897 lines
34 KiB
Python
897 lines
34 KiB
Python
"""
|
||
AI分析服务
|
||
处理Gemini API集成和AI分析功能
|
||
"""
|
||
|
||
from typing import Dict, Any, Optional, List
|
||
import httpx
|
||
import asyncio
|
||
import json
|
||
import logging
|
||
from datetime import datetime
|
||
|
||
from ..core.exceptions import (
|
||
AIAnalysisError,
|
||
APIError,
|
||
AuthenticationError,
|
||
RateLimitError
|
||
)
|
||
from ..schemas.report import AIAnalysisRequest, AIAnalysisResponse
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
|
||
class GeminiAnalyzer:
|
||
"""Gemini AI分析器"""
|
||
|
||
def __init__(self, api_key: str, config: Optional[Dict[str, Any]] = None):
|
||
if not api_key:
|
||
raise AuthenticationError("gemini", {"message": "Gemini API密钥未配置"})
|
||
|
||
self.api_key = api_key
|
||
self.config = config or {}
|
||
self.base_url = self.config.get("base_url", "https://generativelanguage.googleapis.com/v1beta")
|
||
self.model = self.config.get("model", "gemini-pro")
|
||
self.timeout = self.config.get("timeout", 60)
|
||
self.max_retries = self.config.get("max_retries", 3)
|
||
self.retry_delay = self.config.get("retry_delay", 2)
|
||
|
||
# 生成配置
|
||
self.generation_config = {
|
||
"temperature": self.config.get("temperature", 0.7),
|
||
"top_p": self.config.get("top_p", 0.8),
|
||
"top_k": self.config.get("top_k", 40),
|
||
"max_output_tokens": self.config.get("max_output_tokens", 8192),
|
||
}
|
||
|
||
async def analyze_business_info(self, symbol: str, market: str, financial_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""分析公司业务信息"""
|
||
if not symbol or not market:
|
||
raise AIAnalysisError("股票代码和市场参数不能为空", self.model)
|
||
|
||
prompt = self._build_business_info_prompt(symbol, market, financial_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
# 验证响应内容
|
||
if not result or len(result.strip()) < 100:
|
||
raise AIAnalysisError("AI返回的业务信息分析内容过短或为空", self.model)
|
||
|
||
# 解析响应
|
||
parsed_content = self._parse_business_info_response(result)
|
||
|
||
# 验证解析结果的质量
|
||
quality = parsed_content.get("content_quality", {})
|
||
if quality.get("quality_score", 0) < 0.3:
|
||
logger.warning(f"业务信息分析质量较低: {symbol} ({market}) - 质量分数: {quality.get('quality_score', 0)}")
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="business_info",
|
||
content=parsed_content,
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"业务信息分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_fundamental(self, symbol: str, market: str, financial_data: Dict[str, Any], business_info: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""基本面分析（景林模型）"""
|
||
prompt = self._build_fundamental_analysis_prompt(symbol, market, financial_data, business_info)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="fundamental_analysis",
|
||
content=self._parse_fundamental_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"基本面分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_bullish_case(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""看涨分析（隐藏资产、护城河分析）"""
|
||
prompt = self._build_bullish_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="bullish_analysis",
|
||
content=self._parse_bullish_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"看涨分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_bearish_case(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""看跌分析（价值底线、最坏情况分析）"""
|
||
prompt = self._build_bearish_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="bearish_analysis",
|
||
content=self._parse_bearish_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"看跌分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_market_sentiment(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""市场情绪分析"""
|
||
prompt = self._build_market_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="market_analysis",
|
||
content=self._parse_market_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"市场分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_news_catalysts(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""新闻催化剂分析"""
|
||
prompt = self._build_news_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="news_analysis",
|
||
content=self._parse_news_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"新闻分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_trading_dynamics(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""交易动态分析"""
|
||
prompt = self._build_trading_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="trading_analysis",
|
||
content=self._parse_trading_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"交易分析失败: {str(e)}", self.model)
|
||
|
||
async def analyze_insider_institutional(self, symbol: str, market: str, context_data: Dict[str, Any]) -> AIAnalysisResponse:
|
||
"""内部人与机构动向分析"""
|
||
prompt = self._build_insider_analysis_prompt(symbol, market, context_data)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="insider_analysis",
|
||
content=self._parse_insider_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"内部人分析失败: {str(e)}", self.model)
|
||
|
||
async def generate_final_conclusion(self, symbol: str, market: str, all_analyses: List[Dict[str, Any]]) -> AIAnalysisResponse:
|
||
"""生成最终结论"""
|
||
prompt = self._build_conclusion_prompt(symbol, market, all_analyses)
|
||
|
||
try:
|
||
result = await self._retry_request(self._call_gemini_api, prompt)
|
||
|
||
return AIAnalysisResponse(
|
||
symbol=symbol,
|
||
market=market,
|
||
analysis_type="final_conclusion",
|
||
content=self._parse_conclusion_response(result),
|
||
model_used=self.model,
|
||
generated_at=datetime.now()
|
||
)
|
||
except Exception as e:
|
||
if isinstance(e, (AIAnalysisError, APIError)):
|
||
raise
|
||
raise AIAnalysisError(f"最终结论生成失败: {str(e)}", self.model)
|
||
|
||
async def _call_gemini_api(self, prompt: str) -> str:
|
||
"""调用Gemini API"""
|
||
url = f"{self.base_url}/models/{self.model}:generateContent"
|
||
|
||
headers = {
|
||
"Content-Type": "application/json",
|
||
}
|
||
|
||
data = {
|
||
"contents": [
|
||
{
|
||
"parts": [
|
||
{
|
||
"text": prompt
|
||
}
|
||
]
|
||
}
|
||
],
|
||
"generationConfig": self.generation_config
|
||
}
|
||
|
||
params = {
|
||
"key": self.api_key
|
||
}
|
||
|
||
try:
|
||
async with httpx.AsyncClient(timeout=self.timeout) as client:
|
||
response = await client.post(url, json=data, headers=headers, params=params)
|
||
response.raise_for_status()
|
||
|
||
result = response.json()
|
||
|
||
if "error" in result:
|
||
error = result["error"]
|
||
error_code = error.get("code", 0)
|
||
error_message = error.get("message", "Unknown error")
|
||
|
||
if error_code == 401 or "API key" in error_message:
|
||
raise AuthenticationError("gemini", {"message": error_message})
|
||
elif error_code == 429 or "quota" in error_message.lower():
|
||
raise RateLimitError("gemini")
|
||
else:
|
||
raise APIError(f"Gemini API错误: {error_message}", error_code, "gemini")
|
||
|
||
candidates = result.get("candidates", [])
|
||
if not candidates:
|
||
raise AIAnalysisError("Gemini API返回空结果", self.model)
|
||
|
||
content = candidates[0].get("content", {})
|
||
parts = content.get("parts", [])
|
||
if not parts:
|
||
raise AIAnalysisError("Gemini API返回内容为空", self.model)
|
||
|
||
return parts[0].get("text", "")
|
||
|
||
except httpx.HTTPStatusError as e:
|
||
if e.response.status_code == 401:
|
||
raise AuthenticationError("gemini", {"status_code": e.response.status_code})
|
||
elif e.response.status_code == 429:
|
||
raise RateLimitError("gemini")
|
||
else:
|
||
raise APIError(f"HTTP错误: {e.response.status_code}", e.response.status_code, "gemini")
|
||
except httpx.RequestError as e:
|
||
raise AIAnalysisError(f"网络请求失败: {str(e)}", self.model)
|
||
|
||
async def _retry_request(self, func, *args, **kwargs):
|
||
"""重试机制"""
|
||
last_exception = None
|
||
|
||
for attempt in range(self.max_retries):
|
||
try:
|
||
return await func(*args, **kwargs)
|
||
except (httpx.TimeoutException, httpx.ConnectError, AIAnalysisError) as e:
|
||
last_exception = e
|
||
if attempt < self.max_retries - 1:
|
||
await asyncio.sleep(self.retry_delay * (2 ** attempt)) # 指数退避
|
||
continue
|
||
except (AuthenticationError, RateLimitError) as e:
|
||
# 认证错误和频率限制错误不重试
|
||
raise e
|
||
except Exception as e:
|
||
# 其他异常不重试
|
||
raise e
|
||
|
||
# 所有重试都失败了
|
||
raise AIAnalysisError(
|
||
f"Gemini API请求失败，已重试 {self.max_retries} 次: {str(last_exception)}",
|
||
self.model
|
||
)
|
||
|
||
def _build_business_info_prompt(self, symbol: str, market: str, financial_data: Dict[str, Any]) -> str:
|
||
"""构建业务信息分析提示词"""
|
||
|
||
# 根据市场调整分析重点
|
||
market_context = {
|
||
"中国": "中国A股市场",
|
||
"香港": "香港联交所",
|
||
"美国": "美国证券市场",
|
||
"日本": "日本证券市场"
|
||
}.get(market, market)
|
||
|
||
return f"""
|
||
请对股票代码 {symbol}（{market_context}）进行全面的业务信息分析。
|
||
|
||
基于以下财务数据：
|
||
{json.dumps(financial_data, ensure_ascii=False, indent=2)}
|
||
|
||
请按照以下结构提供详细分析，每个部分都要有具体的信息和数据支撑：
|
||
|
||
## 1. 公司概览
|
||
请提供公司的基本信息，包括：
|
||
- 公司全称、成立时间、注册地
|
||
- 主要业务领域和所属行业
|
||
- 在行业中的市场地位和排名
|
||
- 公司规模（员工数量、资产规模等）
|
||
- 主要竞争对手
|
||
|
||
## 2. 主营业务分析
|
||
请详细分析公司的核心业务：
|
||
- 主要产品和服务的详细描述
|
||
- 各业务板块的收入占比和盈利贡献
|
||
- 业务模式和价值链分析
|
||
- 核心竞争优势和差异化特点
|
||
- 产品或服务的市场需求和增长前景
|
||
|
||
## 3. 发展历程
|
||
请梳理公司的重要发展节点：
|
||
- 公司成立和上市历程
|
||
- 重要的业务转型和战略调整
|
||
- 关键的并购重组事件
|
||
- 重大技术突破或产品创新
|
||
- 国际化扩张历程（如适用）
|
||
|
||
## 4. 核心团队
|
||
请分析公司的管理层情况：
|
||
- 董事长、CEO等核心高管的背景和经验
|
||
- 管理团队的稳定性和变动情况
|
||
- 创始人或控股股东的影响力
|
||
- 管理层的激励机制和持股情况
|
||
- 公司治理结构的特点
|
||
|
||
## 5. 供应链分析
|
||
请分析公司的供应链体系：
|
||
- 主要原材料或服务的供应商情况
|
||
- 供应链的集中度和依赖性风险
|
||
- 供应链管理的优势和挑战
|
||
- 重要客户的构成和集中度
|
||
- 客户关系的稳定性和粘性
|
||
|
||
## 6. 销售模式
|
||
请分析公司的销售和营销策略：
|
||
- 主要销售渠道和分销网络
|
||
- 直销与经销的比例和策略
|
||
- 线上线下销售模式的结合
|
||
- 目标客户群体和市场定位
|
||
- 品牌建设和市场推广策略
|
||
|
||
## 7. 未来展望
|
||
请分析公司的发展前景：
|
||
- 公司的中长期发展战略和规划
|
||
- 新业务或新市场的拓展计划
|
||
- 技术创新和研发投入方向
|
||
- 面临的主要机遇和挑战
|
||
- 行业发展趋势对公司的影响
|
||
|
||
请确保分析内容：
|
||
1. 基于公开可获得的信息和财务数据
|
||
2. 客观、准确，避免过度乐观或悲观
|
||
3. 具体详实，有数据和事实支撑
|
||
4. 结构清晰，便于理解和阅读
|
||
5. 用中文回答，语言专业但易懂
|
||
|
||
注意：如果某些信息无法获得或不确定，请明确说明，不要编造信息。
|
||
"""
|
||
|
||
def _build_fundamental_analysis_prompt(self, symbol: str, market: str, financial_data: Dict[str, Any], business_info: Dict[str, Any]) -> str:
|
||
"""构建基本面分析提示词（景林模型）"""
|
||
return f"""
|
||
请使用景林投资的基本面分析框架，对股票代码 {symbol}（{market}市场）进行深度分析。
|
||
|
||
财务数据：
|
||
{json.dumps(financial_data, ensure_ascii=False, indent=2)}
|
||
|
||
业务信息：
|
||
{json.dumps(business_info, ensure_ascii=False, indent=2)}
|
||
|
||
请按照以下景林模型问题集进行分析：
|
||
|
||
1. 商业模式分析
|
||
- 这是一门什么样的生意？
|
||
- 商业模式的核心竞争力是什么？
|
||
- 盈利模式是否可持续？
|
||
|
||
2. 行业地位分析
|
||
- 公司在行业中的地位如何？
|
||
- 市场份额和竞争优势？
|
||
- 行业发展趋势对公司的影响？
|
||
|
||
3. 财务质量分析
|
||
- 收入增长的质量如何？
|
||
- 盈利能力和现金流状况？
|
||
- 资产负债结构是否健康？
|
||
|
||
4. 管理层评估
|
||
- 管理层的能力和诚信度？
|
||
- 公司治理结构是否完善？
|
||
- 股东利益是否得到保护？
|
||
|
||
5. 估值分析
|
||
- 当前估值水平是否合理？
|
||
- 与同行业公司比较如何？
|
||
- 未来增长预期是否支撑估值？
|
||
|
||
请用中文提供详细、专业的分析，每个方面都要有具体的数据支撑和逻辑推理。
|
||
"""
|
||
|
||
def _build_bullish_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建看涨分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行看涨分析，重点关注隐藏资产和护城河竞争优势。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行看涨分析：
|
||
|
||
1. 隐藏资产发现
|
||
- 账面价值被低估的资产
|
||
- 无形资产的真实价值
|
||
- 潜在的资产重估机会
|
||
|
||
2. 护城河分析
|
||
- 品牌价值和客户忠诚度
|
||
- 技术壁垒和专利保护
|
||
- 规模经济和网络效应
|
||
- 转换成本和客户粘性
|
||
|
||
3. 成长潜力
|
||
- 新业务和新市场机会
|
||
- 产品创新和技术升级
|
||
- 市场扩张的可能性
|
||
|
||
4. 催化剂识别
|
||
- 短期可能的积极因素
|
||
- 政策支持和行业利好
|
||
- 公司内部改革和优化
|
||
|
||
5. 最佳情况假设
|
||
- 如果一切顺利，公司价值可能达到什么水平？
|
||
- 关键假设和实现路径
|
||
|
||
请用中文提供乐观但理性的分析，要有具体的逻辑支撑。
|
||
"""
|
||
|
||
def _build_bearish_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建看跌分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行看跌分析，重点关注价值底线和最坏情况。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行看跌分析：
|
||
|
||
1. 价值底线分析
|
||
- 清算价值估算
|
||
- 资产的最低合理价值
|
||
- 下行风险的底线在哪里
|
||
|
||
2. 主要风险因素
|
||
- 行业周期性风险
|
||
- 竞争加剧的威胁
|
||
- 技术替代的可能性
|
||
- 监管政策变化风险
|
||
|
||
3. 财务脆弱性
|
||
- 债务压力和流动性风险
|
||
- 现金流恶化的可能性
|
||
- 盈利能力下降的风险
|
||
|
||
4. 管理层风险
|
||
- 决策失误的历史
|
||
- 治理结构的缺陷
|
||
- 利益冲突的可能性
|
||
|
||
5. 最坏情况假设
|
||
- 如果一切都出错，公司价值可能跌到什么水平？
|
||
- 关键风险因素和触发条件
|
||
|
||
请用中文提供谨慎但客观的分析，要有具体的风险量化。
|
||
"""
|
||
|
||
def _build_market_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建市场分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行市场情绪分析，重点关注分歧点与变化驱动。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行市场分析：
|
||
|
||
1. 市场情绪评估
|
||
- 当前市场对该股票的主流观点
|
||
- 机构投资者的持仓变化
|
||
- 散户投资者的情绪指标
|
||
|
||
2. 分歧点识别
|
||
- 市场存在哪些主要分歧？
|
||
- 乐观派和悲观派的核心观点
|
||
- 分歧的根本原因是什么？
|
||
|
||
3. 变化驱动因素
|
||
- 什么因素可能改变市场共识？
|
||
- 关键数据点和时间节点
|
||
- 外部环境变化的影响
|
||
|
||
4. 资金流向分析
|
||
- 主力资金的进出情况
|
||
- 不同类型投资者的行为模式
|
||
- 流动性状况评估
|
||
|
||
5. 市场预期vs现实
|
||
- 市场预期是否过于乐观或悲观？
|
||
- 预期差的投资机会在哪里？
|
||
|
||
请用中文提供专业的市场分析，要有数据支撑和逻辑推理。
|
||
"""
|
||
|
||
def _build_news_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建新闻分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行新闻催化剂分析，重点关注股价拐点预判。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行新闻分析：
|
||
|
||
1. 近期重要新闻梳理
|
||
- 公司公告和重大事件
|
||
- 行业政策和监管变化
|
||
- 宏观经济相关新闻
|
||
|
||
2. 催化剂识别
|
||
- 正面催化剂（业绩超预期、政策利好等）
|
||
- 负面催化剂（风险事件、竞争加剧等）
|
||
- 中性但重要的信息
|
||
|
||
3. 拐点预判
|
||
- 基本面拐点的可能时间
|
||
- 市场情绪拐点的信号
|
||
- 技术面拐点的确认
|
||
|
||
4. 新闻影响评估
|
||
- 短期影响vs长期影响
|
||
- 市场反应是否充分
|
||
- 后续发展的可能路径
|
||
|
||
5. 关注要点
|
||
- 未来需要重点关注的事件
|
||
- 可能的风险点和机会点
|
||
- 时间窗口和操作建议
|
||
|
||
请用中文提供前瞻性的分析，要有时间维度和影响程度的判断。
|
||
"""
|
||
|
||
def _build_trading_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建交易分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行交易分析，重点关注市场体量与增长路径。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行交易分析：
|
||
|
||
1. 市场体量分析
|
||
- 总市值和流通市值
|
||
- 日均成交量和换手率
|
||
- 市场容量和流动性评估
|
||
|
||
2. 增长路径分析
|
||
- 历史增长轨迹和驱动因素
|
||
- 未来增长的可能路径
|
||
- 增长的可持续性评估
|
||
|
||
3. 交易特征分析
|
||
- 股价波动特征和规律
|
||
- 主要交易时段和模式
|
||
- 大宗交易和异常交易情况
|
||
|
||
4. 技术面分析
|
||
- 关键技术位和支撑阻力
|
||
- 趋势线和形态分析
|
||
- 技术指标的信号
|
||
|
||
5. 交易策略建议
|
||
- 适合的交易时机和方式
|
||
- 风险控制和仓位管理
|
||
- 进出场点位的选择
|
||
|
||
请用中文提供实用的交易分析，要有具体的数据和操作建议。
|
||
"""
|
||
|
||
def _build_insider_analysis_prompt(self, symbol: str, market: str, context_data: Dict[str, Any]) -> str:
|
||
"""构建内部人分析提示词"""
|
||
return f"""
|
||
请对股票代码 {symbol}（{market}市场）进行内部人与机构动向分析。
|
||
|
||
基础数据：
|
||
{json.dumps(context_data, ensure_ascii=False, indent=2)}
|
||
|
||
请从以下角度进行分析：
|
||
|
||
1. 内部人交易分析
|
||
- 高管和大股东的买卖行为
|
||
- 内部人交易的时机和规模
|
||
- 内部人交易的信号意义
|
||
|
||
2. 机构持仓分析
|
||
- 主要机构投资者的持仓变化
|
||
- 新进和退出的机构情况
|
||
- 机构持仓集中度分析
|
||
|
||
3. 股东结构变化
|
||
- 股权结构的演变趋势
|
||
- 重要股东的进出情况
|
||
- 股权激励和员工持股情况
|
||
|
||
4. 资金流向追踪
|
||
- 大资金的进出时机
|
||
- 不同类型资金的偏好
|
||
- 资金成本和收益预期
|
||
|
||
5. 动向信号解读
|
||
- 内部人和机构行为的一致性
|
||
- 与股价走势的相关性
|
||
- 对未来走势的指示意义
|
||
|
||
请用中文提供专业的分析，要有数据支撑和逻辑推理。
|
||
"""
|
||
|
||
def _build_conclusion_prompt(self, symbol: str, market: str, all_analyses: List[Dict[str, Any]]) -> str:
|
||
"""构建最终结论提示词"""
|
||
analyses_text = "\n\n".join([
|
||
f"{analysis.get('analysis_type', '未知分析')}:\n{json.dumps(analysis.get('content', {}), ensure_ascii=False, indent=2)}"
|
||
for analysis in all_analyses
|
||
])
|
||
|
||
return f"""
|
||
基于以下所有分析结果，请对股票代码 {symbol}（{market}市场）给出最终投资结论。
|
||
|
||
所有分析结果：
|
||
{analyses_text}
|
||
|
||
请从以下角度进行综合分析：
|
||
|
||
1. 关键矛盾识别
|
||
- 当前最核心的投资矛盾是什么？
|
||
- 不同分析维度的结论是否一致？
|
||
- 主要的不确定性因素有哪些？
|
||
|
||
2. 预期差分析
|
||
- 市场预期与实际情况的差异
|
||
- 可能被忽视或误解的关键信息
|
||
- 预期差带来的投资机会
|
||
|
||
3. 拐点临近性判断
|
||
- 基本面拐点的时间和概率
|
||
- 市场情绪拐点的信号
|
||
- 催化剂的时效性分析
|
||
|
||
4. 风险收益评估
|
||
- 上行空间和下行风险的量化
|
||
- 风险调整后的收益预期
|
||
- 投资的风险收益比
|
||
|
||
5. 最终投资建议
|
||
- 明确的投资观点（看多/看空/中性）
|
||
- 建议的投资时间框架
|
||
- 关键的跟踪指标和退出条件
|
||
|
||
请用中文提供清晰、明确的投资结论，要有逻辑性和可操作性。
|
||
"""
|
||
|
||
def _parse_business_info_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析业务信息分析响应"""
|
||
return {
|
||
"company_overview": self._extract_section(response, "公司概览"),
|
||
"main_business": self._extract_section(response, "主营业务分析"),
|
||
"development_history": self._extract_section(response, "发展历程"),
|
||
"core_team": self._extract_section(response, "核心团队"),
|
||
"supply_chain": self._extract_section(response, "供应链分析"),
|
||
"sales_model": self._extract_section(response, "销售模式"),
|
||
"future_outlook": self._extract_section(response, "未来展望"),
|
||
"full_analysis": response,
|
||
"analysis_timestamp": datetime.now().isoformat(),
|
||
"content_quality": self._assess_content_quality(response)
|
||
}
|
||
|
||
def _parse_fundamental_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析基本面分析响应"""
|
||
return {
|
||
"business_model": self._extract_section(response, "商业模式"),
|
||
"industry_position": self._extract_section(response, "行业地位"),
|
||
"financial_quality": self._extract_section(response, "财务质量"),
|
||
"management_assessment": self._extract_section(response, "管理层"),
|
||
"valuation_analysis": self._extract_section(response, "估值分析"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_bullish_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析看涨分析响应"""
|
||
return {
|
||
"hidden_assets": self._extract_section(response, "隐藏资产"),
|
||
"moat_analysis": self._extract_section(response, "护城河"),
|
||
"growth_potential": self._extract_section(response, "成长潜力"),
|
||
"catalysts": self._extract_section(response, "催化剂"),
|
||
"best_case": self._extract_section(response, "最佳情况"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_bearish_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析看跌分析响应"""
|
||
return {
|
||
"value_floor": self._extract_section(response, "价值底线"),
|
||
"risk_factors": self._extract_section(response, "风险因素"),
|
||
"financial_vulnerability": self._extract_section(response, "财务脆弱性"),
|
||
"management_risks": self._extract_section(response, "管理层风险"),
|
||
"worst_case": self._extract_section(response, "最坏情况"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_market_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析市场分析响应"""
|
||
return {
|
||
"market_sentiment": self._extract_section(response, "市场情绪"),
|
||
"disagreement_points": self._extract_section(response, "分歧点"),
|
||
"change_drivers": self._extract_section(response, "变化驱动"),
|
||
"capital_flow": self._extract_section(response, "资金流向"),
|
||
"expectation_vs_reality": self._extract_section(response, "预期vs现实"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_news_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析新闻分析响应"""
|
||
return {
|
||
"recent_news": self._extract_section(response, "重要新闻"),
|
||
"catalysts": self._extract_section(response, "催化剂"),
|
||
"inflection_points": self._extract_section(response, "拐点预判"),
|
||
"news_impact": self._extract_section(response, "影响评估"),
|
||
"focus_points": self._extract_section(response, "关注要点"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_trading_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析交易分析响应"""
|
||
return {
|
||
"market_size": self._extract_section(response, "市场体量"),
|
||
"growth_path": self._extract_section(response, "增长路径"),
|
||
"trading_characteristics": self._extract_section(response, "交易特征"),
|
||
"technical_analysis": self._extract_section(response, "技术面"),
|
||
"trading_strategy": self._extract_section(response, "交易策略"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_insider_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析内部人分析响应"""
|
||
return {
|
||
"insider_trading": self._extract_section(response, "内部人交易"),
|
||
"institutional_holdings": self._extract_section(response, "机构持仓"),
|
||
"ownership_changes": self._extract_section(response, "股东结构"),
|
||
"capital_flow": self._extract_section(response, "资金流向"),
|
||
"signal_interpretation": self._extract_section(response, "信号解读"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _parse_conclusion_response(self, response: str) -> Dict[str, Any]:
|
||
"""解析最终结论响应"""
|
||
return {
|
||
"key_contradictions": self._extract_section(response, "关键矛盾"),
|
||
"expectation_gap": self._extract_section(response, "预期差"),
|
||
"inflection_timing": self._extract_section(response, "拐点临近性"),
|
||
"risk_return": self._extract_section(response, "风险收益"),
|
||
"investment_recommendation": self._extract_section(response, "投资建议"),
|
||
"full_analysis": response
|
||
}
|
||
|
||
def _extract_section(self, text: str, section_name: str) -> str:
|
||
"""从文本中提取特定章节内容"""
|
||
lines = text.split('\n')
|
||
section_content = []
|
||
in_section = False
|
||
|
||
for line in lines:
|
||
# 匹配章节标题（支持多种格式）
|
||
if section_name in line and any(marker in line for marker in ['##', '1.', '2.', '3.', '4.', '5.', '6.', '7.', '：', ':']):
|
||
in_section = True
|
||
# 不包含标题行，只要内容
|
||
continue
|
||
|
||
if in_section:
|
||
# 遇到下一个主要章节，停止
|
||
if line.strip() and any(marker in line for marker in ['##', '1.', '2.', '3.', '4.', '5.', '6.', '7.']) and section_name not in line:
|
||
break
|
||
section_content.append(line)
|
||
|
||
# 清理内容
|
||
content = '\n'.join(section_content).strip()
|
||
# 移除多余的空行
|
||
content = '\n'.join(line for line in content.split('\n') if line.strip() or not content.split('\n')[content.split('\n').index(line):content.split('\n').index(line)+2] == ['', ''])
|
||
return content
|
||
|
||
def _assess_content_quality(self, response: str) -> Dict[str, Any]:
|
||
"""评估内容质量"""
|
||
word_count = len(response)
|
||
sections_found = 0
|
||
|
||
# 检查各个章节是否存在
|
||
required_sections = ["公司概览", "主营业务", "发展历程", "核心团队", "供应链", "销售模式", "未来展望"]
|
||
for section in required_sections:
|
||
if section in response:
|
||
sections_found += 1
|
||
|
||
# 计算完整度
|
||
completeness = sections_found / len(required_sections)
|
||
|
||
# 评估详细程度
|
||
detail_level = "简要" if word_count < 500 else "详细" if word_count < 1500 else "非常详细"
|
||
|
||
return {
|
||
"word_count": word_count,
|
||
"sections_found": sections_found,
|
||
"total_sections": len(required_sections),
|
||
"completeness_ratio": completeness,
|
||
"detail_level": detail_level,
|
||
"quality_score": min(1.0, (completeness * 0.7 + min(1.0, word_count / 1000) * 0.3))
|
||
}
|
||
|
||
|
||
class AIAnalyzerFactory:
|
||
"""AI分析器工厂"""
|
||
|
||
@classmethod
|
||
def create_gemini_analyzer(cls, api_key: str, config: Optional[Dict[str, Any]] = None) -> GeminiAnalyzer:
|
||
"""创建Gemini分析器"""
|
||
return GeminiAnalyzer(api_key, config)
|
||
|
||
@classmethod
|
||
def create_analyzer(cls, analyzer_type: str, **kwargs) -> GeminiAnalyzer:
|
||
"""创建分析器（可扩展支持其他AI服务）"""
|
||
if analyzer_type.lower() == "gemini":
|
||
return cls.create_gemini_analyzer(kwargs.get("api_key"), kwargs.get("config"))
|
||
else:
|
||
raise AIAnalysisError(f"不支持的AI分析器类型: {analyzer_type}") |