
# 调用大模型
|
||
response_text = await llm_client.call(
|
||
prompt=prompt,
|
||
system_prompt=SYSTEM_PROMPT,
|
||
temperature=temperature,
|
||
model=model
|
||
)
|
||
|
||
# 解析结果
|
||
llm_result = llm_client.parse_json_response(response_text)
|
||
logger.info("大模型返回结果解析成功")
|
||
|
||
# 转换为标准格式
|
||
recommended_scenarios = []
|
||
scenarios_data = llm_result.get("recommended_scenarios", [])
|
||
|
||
for idx, scenario_data in enumerate(scenarios_data):
|
||
# 过滤排除的场景类型
|
||
if exclude_types and scenario_data.get("type") in exclude_types:
|
||
continue
|
||
|
||
scenario = RecommendedScenario(
|
||
id=scenario_data.get("id", idx + 1),
|
||
name=scenario_data.get("name", ""),
|
||
type=scenario_data.get("type", ""),
|
||
recommendation_index=scenario_data.get("recommendation_index", 3),
|
||
desc=scenario_data.get("desc", ""),
|
||
dependencies=scenario_data.get("dependencies", []),
|
||
business_value=scenario_data.get("business_value", ""),
|
||
implementation_difficulty=scenario_data.get("implementation_difficulty", "中等"),
|
||
estimated_roi=scenario_data.get("estimated_roi", "中"),
|
||
technical_requirements=scenario_data.get("technical_requirements", []),
|
||
data_requirements=scenario_data.get("data_requirements", [])
|
||
)
|
||
recommended_scenarios.append(scenario)
|
||
|
||
# 计算生成时间
|
||
generation_time = time.time() - start_time
|
||
|
||
# 构建响应数据
|
||
response_data = {
|
||
"recommended_scenarios": [scenario.dict() for scenario in recommended_scenarios],
|
||
"total_count": len(recommended_scenarios),
|
||
"generation_time": round(generation_time, 2),
|
||
"model_used": model
|
||
}
|
||
|
||
logger.info(
|
||
f"场景推荐完成 - 推荐数: {len(recommended_scenarios)}, "
|
||
f"耗时: {generation_time:.2f}秒"
|
||
)
|
||
|
||
return response_data
|
||
|
||
except Exception as e:
|
||
logger.exception(f"场景推荐失败: {str(e)}")
|
||
raise LLMAPIException(
|
||
f"场景推荐失败: {str(e)}",
|
||
error_detail=str(e),
|
||
retryable="Rate limit" in str(e) or "timeout" in str(e).lower()
|
||
)
|