How to Build a Real-Time Market Sentiment Analyzer with FastAPI, VADER NLP & Docker

The Project

EchoSentinel is an autonomous, end-to-end market sentiment analysis platform. It harvests global news in real-time, analyzes emotional context using NLP, and surfaces market trends through a live dashboard — all running in a fully containerized environment.

The problem it solves: financial and market data is noisy, scattered, and slow to react. By the time a human reads the news and forms an opinion, the market has already moved. EchoSentinel automates that pipeline end-to-end.

Key Features

Autonomous Ingestion

A dedicated background worker (watcher.py) monitors high-value keywords 24/7 — Bitcoin, AI, Tesla, inflation, interest rates, and more. It polls NewsAPI on a schedule, fetches fresh headlines, and pushes them into the pipeline without human intervention.

# Simplified watcher loop
import time
from harvester import fetch_news, analyze_sentiment

KEYWORDS = ["Bitcoin", "AI", "Tesla", "Inflation", "Fed"]

while True:
    for keyword in KEYWORDS:
        articles = fetch_news(keyword, hours_back=1)
        for article in articles:
            score = analyze_sentiment(article["title"])
            store_in_db(article, score)
    time.sleep(3600)  # Run every hour

AI Sentiment Engine

Powered by the VADER (Valence Aware Dictionary and sEntiment Reasoner) NLP model, each headline is classified into Positive, Negative, or Neutral with a compound sentiment score. VADER is specifically tuned for social media and news text, making it ideal for headline analysis.

The engine doesn't just classify — it quantifies. Each headline gets a score from -1 (extremely negative) to +1 (extremely positive), and the dashboard aggregates these scores over time to show sentiment trends.

Persistent Data Vault

Using SQLAlchemy ORM with SQLite, every article, score, and timestamp is stored permanently. This enables historical trend analysis — you can look back at sentiment patterns before major market events and correlate them.

from sqlalchemy import create_engine, Column, String, Float, DateTime
from sqlalchemy.ext.declarative import declarative_base

Base = declarative_base()

class Article(Base):
    __tablename__ = "articles"
    
    id = Column(String, primary_key=True)
    title = Column(String)
    source = Column(String)
    keyword = Column(String)
    sentiment_score = Column(Float)
    sentiment_label = Column(String)  # Positive / Negative / Neutral
    published_at = Column(DateTime)
    url = Column(String)

Real-Time Dashboard

A modern, responsive interface built with vanilla JavaScript (ES6+), CSS3 Flexbox/Grid, and HTML5. The dashboard calculates market "Bullish/Bearish" scores on the fly by aggregating recent sentiment data.

Cloud-Ready Architecture

Fully containerized with Docker and Docker Compose. One command spins up the entire stack:

docker-compose up --build

This makes deployment to any VPS or cloud environment trivial — no dependency hell, no manual setup.

System Architecture

EchoSentinel follows a Distributed Worker Pattern:

┌─────────────────────────────────────────────┐
│              Docker Container                │
│                                              │
│  ┌──────────┐   ┌───────────────────────┐   │
│  │ Watcher  │──→│  FastAPI + SQLAlchemy │   │
│  │ (Worker) │   │  (Web Server + DB)    │   │
│  └──────────┘   └──────────┬────────────┘   │
│                            │                 │
│                   ┌────────▼────────┐        │
│                   │  SQLite DB      │        │
│                   └─────────────────┘        │
│                            │                 │
│                   ┌────────▼────────┐        │
│                   │  Static UI      │        │
│                   │  (JS Dashboard) │        │
│                   └─────────────────┘        │
└──────────────────────────────────────────────┘

• The Engine (Harvester) — Core logic that communicates with NewsAPI and performs VADER NLP analysis
• The Scout (Watcher) — Autonomous background process that populates the database while idle
• The Waiter (FastAPI) — High-performance web server bridging database and user
• The Face (UI) — Dynamic frontend calculating Bullish/Bearish scores on the fly

Tech Stack

Installation & Setup

git clone https://github.com/MUIZ-UDDIN/EchoSentinel.git
cd EchoSentinel

# Add your NewsAPI key to .env
echo "API_KEY=your_news_api_key_here" > .env

# Launch
docker-compose up --build

Open http://localhost:8000/index.html and the dashboard is live.

Testing

uv run pytest

What I Learned

1. VADER punches above its weight — For headline-level sentiment, VADER's lexicon-based approach is surprisingly accurate and orders of magnitude faster than transformer models.

2. Worker patterns keep the UI fast — By separating ingestion (watcher) from serving (FastAPI), the dashboard never blocks on data collection.

3. Docker Compose is perfect for side projects — One-file deployment makes the project instantly portable anywhere.

4. SQLAlchemy adds unnecessary complexity for SQLite — For a single-user app, raw SQL or a lighter ORM would have been simpler. But SQLAlchemy made it easy to migrate to PostgreSQL later if needed.