🧠 Sentiment Radar

An end-to-end NLP web app that classifies product reviews as positive or negative in real time using a scikit-learn pipeline, MLflow tracking, a FastAPI backend, and a modern vanilla frontend.

LIVE DEMO

---

✨ Features

• NLP pipeline — text cleaning, TF-IDF vectorization, and Logistic Regression for binary sentiment classification
• Experiment tracking — MLflow logs parameters, metrics, and model artifacts for each run
• REST API — FastAPI backend with /predict and /health endpoints
• Modern UI — responsive dark interface built with plain HTML, CSS, and JavaScript
• Docker ready — single-container deployment with a Render-friendly PORT setup
• Render ready — easy cloud deployment with a stable public URL

---

🛠 Tech stack

Layer	Technology
Model	scikit-learn · LogisticRegression · TF-IDF
Tracking	MLflow
API	FastAPI · Uvicorn
Frontend	Vanilla HTML · CSS · JavaScript
Container	Docker
Deploy	Render

---

📁 Project structure

sentiment-radar/
├── assets/
│   └── website.png           # Frontend screenshot used in this README
├── data/                      # Local dataset and SQLite database files
├── mlruns/                    # MLflow run artifacts
├── src/
│   ├── api.py                 # FastAPI app and prediction endpoints
│   ├── config.py              # Paths and environment configuration
│   ├── db.py                  # SQLAlchemy models and SQLite engine
│   ├── seed_db.py             # Dataset seeding helper
│   ├── train.py               # Training pipeline + MLflow logging
│   └── static/
│       ├── index.html         # Frontend entry point
│       ├── styles.css         # UI styles
│       └── app.js             # Frontend logic
├── Dockerfile
├── requirements.txt
└── README.md

---

🚀 Quick start

Option 1: Run with Docker

docker build -t sentiment-radar .
docker run -p 10000:10000 sentiment-radar

Open:

• http://localhost:10000

Option 2: Run locally without Docker

pip install -r requirements.txt

Then start the API:

set PYTHONPATH=src
uvicorn api:app --host 0.0.0.0 --port 8000 --app-dir src

Open:

• http://localhost:8000

If you open the frontend with Live Server, keep the API running separately on port 8000 or 10000.

---

🧪 Model workflow

1. Reviews are stored in the local SQLite database.
2. src/train.py loads the data and trains a text classification pipeline.
3. MLflow logs the experiment run and model artifact.
4. src/api.py loads the latest trained model and serves predictions through the API.

Training the model

set PYTHONPATH=src
python src/train.py

---

🌐 Deploy on Render

1. Push the repository to GitHub.
2. Go to render.com and create a New Web Service.
3. Connect your GitHub repository.
4. Select Docker as the environment.
5. Deploy the service.
6. Render will assign a public URL like https://your-app.onrender.com.

The container is configured to respect the PORT environment variable, which is the safest setup for Render.

---

📡 API endpoints

Method	Endpoint	Description
GET	/	Serves the frontend
GET	/health	Returns API and model status
POST	/predict	Returns the predicted sentiment

Example /predict request

{
  "text": "I absolutely love this phone. The screen is sharp and the battery lasts all day."
}

Example response

{
  "sentiment": "Positive",
  "confidence": 0.97
}

---

🖼 Frontend preview

The UI is intentionally minimal, modern, and responsive. It includes:

• a clear hero section
• a live prediction card
• sample review buttons
• confidence visualization
• green/red animated feedback based on the prediction

---

👤 Author

@raess1593