🚀 AlphaPredict: Full-Stack Stock Prediction Platform

AlphaPredict is an advanced, full-stack stock analysis platform featuring a decoupled, cloud-native architecture. It leverages a high-performance LightGBM model for predictions, served via a FastAPI backend that is automatically built, containerized, and deployed to Microsoft Azure using a complete CI/CD pipeline.

✨ Core Features

• Azure-Native Backend: The containerized FastAPI backend is deployed to Azure App Service, demonstrating enterprise-grade cloud skills.
• Automated CI/CD Pipeline: On every push to main, a GitHub Actions workflow automatically builds a Docker image, pushes it to Azure Container Registry, and deploys the new version to production.
• Advanced ML Model: Utilizes a LightGBM Gradient Boosting model for superior prediction accuracy.
• Decoupled Data Pipeline: A separate, scheduled GitHub Action handles all data ingestion, ensuring the live API is never impacted by heavy data-fetching tasks.
• Interactive UI: A sleek, modern dashboard built with React and TailwindCSS, hosted on Vercel's global edge network.

🏛️ System Architecture

The application is composed of distinct services that work in concert. This decoupled design is a key feature, enhancing scalability, resilience, and performance.

graph TD
    subgraph "User Interaction"
        A[User] -->|Visits Site| FE(React Frontend @ Vercel)
    end

    subgraph "Live Application (Microsoft Azure)"
        FE -->|API Calls| BE(FastAPI Backend @ Azure App Service)
        BE -->|Reads Data| DB[(PostgreSQL @ Supabase)]
    end

    subgraph "CI/CD & Automation (GitHub Actions)"
        GHA_Data(Data Pipeline) -->|Writes Daily Prices| DB
        GHA_CICD(Backend CI/CD) -->|Deploys Container| BE
    end

    style FE fill:#61DAFB,stroke:#000,stroke-width:2px
    style BE fill:#0078D4,stroke:#000,stroke-width:2px
    style DB fill:#336791,stroke:#000,stroke-width:2px
    style GHA_Data fill:#2088FF,stroke:#000,stroke-width:2px
    style GHA_CICD fill:#2088FF,stroke:#000,stroke-width:2px

Loading

DevOps & CI/CD Highlights

This project demonstrates several core DevOps principles:

• CI/CD for Applications: The .github/workflows/deploy-to-azure.yml workflow automates the build, test, and deployment of the containerized backend to Azure on every push to main.
• CI/CD for Data: The .github/workflows/daily_data_fetch.yml pipeline manages the automated, scheduled ingestion of data, treating data operations as a first-class citizen.
• Infrastructure as Code (IaC): The database schema is version-controlled (data_pipeline/db_setup.sql), and the entire cloud deployment is defined as code within GitHub Actions workflows.
• Containerization: The backend is fully containerized with Docker, ensuring consistency from local development to production on Azure.
• Configuration Management: All secrets are managed securely using GitHub Secrets and Azure App Service configuration, following 12-Factor App principles.

🛠️ Tech Stack

Category	Technology
Frontend	React 19, Vite, TailwindCSS, Recharts
Backend	FastAPI, Python 3.11, Docker
Database	PostgreSQL (hosted on Supabase)
ML Model	LightGBM, Scikit-learn, Pandas
CI/CD	GitHub Actions
Deployment	Vercel (Frontend), Azure App Service (Backend), Azure Container Registry

🌐 Deployment

This project is deployed as a set of independent, communicating services.

1. Backend (Azure App Service via CI/CD)

The backend is deployed automatically by the .github/workflows/deploy-to-azure.yml workflow.

Setup required:

1. Create Azure Resources:
   • An Azure Container Registry (ACR) to store the Docker images.
   • An Azure App Service configured to run Docker containers.
2. Create Azure Credentials: Generate a Service Principal for GitHub Actions to use.

   # Install Azure CLI, then run:
   az login
   az ad sp create-for-rbac --name "GitHubActions-StockPredictor" --role contributor --scopes /subscriptions/{your-subscription-id} --sdk-auth

3. Configure GitHub Secrets:
   • In your GitHub repo, go to Settings > Secrets and variables > Actions.
   • Create AZURE_CREDENTIALS and paste the JSON output from the previous step.
   • Create DATABASE_URL, SECRET_KEY, etc., which will be used by the data pipeline.
4. Configure App Service: In the Azure Portal, add your DATABASE_URL, SECRET_KEY, etc., to the App Service's configuration.

2. Frontend (Vercel)

1. Import your repository on Vercel.
2. Configure:
   • Framework Preset: Vite
   • Root Directory: frontend
3. Add environment variables:
   • VITE_API_BASE: Your backend URL from Azure App Service.
   • VITE_GOOGLE_CLIENT_ID: Your Google Client ID.
4. Deploy.

⚙️ Local Development

(Instructions for local setup remain the same, allowing developers to work on any component without needing a full cloud environment.)

Prerequisites

• Python 3.11+
• Node.js 18+
• A PostgreSQL database instance

Setup

1. Clone & Navigate: git clone https://github.com/dhruvmohan867/stock-predictor.git && cd stock-predi
2. Activate Python Env: python -m venv venv && source venv/bin/activate (or .\venv\Scripts\activate on Windows)
3. Install Dependencies: pip install -r backend/requirements.txt -r data_pipeline/requirements.txt
4. Setup Database: Run the script from data_pipeline/db_setup.sql on your database.
5. Run Data Pipeline: python data_pipeline/fetch_data.py
6. Train Model: python ml_model/train.py
7. Run Backend: uvicorn backend.main:app --reload
8. Run Frontend: cd frontend && npm install && npm run dev

📝 License

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