Update app.py

app.py

@@ -1,18 +1,17 @@
-import gymnasium as gym
+import gym
 import numpy as np
+import matplotlib.pyplot as plt
 import requests
 import pandas as pd
 from datetime import datetime, timedelta
 from stable_baselines3 import PPO
 from stable_baselines3.common.vec_env import DummyVecEnv
-from gymnasium import spaces
+from gym import spaces
+import time
 import firebase_admin
 from firebase_admin import credentials, db
 import os
-import threading
-import time
 
-# Firebase initialization
 cred = credentials.Certificate("credentials.json")
 firebase_admin.initialize_app(cred, {"databaseURL": "https://socail-swap-default-rtdb.asia-southeast1.firebasedatabase.app/"})
 ref = db.reference()
@@ -21,8 +20,8 @@ buy_signals = []
 sell_signals = []
 
 class TradingEnv(gym.Env):
-    def __init__(self, data, window_size=50):
-        super(TradingEnv, self).__init__()
+    def _init_(self, data, window_size=50):
+        super(TradingEnv, self)._init_()
         self.data = data
         self.window_size = window_size
         self.current_step = window_size
@@ -30,11 +29,10 @@ class TradingEnv(gym.Env):
         self.observation_space = spaces.Box(
             low=0, high=1, shape=(window_size, 2), dtype=np.float32)
         
-    def reset(self, seed=None, options=None):
-        super().reset(seed=seed)
+    def reset(self):
         self.current_step = self.window_size
-        return self._get_observation(), {}
-
+        return self._get_observation()
+    
     def _get_observation(self):
         window_data = self.data[self.current_step-self.window_size:self.current_step]
         obs = window_data[['close', 'EMA']].values
@@ -54,9 +52,9 @@ class TradingEnv(gym.Env):
             elif action == 2: 
                 reward = self.data['close'].iloc[self.current_step - 1] - self.data['close'].iloc[self.current_step]
 
-        return self._get_observation(), reward, done, {}, {}
-
-def fetch_data(symbol='ETH', tsym='USD', start_date='2021-01-01', api_key='YOUR_API_KEY'):
+        return self._get_observation(), reward, done, {}
+    
+def fetch_data(symbol='ETH', tsym='USD', start_date='2021-01-01', api_key='66bc686cb714fadda1fad0320704c98869d4b31ce7d9d27560c6c574b4d04c54'):
     start_date = datetime.strptime(start_date, '%Y-%m-%d')
     end_date = datetime.utcnow()
     to_ts = int(end_date.timestamp())
@@ -70,7 +68,10 @@ def fetch_data(symbol='ETH', tsym='USD', start_date='2021-01-01', api_key='YOUR_
         df = pd.DataFrame(data_points)
         df['time'] = pd.to_datetime(df['time'], unit='s')
         df.set_index('time', inplace=True)
+        
+        # Filter data based on start_date
         df = df[df.index >= start_date]
+        
         return df[['close']]
     else:
         print(f"Error fetching data: {data['Message']}")
@@ -80,65 +81,68 @@ def calculate_ema(data, span=20):
     data['EMA'] = data['close'].ewm(span=span, adjust=False).mean()
     return data
 
-def load_or_train_model(env):
-    custom_objects = {"clip_range": 0.2, "lr_schedule": 0.0003}  # Update with your actual values or leave as is
-    try:
-        model = PPO.load("ppo_trading_agent", env=env, custom_objects=custom_objects)
-    except Exception as e:
-        print(f"Loading model failed: {e}. Training a new model.")
-        model = PPO('MlpPolicy', env, verbose=1)
-        model.learn(total_timesteps=10000)
-        model.save("ppo_trading_agent")
-    return model
-
-def run_model(model, env, data):
-    obs, _ = env.reset()
-    dates = data.index[50:]
-    prices = data['close'][50:]
-    emas = data['EMA'][50:]
-    actions = []
-
-    for date, price, ema in zip(dates, prices, emas):
-        action, _ = model.predict(obs)
-        actions.append(action[0])
-        obs, _, done, _, _ = env.step(action)
-        if done:
-            break
-
-    new_buy_signals = [(date, price, ema) for date, price, ema, action in zip(dates, prices, emas, actions) if action == 1]
-    new_sell_signals = [(date, price, ema) for date, price, ema, action in zip(dates, prices, emas, actions) if action == 2]
-
-    for signal in new_buy_signals:
-        if signal[0] not in [s[0] for s in buy_signals] and signal[0] not in [s[0] for s in sell_signals]:
-            buy_signals.append(signal)
-    
-    for signal in new_sell_signals:
-        if signal[0] not in [s[0] for s in sell_signals] and signal[0] not in [s[0] for s in buy_signals]:
-            sell_signals.append(signal)
-    
-    buy_signals_data = [{'timestamp': signal[0].strftime('%Y-%m-%d %H:%M:%S'), 'type': 'b', 'price': round(signal[1], 2), 'ema': round(signal[2], 2)} for signal in buy_signals]
-    sell_signals_data = [{'timestamp': signal[0].strftime('%Y-%m-%d %H:%M:%S'), 'type': 's', 'price': round(signal[1], 2), 'ema': round(signal[2], 2)} for signal in sell_signals]
-
-    all_signals_data = buy_signals_data + sell_signals_data
-
-    ref.child('signals').child('data').set(all_signals_data)
-
-def background_task():
+def run_model():
     while True:
         try:
             new_data = fetch_data(start_date=(datetime.utcnow() - timedelta(days=3)).strftime('%Y-%m-%d'))
-            if new_data is not None:
-                new_data = calculate_ema(new_data)
-                if len(new_data) >= 50:
-                    env = DummyVecEnv([lambda: TradingEnv(new_data)])
-                    model = load_or_train_model(env)
-                    run_model(model, env, new_data)
-            time.sleep(3600)
+            new_data = calculate_ema(new_data)
+            
+            if len(new_data) < 50:
+                print("Not enough data to update the environment.")
+                time.sleep(3600)
+                continue
+            
+            env = DummyVecEnv([lambda: TradingEnv(new_data)])
+            model.set_env(env)
+            
+            obs = env.reset()
+            dates = new_data.index[50:]
+            prices = new_data['close'][50:]
+            emas = new_data['EMA'][50:]
+            actions = []
+
+            for date, price, ema in zip(dates, prices, emas):
+                action, _ = model.predict(obs)
+                actions.append(action[0])
+                obs, _, done, _ = env.step(action)
+                if done:
+                    break
+
+            new_buy_signals = [(date, price, ema) for date, price, ema, action in zip(dates, prices, emas, actions) if action == 1 and date not in [signal[0] for signal in buy_signals]]
+            new_sell_signals = [(date, price, ema) for date, price, ema, action in zip(dates, prices, emas, actions) if action == 2 and date not in [signal[0] for signal in sell_signals]]
+
+            for signal in new_buy_signals:
+                if signal[0] not in [s[0] for s in buy_signals] and signal[0] not in [s[0] for s in sell_signals]:
+                    buy_signals.append(signal)
+                
+
+            for signal in new_sell_signals:
+                if signal[0] not in [s[0] for s in sell_signals] and signal[0] not in [s[0] for s in buy_signals]:
+                    sell_signals.append(signal)
+                    
+            buy_signals_data = [{'timestamp': signal[0].strftime('%Y-%m-%d %H:%M:%S'), 'type': 'b', 'price': round(signal[1], 2), 'ema': round(signal[2],2)} for signal in buy_signals]
+            sell_signals_data = [{'timestamp': signal[0].strftime('%Y-%m-%d %H:%M:%S'), 'type': 's', 'price': round(signal[1], 2), 'ema': round(signal[2],2)} for signal in sell_signals]
+
+            all_signals_data = buy_signals_data + sell_signals_data
+
+            ref.child('signals').child('data').set(all_signals_data)
+            time.sleep(3600) 
         except Exception as e:
             print(f"An error occurred: {e}")
             break
+        
+if _name_ == "_main_":
+    data = fetch_data()
+    data = calculate_ema(data)
+    if len(data) < 50:
+        raise ValueError("Not enough data to fill the window size.")
+
+    env = DummyVecEnv([lambda: TradingEnv(data)])
 
-if __name__ == "__main__":
-    background_thread = threading.Thread(target=background_task)
-    background_thread.start()
-    background_thread.join()
+    if os.path.exists("./ppo_trading_agent.zip"):
+        model = PPO.load("ppo_trading_agent", env=env)
+    else:
+        model = PPO('MlpPolicy', env, verbose=1)
+        model.learn(total_timesteps=10000)
+    print("Ender function")
+    run_model()