Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import yfinance as yf
|
3 |
+
import pandas as pd
|
4 |
+
import numpy as np
|
5 |
+
import plotly.graph_objects as go
|
6 |
+
from plotly.subplots import make_subplots
|
7 |
+
from datetime import datetime, timedelta
|
8 |
+
from langchain_huggingface import HuggingFaceEndpoint
|
9 |
+
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
|
10 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
11 |
+
import chromadb
|
12 |
+
import requests
|
13 |
+
from bs4 import BeautifulSoup
|
14 |
+
import warnings
|
15 |
+
from typing import Dict, List, Tuple
|
16 |
+
import feedparser
|
17 |
+
from sentence_transformers import SentenceTransformer
|
18 |
+
import faiss
|
19 |
+
import json
|
20 |
+
import os
|
21 |
+
|
22 |
+
warnings.filterwarnings('ignore')
|
23 |
+
|
24 |
+
|
25 |
+
|
26 |
+
COMPANIES = {
|
27 |
+
'Apple (AAPL)': 'AAPL',
|
28 |
+
'Microsoft (MSFT)': 'MSFT',
|
29 |
+
'Amazon (AMZN)': 'AMZN',
|
30 |
+
'Google (GOOGL)': 'GOOGL',
|
31 |
+
'Meta (META)': 'META',
|
32 |
+
'Tesla (TSLA)': 'TSLA',
|
33 |
+
'NVIDIA (NVDA)': 'NVDA',
|
34 |
+
'JPMorgan Chase (JPM)': 'JPM',
|
35 |
+
'Johnson & Johnson (JNJ)': 'JNJ',
|
36 |
+
'Walmart (WMT)': 'WMT',
|
37 |
+
'Visa (V)': 'V',
|
38 |
+
'Mastercard (MA)': 'MA',
|
39 |
+
'Procter & Gamble (PG)': 'PG',
|
40 |
+
'UnitedHealth (UNH)': 'UNH',
|
41 |
+
'Home Depot (HD)': 'HD',
|
42 |
+
'Bank of America (BAC)': 'BAC',
|
43 |
+
'Coca-Cola (KO)': 'KO',
|
44 |
+
'Pfizer (PFE)': 'PFE',
|
45 |
+
'Disney (DIS)': 'DIS',
|
46 |
+
'Netflix (NFLX)': 'NFLX'
|
47 |
+
}
|
48 |
+
|
49 |
+
# Initialize models
|
50 |
+
print("Initializing models...")
|
51 |
+
api_token = os.getenv(TOKEN)
|
52 |
+
llm = HuggingFaceEndpoint(
|
53 |
+
repo_id="mistralai/Mistral-7B-Instruct-v0.2",
|
54 |
+
huggingfacehub_api_token=api_token,
|
55 |
+
temperature=0.7,
|
56 |
+
max_new_tokens=1000
|
57 |
+
)
|
58 |
+
vader = SentimentIntensityAnalyzer()
|
59 |
+
finbert = pipeline("sentiment-analysis",
|
60 |
+
model="ProsusAI/finbert")
|
61 |
+
print("Models initialized successfully!")
|
62 |
+
class AgenticRAGFramework:
|
63 |
+
"""Main framework coordinating all agents"""
|
64 |
+
def __init__(self):
|
65 |
+
self.technical_agent = TechnicalAnalysisAgent()
|
66 |
+
self.sentiment_agent = SentimentAnalysisAgent()
|
67 |
+
self.llama_agent = LLMAgent()
|
68 |
+
self.knowledge_base = chromadb.Client()
|
69 |
+
|
70 |
+
def analyze(self, symbol: str, data: pd.DataFrame) -> Dict:
|
71 |
+
"""Perform comprehensive analysis"""
|
72 |
+
technical_analysis = self.technical_agent.analyze(data)
|
73 |
+
sentiment_analysis = self.sentiment_agent.analyze(symbol)
|
74 |
+
llm_analysis = self.llama_agent.generate_analysis(
|
75 |
+
technical_analysis,
|
76 |
+
sentiment_analysis
|
77 |
+
)
|
78 |
+
|
79 |
+
return {
|
80 |
+
'technical_analysis': technical_analysis,
|
81 |
+
'sentiment_analysis': sentiment_analysis,
|
82 |
+
'llm_analysis': llm_analysis
|
83 |
+
}
|
84 |
+
|
85 |
+
|
86 |
+
class NewsSource:
|
87 |
+
"""Base class for news sources"""
|
88 |
+
def get_news(self, company: str) -> List[Dict]:
|
89 |
+
raise NotImplementedError
|
90 |
+
|
91 |
+
class FinvizNews(NewsSource):
|
92 |
+
"""Fetch news from FinViz"""
|
93 |
+
def get_news(self, company: str) -> List[Dict]:
|
94 |
+
try:
|
95 |
+
ticker = company.split('(')[-1].replace(')', '')
|
96 |
+
url = f"https://finviz.com/quote.ashx?t={ticker}"
|
97 |
+
headers = {
|
98 |
+
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
|
99 |
+
}
|
100 |
+
|
101 |
+
response = requests.get(url, headers=headers)
|
102 |
+
soup = BeautifulSoup(response.text, 'html.parser')
|
103 |
+
news_table = soup.find('table', {'class': 'news-table'})
|
104 |
+
|
105 |
+
if not news_table:
|
106 |
+
return []
|
107 |
+
|
108 |
+
news_list = []
|
109 |
+
for row in news_table.find_all('tr')[:5]:
|
110 |
+
cols = row.find_all('td')
|
111 |
+
if len(cols) >= 2:
|
112 |
+
date = cols[0].text.strip()
|
113 |
+
title = cols[1].a.text.strip()
|
114 |
+
link = cols[1].a['href']
|
115 |
+
|
116 |
+
news_list.append({
|
117 |
+
'title': title,
|
118 |
+
'description': title,
|
119 |
+
'date': date,
|
120 |
+
'source': 'FinViz',
|
121 |
+
'url': link
|
122 |
+
})
|
123 |
+
|
124 |
+
return news_list
|
125 |
+
except Exception as e:
|
126 |
+
print(f"FinViz Error: {str(e)}")
|
127 |
+
return []
|
128 |
+
|
129 |
+
class MarketWatchNews(NewsSource):
|
130 |
+
"""Fetch news from MarketWatch"""
|
131 |
+
def get_news(self, company: str) -> List[Dict]:
|
132 |
+
try:
|
133 |
+
ticker = company.split('(')[-1].replace(')', '')
|
134 |
+
url = f"https://www.marketwatch.com/investing/stock/{ticker}"
|
135 |
+
headers = {
|
136 |
+
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
|
137 |
+
}
|
138 |
+
|
139 |
+
response = requests.get(url, headers=headers)
|
140 |
+
soup = BeautifulSoup(response.text, 'html.parser')
|
141 |
+
news_elements = soup.find_all('div', {'class': 'article__content'})
|
142 |
+
|
143 |
+
news_list = []
|
144 |
+
for element in news_elements[:5]:
|
145 |
+
title_elem = element.find('a', {'class': 'link'})
|
146 |
+
if title_elem:
|
147 |
+
title = title_elem.text.strip()
|
148 |
+
link = title_elem['href']
|
149 |
+
date_elem = element.find('span', {'class': 'article__timestamp'})
|
150 |
+
date = date_elem.text if date_elem else 'Recent'
|
151 |
+
|
152 |
+
news_list.append({
|
153 |
+
'title': title,
|
154 |
+
'description': title,
|
155 |
+
'date': date,
|
156 |
+
'source': 'MarketWatch',
|
157 |
+
'url': link
|
158 |
+
})
|
159 |
+
|
160 |
+
return news_list
|
161 |
+
except Exception as e:
|
162 |
+
print(f"MarketWatch Error: {str(e)}")
|
163 |
+
return []
|
164 |
+
|
165 |
+
class YahooRSSNews(NewsSource):
|
166 |
+
"""Fetch news from Yahoo Finance RSS feed"""
|
167 |
+
def get_news(self, company: str) -> List[Dict]:
|
168 |
+
try:
|
169 |
+
ticker = company.split('(')[-1].replace(')', '')
|
170 |
+
url = f"https://feeds.finance.yahoo.com/rss/2.0/headline?s={ticker}®ion=US&lang=en-US"
|
171 |
+
|
172 |
+
feed = feedparser.parse(url)
|
173 |
+
news_list = []
|
174 |
+
|
175 |
+
for entry in feed.entries[:5]:
|
176 |
+
news_list.append({
|
177 |
+
'title': entry.title,
|
178 |
+
'description': entry.description,
|
179 |
+
'date': entry.published,
|
180 |
+
'source': 'Yahoo Finance',
|
181 |
+
'url': entry.link
|
182 |
+
})
|
183 |
+
|
184 |
+
return news_list
|
185 |
+
except Exception as e:
|
186 |
+
print(f"Yahoo RSS Error: {str(e)}")
|
187 |
+
return []
|
188 |
+
|
189 |
+
class TechnicalAnalysisAgent:
|
190 |
+
"""Agent for technical analysis"""
|
191 |
+
def __init__(self):
|
192 |
+
self.required_periods = {
|
193 |
+
'sma': [20, 50, 200],
|
194 |
+
'rsi': 14,
|
195 |
+
'volatility': 20,
|
196 |
+
'macd': [12, 26, 9]
|
197 |
+
}
|
198 |
+
|
199 |
+
def analyze(self, data: pd.DataFrame) -> Dict:
|
200 |
+
df = data.copy()
|
201 |
+
close_col = ('Close', df.columns.get_level_values(1)[0])
|
202 |
+
|
203 |
+
# Calculate metrics
|
204 |
+
df['Returns'] = df[close_col].pct_change()
|
205 |
+
|
206 |
+
# SMAs
|
207 |
+
for period in self.required_periods['sma']:
|
208 |
+
df[f'SMA_{period}'] = df[close_col].rolling(window=period).mean()
|
209 |
+
|
210 |
+
# RSI
|
211 |
+
delta = df[close_col].diff()
|
212 |
+
gain = delta.where(delta > 0, 0).rolling(window=14).mean()
|
213 |
+
loss = -delta.where(delta < 0, 0).rolling(window=14).mean()
|
214 |
+
rs = gain / loss
|
215 |
+
df['RSI'] = 100 - (100 / (1 + rs))
|
216 |
+
|
217 |
+
# MACD
|
218 |
+
exp1 = df[close_col].ewm(span=12, adjust=False).mean()
|
219 |
+
exp2 = df[close_col].ewm(span=26, adjust=False).mean()
|
220 |
+
df['MACD'] = exp1 - exp2
|
221 |
+
df['Signal_Line'] = df['MACD'].ewm(span=9, adjust=False).mean()
|
222 |
+
|
223 |
+
# Bollinger Bands
|
224 |
+
df['BB_middle'] = df[close_col].rolling(window=20).mean()
|
225 |
+
rolling_std = df[close_col].rolling(window=20).std()
|
226 |
+
df['BB_upper'] = df['BB_middle'] + (2 * rolling_std)
|
227 |
+
df['BB_lower'] = df['BB_middle'] - (2 * rolling_std)
|
228 |
+
|
229 |
+
return {
|
230 |
+
'processed_data': df,
|
231 |
+
'current_signals': self._generate_signals(df, close_col)
|
232 |
+
}
|
233 |
+
|
234 |
+
def _generate_signals(self, df: pd.DataFrame, close_col) -> Dict:
|
235 |
+
if df.empty:
|
236 |
+
return {
|
237 |
+
'trend': 'Unknown',
|
238 |
+
'rsi_signal': 'Unknown',
|
239 |
+
'macd_signal': 'Unknown',
|
240 |
+
'bb_position': 'Unknown'
|
241 |
+
}
|
242 |
+
|
243 |
+
current = df.iloc[-1]
|
244 |
+
|
245 |
+
trend = 'Bullish' if float(current['SMA_20']) > float(current['SMA_50']) else 'Bearish'
|
246 |
+
|
247 |
+
rsi_value = float(current['RSI'])
|
248 |
+
if rsi_value > 70:
|
249 |
+
rsi_signal = 'Overbought'
|
250 |
+
elif rsi_value < 30:
|
251 |
+
rsi_signal = 'Oversold'
|
252 |
+
else:
|
253 |
+
rsi_signal = 'Neutral'
|
254 |
+
|
255 |
+
macd_signal = 'Buy' if float(current['MACD']) > float(current['Signal_Line']) else 'Sell'
|
256 |
+
|
257 |
+
close_value = float(current[close_col])
|
258 |
+
bb_upper = float(current['BB_upper'])
|
259 |
+
bb_lower = float(current['BB_lower'])
|
260 |
+
|
261 |
+
if close_value > bb_upper:
|
262 |
+
bb_position = 'Above Upper Band'
|
263 |
+
elif close_value < bb_lower:
|
264 |
+
bb_position = 'Below Lower Band'
|
265 |
+
else:
|
266 |
+
bb_position = 'Within Bands'
|
267 |
+
|
268 |
+
return {
|
269 |
+
'trend': trend,
|
270 |
+
'rsi_signal': rsi_signal,
|
271 |
+
'macd_signal': macd_signal,
|
272 |
+
'bb_position': bb_position
|
273 |
+
}
|
274 |
+
|
275 |
+
class SentimentAnalysisAgent:
|
276 |
+
"""Agent for sentiment analysis"""
|
277 |
+
def __init__(self):
|
278 |
+
self.news_sources = [
|
279 |
+
FinvizNews(),
|
280 |
+
MarketWatchNews(),
|
281 |
+
YahooRSSNews()
|
282 |
+
]
|
283 |
+
|
284 |
+
def analyze(self, symbol: str) -> Dict:
|
285 |
+
all_news = []
|
286 |
+
for source in self.news_sources:
|
287 |
+
news_items = source.get_news(symbol)
|
288 |
+
all_news.extend(news_items)
|
289 |
+
|
290 |
+
vader_scores = []
|
291 |
+
finbert_scores = []
|
292 |
+
|
293 |
+
for article in all_news:
|
294 |
+
vader_scores.append(vader.polarity_scores(article['title']))
|
295 |
+
finbert_scores.append(
|
296 |
+
finbert(article['title'][:512])[0]
|
297 |
+
)
|
298 |
+
|
299 |
+
return {
|
300 |
+
'articles': all_news,
|
301 |
+
'vader_scores': vader_scores,
|
302 |
+
'finbert_scores': finbert_scores,
|
303 |
+
'aggregated': self._aggregate_sentiment(vader_scores, finbert_scores)
|
304 |
+
}
|
305 |
+
|
306 |
+
def _aggregate_sentiment(self, vader_scores: List[Dict],
|
307 |
+
finbert_scores: List[Dict]) -> Dict:
|
308 |
+
if not vader_scores or not finbert_scores:
|
309 |
+
return {
|
310 |
+
'sentiment': 'Neutral',
|
311 |
+
'confidence': 0,
|
312 |
+
'vader_sentiment': 0,
|
313 |
+
'finbert_sentiment': 0
|
314 |
+
}
|
315 |
+
|
316 |
+
avg_vader = np.mean([score['compound'] for score in vader_scores])
|
317 |
+
avg_finbert = np.mean([
|
318 |
+
1 if score['label'] == 'positive' else -1
|
319 |
+
for score in finbert_scores
|
320 |
+
])
|
321 |
+
|
322 |
+
combined_score = (avg_vader + avg_finbert) / 2
|
323 |
+
|
324 |
+
return {
|
325 |
+
'sentiment': 'Bullish' if combined_score > 0.1 else 'Bearish' if combined_score < -0.1 else 'Neutral',
|
326 |
+
'confidence': abs(combined_score),
|
327 |
+
'vader_sentiment': avg_vader,
|
328 |
+
'finbert_sentiment': avg_finbert
|
329 |
+
}
|
330 |
+
|
331 |
+
class LLMAgent:
|
332 |
+
"""Agent for LLM-based analysis using HuggingFace API"""
|
333 |
+
def __init__(self):
|
334 |
+
self.llm = llm
|
335 |
+
|
336 |
+
def generate_analysis(self, technical_data: Dict, sentiment_data: Dict) -> str:
|
337 |
+
prompt = self._create_prompt(technical_data, sentiment_data)
|
338 |
+
|
339 |
+
response = self.llm.invoke(prompt)
|
340 |
+
return response
|
341 |
+
|
342 |
+
def _create_prompt(self, technical_data: Dict, sentiment_data: Dict) -> str:
|
343 |
+
return f"""Based on technical and sentiment indicators:
|
344 |
+
|
345 |
+
Technical Signals:
|
346 |
+
- Trend: {technical_data['current_signals']['trend']}
|
347 |
+
- RSI: {technical_data['current_signals']['rsi_signal']}
|
348 |
+
- MACD: {technical_data['current_signals']['macd_signal']}
|
349 |
+
- BB Position: {technical_data['current_signals']['bb_position']}
|
350 |
+
- Sentiment: {sentiment_data['aggregated']['sentiment']} (Confidence: {sentiment_data['aggregated']['confidence']:.2f})
|
351 |
+
|
352 |
+
Provide:
|
353 |
+
1. Current Trend Analysis
|
354 |
+
2. Key Risk Factors
|
355 |
+
3. Trading Recommendations
|
356 |
+
4. Price Targets
|
357 |
+
5. Near-term Outlook (1-2 weeks)
|
358 |
+
|
359 |
+
Note: return only required information and nothing unnecessary"""
|
360 |
+
|
361 |
+
# class ChatbotRouter:
|
362 |
+
# """Routes chatbot queries to appropriate data sources and generates responses"""
|
363 |
+
# def __init__(self):
|
364 |
+
# self.llm = llm
|
365 |
+
# self.encoder = SentenceTransformer('all-MiniLM-L6-v2')
|
366 |
+
# self.faiss_index = None
|
367 |
+
# self.company_data = {}
|
368 |
+
# self.news_sources = [
|
369 |
+
# FinvizNews(),
|
370 |
+
# MarketWatchNews(),
|
371 |
+
# YahooRSSNews()
|
372 |
+
# ]
|
373 |
+
# self.load_faiss_index()
|
374 |
+
|
375 |
+
# def route_and_respond(self, query: str, company: str) -> str:
|
376 |
+
# query_type = self._classify_query(query.lower())
|
377 |
+
# route_message = f"\n[Taking {query_type.upper()} route]\n\n"
|
378 |
+
|
379 |
+
# if query_type == "company_info":
|
380 |
+
# context = self._get_company_context(query, company)
|
381 |
+
# elif query_type == "news":
|
382 |
+
# context = self._get_news_context(company)
|
383 |
+
# elif query_type == "price":
|
384 |
+
# context = self._get_price_context(company)
|
385 |
+
# else:
|
386 |
+
# return route_message + "I'm not sure how to handle this query. Please ask about company information, news, or price data."
|
387 |
+
|
388 |
+
# prompt = self._create_prompt(query, context, query_type)
|
389 |
+
# response = self.llm.invoke(prompt)
|
390 |
+
|
391 |
+
# return route_message + response
|
392 |
+
|
393 |
+
class ChatbotRouter:
|
394 |
+
"""Routes chatbot queries to appropriate data sources and generates responses"""
|
395 |
+
def __init__(self):
|
396 |
+
self.llm = llm
|
397 |
+
self.encoder = SentenceTransformer('all-MiniLM-L6-v2')
|
398 |
+
self.faiss_index = None
|
399 |
+
self.company_data = {}
|
400 |
+
self.news_sources = [
|
401 |
+
FinvizNews(),
|
402 |
+
MarketWatchNews(),
|
403 |
+
YahooRSSNews()
|
404 |
+
]
|
405 |
+
self.load_faiss_index()
|
406 |
+
|
407 |
+
def load_faiss_index(self):
|
408 |
+
try:
|
409 |
+
self.faiss_index = faiss.read_index("company_profiles.index")
|
410 |
+
for file in os.listdir('company_data'):
|
411 |
+
with open(f'company_data/{file}', 'r') as f:
|
412 |
+
company_name = file.replace('.txt', '')
|
413 |
+
self.company_data[company_name] = json.load(f)
|
414 |
+
except Exception as e:
|
415 |
+
print(f"Error loading FAISS index: {e}")
|
416 |
+
|
417 |
+
def route_and_respond(self, query: str, company: str) -> str:
|
418 |
+
query_type = self._classify_query(query.lower())
|
419 |
+
route_message = f"\n[Taking {query_type.upper()} route]\n\n"
|
420 |
+
|
421 |
+
if query_type == "company_info":
|
422 |
+
context = self._get_company_context(query, company)
|
423 |
+
elif query_type == "news":
|
424 |
+
context = self._get_news_context(company)
|
425 |
+
elif query_type == "price":
|
426 |
+
context = self._get_price_context(company)
|
427 |
+
else:
|
428 |
+
return route_message + "I'm not sure how to handle this query. Please ask about company information, news, or price data."
|
429 |
+
|
430 |
+
prompt = self._create_prompt(query, context, query_type)
|
431 |
+
response = self.llm.invoke(prompt)
|
432 |
+
|
433 |
+
return route_message + response
|
434 |
+
|
435 |
+
def _classify_query(self, query: str) -> str:
|
436 |
+
"""Classify query type"""
|
437 |
+
if any(word in query for word in ["profile", "about", "information", "details", "what", "who", "describe"]):
|
438 |
+
return "company_info"
|
439 |
+
elif any(word in query for word in ["news", "latest", "recent", "announcement", "update"]):
|
440 |
+
return "news"
|
441 |
+
elif any(word in query for word in ["price", "stock", "value", "market", "trading", "cost"]):
|
442 |
+
return "price"
|
443 |
+
return "unknown"
|
444 |
+
|
445 |
+
def _get_company_context(self, query: str, company: str) -> str:
|
446 |
+
"""Get relevant company information using FAISS"""
|
447 |
+
try:
|
448 |
+
query_vector = self.encoder.encode([query])
|
449 |
+
D, I = self.faiss_index.search(query_vector, 1)
|
450 |
+
|
451 |
+
company_name = company.split(" (")[0]
|
452 |
+
company_info = self.company_data.get(company_name, {})
|
453 |
+
print(company_info)
|
454 |
+
return company_info
|
455 |
+
|
456 |
+
except Exception as e:
|
457 |
+
return f"Error retrieving company information: {str(e)}"
|
458 |
+
|
459 |
+
def _get_news_context(self, company: str) -> str:
|
460 |
+
"""Get news from multiple sources"""
|
461 |
+
all_news = []
|
462 |
+
|
463 |
+
for source in self.news_sources:
|
464 |
+
news_items = source.get_news(company)
|
465 |
+
all_news.extend(news_items)
|
466 |
+
|
467 |
+
seen_titles = set()
|
468 |
+
unique_news = []
|
469 |
+
for news in all_news:
|
470 |
+
if news['title'] not in seen_titles:
|
471 |
+
seen_titles.add(news['title'])
|
472 |
+
unique_news.append(news)
|
473 |
+
|
474 |
+
if not unique_news:
|
475 |
+
return "No recent news found."
|
476 |
+
|
477 |
+
news_context = "Recent news articles:\n\n"
|
478 |
+
for news in unique_news[:5]:
|
479 |
+
news_context += f"Source: {news['source']}\n"
|
480 |
+
news_context += f"Title: {news['title']}\n"
|
481 |
+
if news['description']:
|
482 |
+
news_context += f"Description: {news['description']}\n"
|
483 |
+
news_context += f"Date: {news['date']}\n\n"
|
484 |
+
|
485 |
+
return news_context
|
486 |
+
|
487 |
+
def _get_price_context(self, company: str) -> str:
|
488 |
+
"""Get current price information"""
|
489 |
+
try:
|
490 |
+
ticker = company.split('(')[-1].replace(')', '')
|
491 |
+
stock = yf.Ticker(ticker)
|
492 |
+
info = stock.info
|
493 |
+
|
494 |
+
return f"""Current Stock Information:
|
495 |
+
Price: ${info.get('currentPrice', 'N/A')}
|
496 |
+
Day Range: ${info.get('dayLow', 'N/A')} - ${info.get('dayHigh', 'N/A')}
|
497 |
+
52 Week Range: ${info.get('fiftyTwoWeekLow', 'N/A')} - ${info.get('fiftyTwoWeekHigh', 'N/A')}
|
498 |
+
Market Cap: ${info.get('marketCap', 'N/A'):,}
|
499 |
+
Volume: {info.get('volume', 'N/A'):,}
|
500 |
+
P/E Ratio: {info.get('trailingPE', 'N/A')}
|
501 |
+
Dividend Yield: {info.get('dividendYield', 'N/A')}%"""
|
502 |
+
|
503 |
+
except Exception as e:
|
504 |
+
return f"Error fetching price data: {str(e)}"
|
505 |
+
|
506 |
+
def _create_prompt(self, query: str, context: str, query_type: str) -> str:
|
507 |
+
"""Create prompt for LLM"""
|
508 |
+
if query_type == "news":
|
509 |
+
return f"""Based on the following news articles, please provide a summary addressing the query.
|
510 |
+
|
511 |
+
Context:
|
512 |
+
{context}
|
513 |
+
|
514 |
+
Query: {query}
|
515 |
+
|
516 |
+
Please analyze the news and provide:
|
517 |
+
1. Key points from the recent articles
|
518 |
+
2. Any significant developments or trends
|
519 |
+
3. Potential impact on the company
|
520 |
+
4. Overall sentiment (positive/negative/neutral)
|
521 |
+
|
522 |
+
Response should be clear, concise, and focused on the most relevant information."""
|
523 |
+
else:
|
524 |
+
return f"""Based on the following {query_type} context, please answer the question.
|
525 |
+
|
526 |
+
Context:
|
527 |
+
{context}
|
528 |
+
|
529 |
+
Question: {query}
|
530 |
+
|
531 |
+
Please provide a clear and concise answer based on the given context."""
|
532 |
+
|
533 |
+
def _generate_response(self, prompt: str) -> str:
|
534 |
+
"""Generate response using LLM"""
|
535 |
+
inputs = self.llm_agent.tokenizer(prompt, return_tensors="pt").to(self.llm_agent.model.device)
|
536 |
+
outputs = self.llm_agent.model.generate(
|
537 |
+
inputs["input_ids"],
|
538 |
+
max_new_tokens=200,
|
539 |
+
temperature=0.7,
|
540 |
+
num_return_sequences=1
|
541 |
+
)
|
542 |
+
# Decode and remove the prompt part from the output
|
543 |
+
response = self.llm_agent.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
544 |
+
response_only = response.replace(prompt, "").strip()
|
545 |
+
print(response)
|
546 |
+
return response_only
|
547 |
+
|
548 |
+
def analyze_stock(company: str, lookback_days: int = 180) -> Tuple[str, go.Figure, go.Figure]:
|
549 |
+
"""Main analysis function"""
|
550 |
+
try:
|
551 |
+
symbol = COMPANIES[company]
|
552 |
+
end_date = datetime.now()
|
553 |
+
start_date = end_date - timedelta(days=lookback_days)
|
554 |
+
|
555 |
+
data = yf.download(symbol, start=start_date, end=end_date)
|
556 |
+
if len(data) == 0:
|
557 |
+
return "No data available.", None, None
|
558 |
+
|
559 |
+
framework = AgenticRAGFramework()
|
560 |
+
analysis = framework.analyze(symbol, data)
|
561 |
+
|
562 |
+
plots = create_plots(analysis)
|
563 |
+
|
564 |
+
return analysis['llm_analysis'], plots[0], plots[1]
|
565 |
+
|
566 |
+
except Exception as e:
|
567 |
+
return f"Error analyzing stock: {str(e)}", None, None
|
568 |
+
|
569 |
+
def create_plots(analysis: Dict) -> List[go.Figure]:
|
570 |
+
"""Create analysis plots"""
|
571 |
+
data = analysis['technical_analysis']['processed_data']
|
572 |
+
|
573 |
+
# Price and Volume Plot
|
574 |
+
fig1 = make_subplots(
|
575 |
+
rows=2, cols=1,
|
576 |
+
shared_xaxes=True,
|
577 |
+
vertical_spacing=0.03,
|
578 |
+
subplot_titles=('Price Analysis', 'Volume'),
|
579 |
+
row_heights=[0.7, 0.3]
|
580 |
+
)
|
581 |
+
|
582 |
+
close_col = ('Close', data.columns.get_level_values(1)[0])
|
583 |
+
open_col = ('Open', data.columns.get_level_values(1)[0])
|
584 |
+
volume_col = ('Volume', data.columns.get_level_values(1)[0])
|
585 |
+
|
586 |
+
fig1.add_trace(
|
587 |
+
go.Scatter(x=data.index, y=data[close_col], name='Price',
|
588 |
+
line=dict(color='blue', width=2)),
|
589 |
+
row=1, col=1
|
590 |
+
)
|
591 |
+
fig1.add_trace(
|
592 |
+
go.Scatter(x=data.index, y=data['SMA_20'], name='SMA20',
|
593 |
+
line=dict(color='orange', width=1.5)),
|
594 |
+
row=1, col=1
|
595 |
+
)
|
596 |
+
fig1.add_trace(
|
597 |
+
go.Scatter(x=data.index, y=data['SMA_50'], name='SMA50',
|
598 |
+
line=dict(color='red', width=1.5)),
|
599 |
+
row=1, col=1
|
600 |
+
)
|
601 |
+
|
602 |
+
colors = ['red' if float(row[close_col]) < float(row[open_col]) else 'green'
|
603 |
+
for idx, row in data.iterrows()]
|
604 |
+
|
605 |
+
fig1.add_trace(
|
606 |
+
go.Bar(x=data.index, y=data[volume_col], marker_color=colors, name='Volume'),
|
607 |
+
row=2, col=1
|
608 |
+
)
|
609 |
+
|
610 |
+
fig1.update_layout(
|
611 |
+
height=400,
|
612 |
+
showlegend=True,
|
613 |
+
xaxis_rangeslider_visible=False,
|
614 |
+
plot_bgcolor='white',
|
615 |
+
paper_bgcolor='white'
|
616 |
+
)
|
617 |
+
|
618 |
+
# Technical Indicators Plot
|
619 |
+
fig2 = make_subplots(
|
620 |
+
rows=3, cols=1,
|
621 |
+
shared_xaxes=True,
|
622 |
+
subplot_titles=('RSI', 'MACD', 'Bollinger Bands'),
|
623 |
+
row_heights=[0.33, 0.33, 0.34],
|
624 |
+
vertical_spacing=0.03
|
625 |
+
)
|
626 |
+
|
627 |
+
# RSI
|
628 |
+
fig2.add_trace(
|
629 |
+
go.Scatter(x=data.index, y=data['RSI'], name='RSI',
|
630 |
+
line=dict(color='purple', width=1.5)),
|
631 |
+
row=1, col=1
|
632 |
+
)
|
633 |
+
fig2.add_hline(y=70, line_dash="dash", line_color="red", row=1, col=1)
|
634 |
+
fig2.add_hline(y=30, line_dash="dash", line_color="green", row=1, col=1)
|
635 |
+
|
636 |
+
# MACD
|
637 |
+
fig2.add_trace(
|
638 |
+
go.Scatter(x=data.index, y=data['MACD'], name='MACD',
|
639 |
+
line=dict(color='blue', width=1.5)),
|
640 |
+
row=2, col=1
|
641 |
+
)
|
642 |
+
fig2.add_trace(
|
643 |
+
go.Scatter(x=data.index, y=data['Signal_Line'], name='Signal',
|
644 |
+
line=dict(color='orange', width=1.5)),
|
645 |
+
row=2, col=1
|
646 |
+
)
|
647 |
+
|
648 |
+
# Bollinger Bands
|
649 |
+
fig2.add_trace(
|
650 |
+
go.Scatter(x=data.index, y=data[close_col], name='Price',
|
651 |
+
line=dict(color='blue', width=2)),
|
652 |
+
row=3, col=1
|
653 |
+
)
|
654 |
+
fig2.add_trace(
|
655 |
+
go.Scatter(x=data.index, y=data['BB_upper'], name='Upper BB',
|
656 |
+
line=dict(color='gray', dash='dash')),
|
657 |
+
row=3, col=1
|
658 |
+
)
|
659 |
+
fig2.add_trace(
|
660 |
+
go.Scatter(x=data.index, y=data['BB_lower'], name='Lower BB',
|
661 |
+
line=dict(color='gray', dash='dash')),
|
662 |
+
row=3, col=1
|
663 |
+
)
|
664 |
+
|
665 |
+
fig2.update_layout(
|
666 |
+
height=400,
|
667 |
+
showlegend=True,
|
668 |
+
plot_bgcolor='white',
|
669 |
+
paper_bgcolor='white'
|
670 |
+
)
|
671 |
+
|
672 |
+
return [fig1, fig2]
|
673 |
+
|
674 |
+
def chatbot_response(message: str, company: str, history: List[Tuple[str, str]]) -> List[Tuple[str, str]]:
|
675 |
+
"""Handle chatbot interactions"""
|
676 |
+
router = ChatbotRouter(LlamaAgent())
|
677 |
+
response = router.route_and_respond(message, company)
|
678 |
+
history = history + [(message, response)]
|
679 |
+
return history
|
680 |
+
|
681 |
+
# def create_interface():
|
682 |
+
# """Create Gradio interface"""
|
683 |
+
# with gr.Blocks() as interface:
|
684 |
+
# gr.Markdown("# Stock Analysis with Multi-Source News")
|
685 |
+
|
686 |
+
# with gr.Row():
|
687 |
+
# with gr.Column(scale=2):
|
688 |
+
# company = gr.Dropdown(
|
689 |
+
# choices=list(COMPANIES.keys()),
|
690 |
+
# value=list(COMPANIES.keys())[0],
|
691 |
+
# label="Company"
|
692 |
+
# )
|
693 |
+
# lookback = gr.Slider(
|
694 |
+
# minimum=30,
|
695 |
+
# maximum=365,
|
696 |
+
# value=180,
|
697 |
+
# step=1,
|
698 |
+
# label="Analysis Period (days)"
|
699 |
+
# )
|
700 |
+
# analyze_btn = gr.Button("Analyze", variant="primary")
|
701 |
+
|
702 |
+
# with gr.Row():
|
703 |
+
# with gr.Column(scale=1):
|
704 |
+
# chatbot = gr.Chatbot(label="Stock Assistant", height=400)
|
705 |
+
# with gr.Row():
|
706 |
+
# msg = gr.Textbox(
|
707 |
+
# label="Ask about company info, news, or prices",
|
708 |
+
# scale=4
|
709 |
+
# )
|
710 |
+
# submit = gr.Button("Submit", scale=1)
|
711 |
+
# clear = gr.Button("Clear", scale=1)
|
712 |
+
|
713 |
+
# with gr.Column(scale=2):
|
714 |
+
# analysis = gr.Textbox(
|
715 |
+
# label="Technical Analysis Summary",
|
716 |
+
# lines=10
|
717 |
+
# )
|
718 |
+
# chart1 = gr.Plot(label="Price and Volume Analysis")
|
719 |
+
# chart2 = gr.Plot(label="Technical Indicators")
|
720 |
+
|
721 |
+
# # Event handlers
|
722 |
+
# analyze_btn.click(
|
723 |
+
# fn=analyze_stock,
|
724 |
+
# inputs=[company, lookback],
|
725 |
+
# outputs=[analysis, chart1, chart2]
|
726 |
+
# )
|
727 |
+
|
728 |
+
# submit.click(
|
729 |
+
# fn=chatbot_response,
|
730 |
+
# inputs=[msg, company, chatbot],
|
731 |
+
# outputs=chatbot
|
732 |
+
# )
|
733 |
+
|
734 |
+
# msg.submit(
|
735 |
+
# fn=chatbot_response,
|
736 |
+
# inputs=[msg, company, chatbot],
|
737 |
+
# outputs=chatbot
|
738 |
+
# )
|
739 |
+
|
740 |
+
# clear.click(lambda: None, None, chatbot, queue=False)
|
741 |
+
|
742 |
+
# return interface
|
743 |
+
|
744 |
+
def create_interface():
|
745 |
+
"""Create Gradio interface"""
|
746 |
+
with gr.Blocks() as interface:
|
747 |
+
gr.Markdown("# Stock Analysis with Multi-Source News")
|
748 |
+
|
749 |
+
# Top section with analysis components
|
750 |
+
with gr.Row():
|
751 |
+
# Left column - Controls and Summary
|
752 |
+
with gr.Column(scale=1):
|
753 |
+
company = gr.Dropdown(
|
754 |
+
choices=list(COMPANIES.keys()),
|
755 |
+
value=list(COMPANIES.keys())[0],
|
756 |
+
label="Company"
|
757 |
+
)
|
758 |
+
lookback = gr.Slider(
|
759 |
+
minimum=30,
|
760 |
+
maximum=365,
|
761 |
+
value=180,
|
762 |
+
step=1,
|
763 |
+
label="Analysis Period (days)"
|
764 |
+
)
|
765 |
+
analyze_btn = gr.Button("Analyze", variant="primary")
|
766 |
+
analysis = gr.Textbox(
|
767 |
+
label="Technical Analysis Summary",
|
768 |
+
lines=30
|
769 |
+
)
|
770 |
+
|
771 |
+
# Right column - Charts
|
772 |
+
with gr.Column(scale=2):
|
773 |
+
chart1 = gr.Plot(label="Price and Volume Analysis")
|
774 |
+
chart2 = gr.Plot(label="Technical Indicators")
|
775 |
+
|
776 |
+
gr.Markdown("---") # Separator
|
777 |
+
|
778 |
+
# Bottom section - Chatbot
|
779 |
+
with gr.Row():
|
780 |
+
chatbot = gr.Chatbot(label="Stock Assistant", height=400)
|
781 |
+
|
782 |
+
with gr.Row():
|
783 |
+
msg = gr.Textbox(
|
784 |
+
label="Ask about company info, news, or prices",
|
785 |
+
scale=4
|
786 |
+
)
|
787 |
+
submit = gr.Button("Submit", scale=1)
|
788 |
+
clear = gr.Button("Clear", scale=1)
|
789 |
+
|
790 |
+
# Event handlers
|
791 |
+
analyze_btn.click(
|
792 |
+
fn=analyze_stock,
|
793 |
+
inputs=[company, lookback],
|
794 |
+
outputs=[analysis, chart1, chart2]
|
795 |
+
)
|
796 |
+
|
797 |
+
submit.click(
|
798 |
+
fn=chatbot_response,
|
799 |
+
inputs=[msg, company, chatbot],
|
800 |
+
outputs=chatbot
|
801 |
+
)
|
802 |
+
|
803 |
+
msg.submit(
|
804 |
+
fn=chatbot_response,
|
805 |
+
inputs=[msg, company, chatbot],
|
806 |
+
outputs=chatbot
|
807 |
+
)
|
808 |
+
|
809 |
+
clear.click(lambda: None, None, chatbot, queue=False)
|
810 |
+
|
811 |
+
return interface
|
812 |
+
|
813 |
+
if __name__ == "__main__":
|
814 |
+
interface = create_interface()
|
815 |
+
interface.launch(debug=True)
|