🚀 Stockify - AI-Powered Stock Market Intelligence Platform

Stockify is a comprehensive AI-powered stock market intelligence platform that combines advanced machine learning, natural language processing, and document analysis for complete investment research. Built with privacy as a core principle, it leverages:

• 🧠 LSTM Neural Networks for accurate 50/100/200-day price predictions
• 🤖 Local LLM Integration via LM Studio for on-device AI processing
• 📚 RAG Technology for intelligent document analysis and Q&A
• 📊 Technical Analysis Tools for in-depth market insights

Whether you're analyzing US stocks (NASDAQ) or Indian markets (NSE), Stockify provides AI-powered recommendations, real-time data visualization, and document-based research capabilities—all while keeping your data completely private on your local machine.

---

✨ Features & Pages

1. 🏠 Main Dashboard - Stock Price Predictions

Core Features:

• ✅ LSTM Price Predictions - 50/100/200-day forecasts trained on 2012-2022 market data
• ✅ Real-time Stock Data - Live market data with dynamic date fetching using Yahoo Finance (up to current date)
• ✅ Historical Charts - Interactive price visualization and trend analysis
• ✅ AI Market Insights - Automated analysis of stock performance

2. 📄 Financial Report Analysis - RAG-Powered Q&A

AI RAG Features:

• ✅ Document Q&A - Ask questions about financial PDFs in natural language
• ✅ Source Citations - Answers grounded in actual document content
• ✅ Local LLM Processing - Privacy-first AI analysis (no data leaves your machine)
• ✅ Vector Embeddings - FAISS-powered semantic search for accurate retrieval

3. 📈 Investment Summary - AI Recommendations

Investment Features:

• ✅ BUY/HOLD/SELL Signals - Intelligent recommendations with reasoning
• ✅ Risk Assessment - Comprehensive risk analysis and evaluation
• ✅ Entry/Exit Points - Strategic price targets for trading
• ✅ Streaming Responses - Real-time AI response generation

4. 📋 Stock Tickers List - Market Browse

Data Features:

• ✅ Multi-Market Support - 1000+ NASDAQ and 1000+ NSE stocks
• ✅ Search & Filter - Easy stock discovery and selection
• ✅ Comprehensive Listings - Full stock metadata and information

5. 📊 Technical Analysis - Indicators & Trends

Technical Features:

• ✅ Moving Averages - 50-day, 100-day and 200-day MA calculations
• ✅ Volatility Analysis - Price fluctuation and risk metrics
• ✅ Trend Detection - Automated pattern recognition
• ✅ AI Interpretation - Natural language explanations of technical data

---

🏗️ Architecture

Streamlit UI → Services (LocalLLM, Embeddings, RAG) → Utils (Stock LSTM, PDF)

Tech Stack: Streamlit • Python • TensorFlow/Keras • LangChain • FAISS • LM Studio • yfinance • PyPDF2 • Pandas • NumPy • Google Colab

---

💻 Installation

Prerequisites: Python 3.8+, Google Colab, LM Studio, 8GB+ RAM

# Clone and setup
git clone https://github.com/Preveen369/Stock-Price-Prediction.git
cd Stock-Price-Prediction
python -m venv venv
venv\Scripts\activate  # Windows 

# Install dependencies
pip install -r requirements.txt
pip install -r rag_requirements.txt

# Setup LM Studio (lmstudio.ai)
# Load models: gemma-3-4b + text-embedding-nomic-embed-text-v1.5
# Start server: http://localhost:1234

# Run application
streamlit run app.py  # Opens at http://localhost:8501

---

⚙️ Configuration

• Environment (.env): LM_STUDIO_URL, LM_STUDIO_MAX_TOKENS, LM_STUDIO_TEMPERATURE, LM_STUDIO_TIMEOUT

• LM Studio: Load gemma-3-4b (chat) + nomic-embed-text-v1.5 (embeddings) → Start server at http://localhost:1234

• Settings: See config/llm_config.py for defaults (max_tokens: 2048, temperature: 0.3)

---

🚦 Usage

Quick Start Guide

# 1. Ensure LM Studio is running with models loaded
# 2. Start Stockify
streamlit run app.py

# 3. Open browser at http://localhost:8501

Workflow Examples

• Analyze Stock: Main Dashboard → Enter Symbol (AAPL) → View Charts & Predictions
• NASDAQ | NSE Stock Symbols: US (AAPL, TSLA) | India (INFY.NS, TCS.NS)
• Technical Analysis: Technical Analysis Page → View Moving Averages & Trends
• Investment Advice: Investment Summary → Risk Assessment → AI Recommendations
• Document Q&A: Financial Report → Upload PDF → Ask Questions → Get Answers

---

📚 API Documentation

For detailed API documentation, see API_DOCUMENTATION.md

Key Modules

Services

Core services layer that handles AI/ML operations and external integrations.

• LocalLLMService - LM Studio integration for AI analysis
• LMStudioEmbeddings - Text embedding generation
• RAGPipeline - Retrieval-augmented generation pipeline

Utilities

Helper functions for data processing, calculations, and file operations.

• stock_utils.py - Stock data fetching and analysis
• pdf_utils.py - PDF processing and chunking

---

📁 Project Structure

Stock-Price-Prediction/
├── 📄 app.py                            # Main dashboard application
├── 📁 pages/                        # Streamlit pages
│   ├── 📄 Financial_Report_Analysis.py  # RAG-based PDF analysis
│   ├── 📄 Investment_Summary.py         # AI investment recommendations
│   ├── 📄 Stock_Tickers_List.py         # Browse stock symbols
│   └── 📄 Technical_Analysis.py         # Technical indicators analysis
├── 📁 services/                     # Core services layer
│   ├── 📄 local_llm_service.py          # LM Studio integration
│   ├── 📄 embeddings_service.py         # Vector embedding generation
│   └── 📄 rag_pipeline.py               # RAG implementation
├── 📁 utils/                        # Utility functions
│   ├── 📄 stock_utils.py                # Stock data and metrics
│   └── 📄 pdf_utils.py                  # PDF processing utilities
├── 📁 config/                       # Configuration
│   └── 📄 llm_config.py                 # LLM settings and env vars
├── 📁 models/                       # Pre-trained models
│   └── 📄Stock Predictions Model.keras  # LSTM neural network
├── 📁 resources/                    # Static resources
│   ├── 📄 NASDAQ_Active_Stocks_List.csv # US stock listings
│   └── 📄 NSE_Active_Stocks_List.csv    # Indian stock listings
├── 📄 requirements.txt                  # Core dependencies
├── 📄 stock_market_prediction.ipynb     # LSTM model training notebook
├── 📄 stocks_list_extraction.ipynb      # Stock tickers extraction notebook
├── 📄 API_DOCUMENTATION.md              # Complete API reference
└── 📄 README.md                         # This file

Module Responsibilities

Module	Responsibility
app.py	Main dashboard, stock selection, LSTM predictions
services/	LLM integration, embeddings, RAG pipeline
utils/	Stock data fetching, calculations, PDF processing
pages/	Individual analysis pages (UI components)
config/	Environment variables, LLM configuration
models/	Pre-trained LSTM model for predictions

---

🧠 Key Concepts Explained

1. LSTM (Long Short-Term Memory)

• What: Neural network architecture for time-series prediction
• How: Uses 100 days of historical prices to predict next 50/100/200 days
• Framework: Built with TensorFlow/Keras for deep learning
• Preprocessing: Scikit-learn for data scaling and normalization
• Accuracy: Typically achieves 75-90% accuracy on test data
• Training: Pre-trained on 2012-2022 historical data across thousands of stock patterns

2. RAG (Retrieval-Augmented Generation)

  Document → Chunks → Embeddings → Vector Store
                                        ↓
         Question → Embedding → Similarity Search → Retrieved Chunks
                                        ↓
                        Chunks + Question → LLM → Answer

• Benefit: Accurate answers grounded in document content
• Advantage: No hallucinations, includes source citations

3. Vector Embeddings

• What: Numerical representations of text (384 dimensions)
• Purpose: Enable semantic similarity search
• Model: Nomic Embed Text v1.5
• Use Case: Finding relevant document sections for RAG

4. Local LLM Processing

• Privacy: All AI processing happens on your machine
• Speed: Depends on your hardware (GPU recommended)
• Models: Gemma 3, Mistral, Llama, etc.
• Benefits: No API costs, full control, offline capable

---

🔧 Key Technologies

• LSTM: 100-day history → 50/100/200-day predictions (75-90% accuracy)
• RAG: Document chunks + embeddings + vector search → cited answers
• Embeddings: 384D vectors (Nomic v1.5)
• Local LLM: Privacy-first AI (Gemma/Mistral/Llama)

---

🛠️ Troubleshooting

Common Issues and Solutions

❌ LM Studio Not Connected

Symptoms: "LM Studio Not Connected" error in sidebar,

Solutions:

1. Verify LM Studio is running
2. Check server URL is http://127.0.0.1:1234

❌ Slow Performance

Symptoms: Long processing times, lag

Solutions:

1. Use smaller chunk sizes for PDFs (500 instead of 1000)
2. Reduce top_k in RAG queries (3 instead of 4)
3. Use lighter LLM models (4B instead of 7B)
4. Close other applications
5. Enable GPU acceleration in LM Studio

❌ Memory Errors

Symptoms: Out of memory, crashes

Solutions:

1. Use quantized models (Q4 or Q5)
2. Reduce max_tokens in config (1024 instead of 2048)

---

🤝 Contributing

Pull requests are welcome! Feel free to fork the repository and suggest improvements.

Steps to contribute:

# 1. Fork the repository
# 2. Create a feature branch
git checkout -b feature-name
# 3. Commit your changes
git commit -m "Add feature description"
# 4. Push to GitHub
git push origin feature-name
# 5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

⚠️ Disclaimer

For educational purposes only. NOT financial advice. Past performance ≠ future results. AI predictions may be inaccurate. Consult qualified advisors. Use at your own risk. No warranties or liability.

---

📧 Contact

For queries or suggestions:

• 📩 Email: spreveen123@gmail.com
• 🌐 LinkedIn: www.linkedin.com/in/preveen-s

---

🌟 Show Your Support

If you like this project, please consider giving it a ⭐ on GitHub!

---

Built with ❤️ using DL/ML LSTM + Local AI LLMs + RAG