442 lines
18 KiB
Python
442 lines
18 KiB
Python
import os
|
||
import sys
|
||
import pandas as pd
|
||
import numpy as np
|
||
from typing import Dict, Any, List, Optional, Tuple
|
||
import datetime
|
||
import time
|
||
import json
|
||
from datetime import datetime, timedelta
|
||
|
||
# 添加项目根目录到Python路径
|
||
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
||
|
||
from api.binance_api import BinanceAPI
|
||
from api.alltick_api import AllTickAPI
|
||
from api.deepseek_api import DeepSeekAPI
|
||
from models.data_processor import DataProcessor
|
||
from utils.config_loader import ConfigLoader
|
||
from utils.dingtalk_bot import DingTalkBot
|
||
from utils.discord_bot import DiscordBot
|
||
|
||
class GoldAgent:
|
||
"""黄金分析智能体，用于分析黄金市场数据并生成交易策略"""
|
||
|
||
def __init__(self, config_path: str = None):
|
||
"""
|
||
初始化黄金分析智能体
|
||
|
||
Args:
|
||
config_path: 配置文件路径，如果为None，则使用默认路径
|
||
"""
|
||
# 加载配置
|
||
self.config_loader = ConfigLoader(config_path)
|
||
|
||
# 获取各部分配置
|
||
self.binance_config = self.config_loader.get_binance_config()
|
||
self.deepseek_config = self.config_loader.get_deepseek_config()
|
||
self.alltick_config = self.config_loader.get_alltick_config()
|
||
self.gold_config = self.config_loader.get_gold_config() # 黄金特定配置
|
||
self.data_config = self.config_loader.get_data_config()
|
||
self.dingtalk_config = self.config_loader.get_dingtalk_config()
|
||
self.discord_config = self.config_loader.get_discord_config()
|
||
|
||
# 初始化API客户端
|
||
self.binance_api = BinanceAPI(
|
||
api_key=self.binance_config['api_key'],
|
||
api_secret=self.binance_config['api_secret'],
|
||
test_mode=self.binance_config['test_mode']
|
||
)
|
||
|
||
# 初始化AllTick API客户端（用于获取黄金数据）
|
||
self.alltick_api = AllTickAPI(
|
||
api_key=self.alltick_config['api_key']
|
||
)
|
||
|
||
self.deepseek_api = DeepSeekAPI()
|
||
|
||
# 初始化数据处理器
|
||
self.data_processor = DataProcessor(storage_path=os.path.join(self.data_config['storage_path'], 'gold'))
|
||
|
||
# 初始化钉钉机器人（如果启用）
|
||
self.dingtalk_bot = None
|
||
if self.dingtalk_config.get('enabled', False):
|
||
self.dingtalk_bot = DingTalkBot(
|
||
webhook_url=self.dingtalk_config['webhook_url'],
|
||
secret=self.dingtalk_config.get('secret')
|
||
)
|
||
print("钉钉机器人已启用")
|
||
|
||
# 初始化Discord机器人（如果启用）
|
||
self.discord_bot = None
|
||
if self.discord_config.get('enabled', False):
|
||
self.discord_bot = DiscordBot(
|
||
webhook_url=self.discord_config['gold_webhook_url']
|
||
)
|
||
print("Discord机器人已启用")
|
||
|
||
# 设置黄金交易对
|
||
self.gold_symbols = self.gold_config.get('symbols', ["XAUUSD"]) # 默认黄金/美元
|
||
|
||
# 设置时间间隔
|
||
self.time_interval = self.gold_config.get('time_interval', '1d') # 默认日线
|
||
|
||
|
||
def fetch_historical_data(self, symbol: str, days: int = 180) -> pd.DataFrame:
|
||
"""
|
||
获取历史数据
|
||
|
||
Args:
|
||
symbol: 黄金交易对符号，例如 'XAUUSD'
|
||
days: 要获取的天数
|
||
|
||
Returns:
|
||
历史数据DataFrame
|
||
"""
|
||
print(f"获取{symbol}的历史数据（{days}天）...")
|
||
|
||
# 计算开始时间
|
||
start_time = datetime.now() - timedelta(days=days)
|
||
start_str = start_time.strftime("%Y-%m-%d")
|
||
|
||
try:
|
||
# 使用AllTick API获取黄金K线数据
|
||
data = self.alltick_api.get_historical_klines(
|
||
symbol=symbol,
|
||
interval=self.time_interval,
|
||
start_str=start_str,
|
||
limit=days # 请求天数对应的K线数量
|
||
)
|
||
except Exception as e:
|
||
print(f"从AllTick获取数据出错: {e}")
|
||
|
||
if data.empty:
|
||
print(f"无法获取{symbol}的历史数据")
|
||
return pd.DataFrame()
|
||
|
||
# 保存原始数据
|
||
raw_data_path = self.data_processor.save_data(symbol, data, data_type='raw')
|
||
print(f"原始数据已保存到: {raw_data_path}")
|
||
|
||
return data
|
||
|
||
def process_data(self, symbol: str, data: pd.DataFrame) -> pd.DataFrame:
|
||
"""
|
||
处理数据
|
||
|
||
Args:
|
||
symbol: 黄金交易对符号，例如 'XAUUSD'
|
||
data: 原始数据
|
||
|
||
Returns:
|
||
处理后的数据
|
||
"""
|
||
print(f"处理{symbol}的数据...")
|
||
|
||
# 预处理数据
|
||
processed_data = self.data_processor.preprocess_market_data(symbol, data)
|
||
|
||
# 计算额外的黄金特定指标
|
||
processed_data = self._calculate_gold_indicators(processed_data)
|
||
|
||
# 保存处理后的数据
|
||
processed_data_path = self.data_processor.save_data(symbol, processed_data, data_type='processed')
|
||
print(f"处理后的数据已保存到: {processed_data_path}")
|
||
|
||
return processed_data
|
||
|
||
def _calculate_gold_indicators(self, data: pd.DataFrame) -> pd.DataFrame:
|
||
"""
|
||
计算黄金特定技术指标
|
||
|
||
Args:
|
||
data: 预处理后的数据
|
||
|
||
Returns:
|
||
添加了黄金特定指标的数据
|
||
"""
|
||
# 复制数据，避免修改原始数据
|
||
df = data.copy()
|
||
|
||
# 黄金波动率指标 (GVZ - Gold Volatility Index)
|
||
df['close_pct_change'] = df['close'].pct_change()
|
||
df['GVZ'] = df['close_pct_change'].rolling(window=20).std() * 100
|
||
|
||
# 黄金与美元指数相关性 (模拟数据，实际应用中需要获取美元指数数据)
|
||
df['USD_correlation'] = np.random.normal(-0.6, 0.2, len(df)) # 模拟负相关
|
||
|
||
# 季节性因素 (根据日期生成季节性指标)
|
||
df['month'] = pd.to_datetime(df.index).month
|
||
df['seasonal_factor'] = df['month'].apply(lambda m: 1 if m in [1, 8, 9, 12] else -1 if m in [3, 4, 6, 7] else 0)
|
||
|
||
# 黄金看涨/看跌指标 (计算基于多个指标的综合信号)
|
||
df['gold_bull_signal'] = np.where(
|
||
(df['RSI'] < 30) & (df['close'] > df['MA50']) & (df['seasonal_factor'] >= 0),
|
||
1, # 看涨信号
|
||
np.where(
|
||
(df['RSI'] > 70) & (df['close'] < df['MA50']) & (df['seasonal_factor'] <= 0),
|
||
-1, # 看跌信号
|
||
0 # 中性
|
||
)
|
||
)
|
||
|
||
return df
|
||
|
||
def _format_market_data(self, market_data: Dict[str, Any]) -> str:
|
||
"""
|
||
格式化市场数据为适合大模型的格式
|
||
|
||
Args:
|
||
market_data: 市场数据
|
||
|
||
Returns:
|
||
格式化的数据字符串
|
||
"""
|
||
# 这里可以根据实际情况调整格式化方式
|
||
return json.dumps(market_data, indent=2)
|
||
|
||
def _build_gold_analysis_prompt(self, formatted_data: str) -> str:
|
||
"""
|
||
构建黄金分析提示词
|
||
|
||
Args:
|
||
formatted_data: 格式化的市场数据
|
||
|
||
Returns:
|
||
提示词
|
||
"""
|
||
return f"""请对以下黄金市场K线数据进行深入分析，并给出交易建议。请使用中文回复。
|
||
|
||
数据:
|
||
{formatted_data}
|
||
|
||
请严格按照以下步骤分析:
|
||
1. 分析主要技术指标(包括RSI、MACD、布林带、均线等)，判断当前黄金市场趋势
|
||
2. 基于K线数据和斐波那契回调水平，确定关键支撑位和压力位
|
||
3. 分析黄金特定指标(GVZ波动率指标、与美元相关性、季节性因素)对价格的影响
|
||
4. 考虑全球宏观经济、地缘政治、通胀预期等因素对黄金价格的影响
|
||
5. 根据技术分析和基本面因素给出清晰的交易建议
|
||
6. 评估当前操作建议的紧迫性
|
||
|
||
请以JSON格式回复，仅包含以下字段:
|
||
- market_trend: 市场趋势 (牛市, 熊市, 震荡)
|
||
- technical_analysis: 技术指标详细分析 (重点分析RSI、MACD、布林带、均线交叉等情况)
|
||
- gold_specific_indicators: 黄金特有指标分析 (波动率、美元相关性、季节性)
|
||
- macro_factors: 宏观经济因素分析 (通胀、利率、地缘政治等)
|
||
- support_levels: 基于斐波那契回调的支撑位列表（标明各个支撑位对应的斐波那契水平）
|
||
- resistance_levels: 基于斐波那契回调的阻力位列表（同样标明对应水平）
|
||
- volume_analysis: 交易量分析，重点关注量价关系
|
||
- recommendation: 操作建议 (买入、卖出或等待)
|
||
- entry_points: 推荐入场点位
|
||
- exit_points: 推荐出场点位
|
||
- stop_loss: 建议止损位
|
||
- take_profit: 建议止盈位
|
||
- urgency_level: 操作紧迫性评级 (1-5，1为最低紧迫性，5为最高紧迫性)
|
||
- urgency_reason: 紧迫性评级的原因说明
|
||
- summary: 分析总结（不超过70字）
|
||
|
||
请确保回复为有效的JSON格式，分析要精准专业。重点关注黄金的避险属性、通胀对冲功能以及与主要货币的相关性。"""
|
||
|
||
def analyze_gold(self) -> Dict[str, Dict[str, Any]]:
|
||
"""
|
||
分析所有支持的黄金交易对
|
||
|
||
Returns:
|
||
所有交易对的分析结果
|
||
"""
|
||
results = {}
|
||
|
||
for symbol in self.gold_symbols:
|
||
print(f"\n开始分析{symbol}...")
|
||
|
||
# 获取并处理数据
|
||
raw_data = self.fetch_historical_data(symbol, days=self.gold_config.get('historical_days', 180))
|
||
|
||
if not raw_data.empty:
|
||
processed_data = self.process_data(symbol, raw_data)
|
||
|
||
# 准备市场数据
|
||
market_data = {
|
||
"symbol": symbol,
|
||
"current_price": float(processed_data['close'].iloc[-1]),
|
||
"price_change_24h": float(processed_data['close'].iloc[-1] - processed_data['close'].iloc[-2]),
|
||
"price_change_percentage_24h": float((processed_data['close'].iloc[-1] - processed_data['close'].iloc[-2]) / processed_data['close'].iloc[-2] * 100),
|
||
"historical_prices": processed_data['close'].tail(100).tolist(),
|
||
"volumes": processed_data['volume'].tail(100).tolist(),
|
||
"technical_indicators": {
|
||
"rsi": float(processed_data['RSI'].iloc[-1]),
|
||
"macd": float(processed_data['MACD'].iloc[-1]),
|
||
"macd_signal": float(processed_data['MACD_Signal'].iloc[-1]),
|
||
"bollinger_upper": float(processed_data['Bollinger_Upper'].iloc[-1]),
|
||
"bollinger_lower": float(processed_data['Bollinger_Lower'].iloc[-1]),
|
||
"ma5": float(processed_data['MA5'].iloc[-1]),
|
||
"ma10": float(processed_data['MA10'].iloc[-1]),
|
||
"ma20": float(processed_data['MA20'].iloc[-1]),
|
||
"ma50": float(processed_data['MA50'].iloc[-1]),
|
||
"ma200": float(processed_data['MA50'].iloc[-1]) if 'MA200' not in processed_data else float(processed_data['MA200'].iloc[-1]),
|
||
"atr": float(processed_data['ATR'].iloc[-1])
|
||
},
|
||
"gold_specific_indicators": {
|
||
"gvz": float(processed_data['GVZ'].iloc[-1]),
|
||
"usd_correlation": float(processed_data['USD_correlation'].iloc[-1]),
|
||
"seasonal_factor": float(processed_data['seasonal_factor'].iloc[-1]),
|
||
"gold_bull_signal": float(processed_data['gold_bull_signal'].iloc[-1])
|
||
},
|
||
"klines": processed_data[['open', 'high', 'low', 'close', 'volume']].tail(100).to_dict('records')
|
||
}
|
||
|
||
# 将市场数据格式化为适合大模型的格式
|
||
formatted_data = self._format_market_data(market_data)
|
||
|
||
# 构建提示词
|
||
prompt = self._build_gold_analysis_prompt(formatted_data)
|
||
|
||
# 调用API获取分析和交易建议
|
||
response, usage = self.deepseek_api.call_model(prompt, task_type="黄金分析", symbol=symbol)
|
||
|
||
# 解析响应
|
||
analysis_result = self.deepseek_api.extract_json_from_response(response)
|
||
|
||
# 添加token使用信息
|
||
if usage:
|
||
analysis_result["_token_usage"] = usage
|
||
|
||
# 整合结果
|
||
results[symbol] = {
|
||
"analysis": analysis_result,
|
||
"timestamp": datetime.now().isoformat()
|
||
}
|
||
print(f"{symbol}分析完成")
|
||
else:
|
||
print(f"跳过{symbol}分析，无法获取数据")
|
||
|
||
return results
|
||
|
||
def run_analysis_cycle(self) -> Dict[str, Any]:
|
||
"""
|
||
运行一个完整的分析周期
|
||
|
||
Returns:
|
||
分析结果
|
||
"""
|
||
print(f"开始新的黄金分析周期，时间：{datetime.now().isoformat()}")
|
||
|
||
results = self.analyze_gold()
|
||
|
||
# 保存分析结果
|
||
timestamp = datetime.now().strftime("%Y%m%d%H%M%S")
|
||
results_dir = os.path.join(self.data_config['storage_path'], 'gold', 'analysis_results')
|
||
os.makedirs(results_dir, exist_ok=True)
|
||
|
||
results_file = os.path.join(results_dir, f"gold_analysis_results_{timestamp}.json")
|
||
|
||
with open(results_file, 'w', encoding='utf-8') as f:
|
||
json.dump(results, f, indent=2, ensure_ascii=False)
|
||
|
||
print(f"分析结果已保存到：{results_file}")
|
||
|
||
# 导出 DeepSeek API token 使用情况
|
||
self._export_token_usage()
|
||
|
||
# 把分析的JSON结果调用大模型转化成交易建议
|
||
trading_suggestions = self.convert_analysis_to_trading_suggestions(results)
|
||
print(f"交易建议：{trading_suggestions}")
|
||
|
||
if self.dingtalk_bot:
|
||
self.dingtalk_bot.send_markdown(title="黄金交易建议", text=trading_suggestions)
|
||
|
||
if self.discord_bot:
|
||
self.discord_bot.send_message(content=trading_suggestions)
|
||
|
||
return results, trading_suggestions
|
||
|
||
def convert_analysis_to_trading_suggestions(self, results: Dict[str, Any]) -> str:
|
||
"""
|
||
把分析的JSON结果调用大模型转化成交易建议
|
||
"""
|
||
prompt = f"""
|
||
请对以下黄金市场分析的JSON结果进行归纳总结：
|
||
|
||
需要输出的内容包括：
|
||
标题：AI Agent 黄金分析报告
|
||
1. 对目标交易对行情进行总结
|
||
2. 对所有交易对给出操作建议：
|
||
2.1 操作建议（做多、做空、观望）
|
||
2.2 操作点位
|
||
2.3 止损止盈点位
|
||
2.4 操作评级
|
||
2.5 建议原因
|
||
以下是每个交易对的分析结果:
|
||
{results}
|
||
|
||
请以优美的Markdown格式输出，通过 emoji 标签来增加可读性。
|
||
"""
|
||
|
||
system_prompt = """
|
||
你是一个专业的黄金分析高手，你擅长分析市场趋势、预测价格走向和提供交易建议，请始终使用中文回复。
|
||
"""
|
||
response, usage = self.deepseek_api.call_model(prompt, system_prompt=system_prompt, task_type="交易建议")
|
||
|
||
message = self.deepseek_api.extract_text_from_response(response)
|
||
|
||
return message
|
||
|
||
def _export_token_usage(self) -> None:
|
||
"""
|
||
导出 DeepSeek API token 使用情况
|
||
"""
|
||
try:
|
||
# 每天导出一次详细的JSON数据
|
||
today = datetime.now().strftime("%Y%m%d")
|
||
token_usage_dir = os.path.join(self.data_config['storage_path'], 'gold', 'token_usage')
|
||
os.makedirs(token_usage_dir, exist_ok=True)
|
||
|
||
json_file = os.path.join(token_usage_dir, f"deepseek_token_usage_{today}.json")
|
||
|
||
# 如果文件不存在，导出一次
|
||
if not os.path.exists(json_file):
|
||
exported_file = self.deepseek_api.export_token_usage(json_file)
|
||
if exported_file:
|
||
print(f"DeepSeek API token 使用情况已导出到：{exported_file}")
|
||
|
||
# 每次都导出CSV格式的统计数据
|
||
csv_file = os.path.join(token_usage_dir, f"deepseek_token_usage_{today}.csv")
|
||
self.deepseek_api.export_token_usage(csv_file, "csv")
|
||
|
||
# 输出当前使用情况统计
|
||
stats = self.deepseek_api.get_token_usage_stats()
|
||
print(f"DeepSeek API token 使用统计：")
|
||
print(f"- 总调用次数: {stats['total_calls']}")
|
||
print(f"- 总token数: {stats['total_tokens']}")
|
||
print(f"- 输入token: {stats['total_prompt_tokens']}")
|
||
print(f"- 输出token: {stats['total_completion_tokens']}")
|
||
print(f"- 平均每次调用: {stats['average_tokens_per_call']:.2f} tokens")
|
||
|
||
except Exception as e:
|
||
print(f"导出 token 使用情况时出错: {e}")
|
||
|
||
def start_agent(self) -> None:
|
||
"""
|
||
启动智能体
|
||
|
||
Args:
|
||
run_once: 是否只运行一次分析周期
|
||
"""
|
||
print("启动黄金分析智能体...")
|
||
|
||
try:
|
||
self.run_analysis_cycle()
|
||
# 导出最终的token使用情况
|
||
self._export_token_usage()
|
||
|
||
except KeyboardInterrupt:
|
||
print("\n智能体已停止")
|
||
# 导出最终的token使用情况
|
||
self._export_token_usage()
|
||
|
||
except Exception as e:
|
||
print(f"智能体运行时出错: {e}")
|
||
import traceback
|
||
traceback.print_exc()
|
||
# 发生错误时也尝试导出token使用情况
|
||
self._export_token_usage() |