import requests from bs4 import BeautifulSoup import pandas as pd from io import StringIO from datetime import datetime, timedelta import gradio as gr from langchain.tools import tool from langchain.agents import initialize_agent, Tool, AgentExecutor from langchain_openai import ChatOpenAI from langchain.callbacks.base import BaseCallbackHandler # === Configuration === OPENROUTER_API_KEY = "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" HEADERS = { 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36', 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8', 'Accept-Language': 'en-US,en;q=0.9', 'Referer': 'https://www.google.com/', 'Connection': 'keep-alive', 'Cache-Control': 'max-age=0', 'Upgrade-Insecure-Requests': '1' } # === Tools === @tool("GetCurrentPrice") def get_current_price(symbol: str) -> str: """here is the doc string""" url = f'https://www.marketwatch.com/investing/stock/{symbol}' response = requests.get(url, headers=HEADERS) if response.status_code == 200: soup = BeautifulSoup(response.text, 'html.parser') price_tag = soup.find('bg-quote', class_='value') if price_tag: price = price_tag.get_text(strip=True) return f"{symbol.upper()} current price: ${price}" else: return "Stock price not found." else: return f"Failed to retrieve stock page. Status code: {response.status_code}" @tool("GetHistoricalData") def get_historical_data(symbol: str, days: int = 5000) -> str: """here is it""" end_date = datetime.now() start_date = end_date - timedelta(days=days) start_date_str = start_date.strftime('%m/%d/%Y 00:00:00') end_date_str = end_date.strftime('%m/%d/%Y 23:59:59') csv_url = ( f'https://www.marketwatch.com/investing/stock/{symbol}/downloaddatapartial' f'?csvdownload=true&downloadpartial=false' f'&startdate={start_date_str}&enddate={end_date_str}' f'&frequency=p1d&newdates=false' ) response = requests.get(csv_url, headers=HEADERS) if response.status_code == 200: try: df = pd.read_csv(StringIO(response.text)) return df.head(10).to_csv(index=False) except Exception as e: return f"Failed to parse CSV data: {e}" else: return f"Failed to download historical data. Status code: {response.status_code}" @tool("GetTechnicalAnalysisDocs") def get_technical_analysis_docs(_: str = "") -> str: """Get documentation for technical analysis indicators.""" return """ Technical Analysis: Core Concepts & Formulas ------------------------------------------- - Market Action Discounts Everything: All known information is reflected in price. - Prices Move in Trends: Uptrend, downtrend, or sideways movement. - History Repeats Itself: Psychological patterns repeat. Key Technical Indicators and Their Formulas: ------------------------------------------- 1. Simple Moving Average (SMA): SMA(time_period) = Sum(Price_t ... Price_{t-n}) / n 2. Exponential Moving Average (EMA): EMA_t = (Price_t * α) + EMA_{t-1} * (1 - α), where α = 2/(n+1) 3. Relative Strength Index (RSI): RSI = 100 - [100 / (1 + Avg Gain / Avg Loss)] 4. MACD (Moving Average Convergence Divergence): MACD Line = 12-period EMA - 26-period EMA; Signal Line = 9-period EMA of MACD Line; Histogram = MACD Line - Signal Line 5. Stochastic Oscillator (STOCH): Fast K = (Current Close - Lowest Low) / (Highest High - Lowest Low) * 100 over N periods; Slow K = 3-day SMA of Fast K; Slow D = 3-day SMA of Slow K 6. Momentum (MOM): MOM = Current Close - Close_N_days_ago 7. Rate of Change (ROC): ROC = [(Current Close - Prior Close) / Prior Close] * 100 8. Volume Weighted Average Price (VWAP): VWAP = Sum(Price * Volume) / Sum(Volume) over intraday period 9. Bollinger Bands: Middle Band = 20-day SMA; Upper Band = 20-day SMA + 2 * 20-day Standard Deviation; Lower Band = 20-day SMA - 2 * 20-day Standard Deviation 10. Ichimoku Cloud: Tenkan-sen = (9-period high + low)/2; Kijun-sen = (26-period high + low)/2; Senkou Span A = (Tenkan-sen + Kijun-sen)/2 shifted forward by 26; Senkou Span B = (52-period high + low)/2 shifted forward by 26; Chikou Span = Current close shifted back by 26 days 11. **Williams %R**: %R = (Highest High - Close) / (Highest High - Lowest Low) * -100 over N periods 12. **Commodity Channel Index (CCI)**: CCI = (Typical Price - 20-day SMA of TP) / (0.015 * Mean Deviation) where Typical Price = (High + Low + Close) / 3 13. **Average Directional Index (ADX)**: ADX = Smoothed average of DX values, which measure directional strength 14. **On-Balance Volume (OBV)**: OBV = previous OBV + volume if close > previous close, else -volume 15. **Moving Average Convergence Divergence (MACD)**: MACD Line = 12-day EMA - 26-day EMA Signal Line = 9-day EMA of MACD Line MACD Histogram = MACD Line - Signal Line 16. **Absolute Price Oscillator (APO)**: APO = Fast EMA - Slow EMA 17. **Balance of Power (BOP)**: BOP = (Close - Open) / (High - Low) 18. **Triple Exponential Moving Average (TEMA)**: TEMA = (3 * EMA1) - (3 * EMA2) + EMA3 where EMA1 = fast EMA, EMA2 = slower EMA, etc. 19. **Double Exponential Moving Average (DEMA)**: DEMA = 2*EMA1 - EMA2 20. **Kaufman Adaptive Moving Average (KAMA)**: KAMA = prior KAMA + SC * (price - prior KAMA) where SC = smoothing constant based on efficiency ratio 21. **Chaikin Money Flow (CMF)**: MF Multiplier = [(Close - Low) - (High - Close)] / (High - Low) MF Volume = MF Multiplier * Volume CMF = Sum(MF Volume) / Sum(Volume) over N days 22. **Aroon Indicator**: Aroon Up = ((N - Periods Since Highest Close) / N) * 100 Aroon Down = ((N - Periods Since Lowest Close) / N) * 100 23. **Parabolic SAR**: SAR_t = SAR_{t-1} + AF * (EP - SAR_{t-1}) 24. **Standard Deviation (Volatility)**: σ = sqrt[1/N * Σ(Close_i - μ)^2] 25. **Candlestick Patterns**: - Hammer, Shooting Star, Engulfing, Doji, Morning/Evening Star, etc. """ # === Register Tools === tools = [ Tool.from_function(get_current_price, name="GetCurrentPrice", description="Get current stock price."), Tool.from_function(get_historical_data, name="GetHistoricalData", description="Get stock historical data as CSV."), Tool.from_function(get_technical_analysis_docs, name="GetTechnicalAnalysisDocs", description="Technical indicator docs.") ] # === Callback handler to stream reasoning to chat === class ReasoningCallbackHandler(BaseCallbackHandler): def __init__(self, chat_callback): self.chat_callback = chat_callback def on_tool_start(self, serialized, input_str, **kwargs): self.chat_callback(f"🛠️ Using Tool: {serialized['name']} with input: {input_str}") def on_tool_end(self, output, **kwargs): self.chat_callback(f"✅ Tool output: {output}") def on_llm_new_token(self, token, **kwargs): if token.strip(): self.chat_callback(token) # === OpenAI-compatible LLM via OpenRouter === llm = ChatOpenAI( openai_api_key=OPENROUTER_API_KEY, openai_api_base="https://openrouter.ai/api/v1", model_name="google/gemma-3n-e4b-it:free" ) # === Agent Execution with live reasoning === def run_agent(question, chat_callback): callback_handler = ReasoningCallbackHandler(chat_callback) agent_executor = initialize_agent( tools, llm, agent="zero-shot-react-description", verbose=True, callbacks=[callback_handler] ) try: result = agent_executor.run(question) except Exception as e: result = f"❌ Error: {str(e)}" chat_callback(f"✅ Final Answer: {result}") # === Gradio UI === with gr.Blocks(theme=gr.themes.Monochrome()) as demo: gr.Markdown("## 📈 Stock Market Analysis Agent") chatbot = gr.Chatbot(label="Stock Agent Chat") with gr.Row(): user_input = gr.Textbox(label="Type your question here...", scale=4) submit_btn = gr.Button("Submit", scale=1) def respond(msg, chat_history): chat_history.append((msg, "")) def chat_callback(new_msg): chat_history.append(("", new_msg)) chatbot.value = chat_history run_agent(msg, chat_callback) return chat_history submit_btn.click(respond, [user_input, chatbot], [chatbot]) demo.launch()