FinanceGuard AI

FinanceGuard AI is a distributed system for sentiment analysis of financial news. It monitors configurable assets (Stocks, Crypto), analyzes news sentiment using FinBERT, and triggers notifications on significant market events.

Architecture

The system consists of two microservices:

1. Sentinel Service:

   • Aggregates news via RSS (Google News, Yahoo Finance) and yfinance
   • Prevents duplicates via hashing and SQLite caching
   • Sends Telegram notifications

2. API Service:

   • Hosts the ML model (ProsusAI/finbert) via FastAPI
   • Asynchronous sentiment analysis of news titles

Tech Stack

• Python 3.11
• FastAPI / Uvicorn
• Transformers / PyTorch (FinBERT)
• Docker & Docker Compose
• SQLite (Persistence)

Installation & Usage

Prerequisites: Docker and Docker Compose.

1. Clone repository and create .env file:

   cp .env.example .env
   # Customize .env (Telegram Token etc.)

2. Start containers:

   docker-compose up --build -d

3. Monitor logs:

   docker-compose logs -f

Configuration

Configuration is handled via environment variables in .env:

• WATCHLIST: Assets to monitor
• SCAN_INTERVAL_MINUTES: Scan interval
• SENTIMENT_THRESHOLD: Alert threshold (0.0 - 1.0)