Sentiment Analysis API

A FastAPI service that classifies text sentiment with Hugging Face transformers.

Overview

• Single-text and batch prediction endpoints
• Startup model loading with a health endpoint
• Dockerfile included for containerized runs

The current app uses cardiffnlp/twitter-roberta-base-sentiment-latest through the Hugging Face pipeline API.

Stack

• Python
• FastAPI
• Hugging Face Transformers
• PyTorch
• Uvicorn
• Docker

API Endpoints

Method	Route	Purpose
POST	/predict	Predict sentiment for one text
POST	/predict/batch	Predict sentiment for up to 32 texts
GET	/health	Check service and model readiness

Quick Start

Local

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app:app --reload --port 8000

Open http://localhost:8000/docs for the interactive API docs.

Docker

docker build -t sentiment-api .
docker run -p 8000:8000 sentiment-api

Example Request

curl -X POST http://localhost:8000/predict \
  -H "Content-Type: application/json" \
  -d '{"text": "This movie was absolutely fantastic!"}'

Example response:

{
  "text": "This movie was absolutely fantastic!",
  "label": "positive",
  "score": 0.9998,
  "model": "cardiffnlp/twitter-roberta-base-sentiment-latest"
}

Project Files

sentiment-api/
├── app.py
├── requirements.txt
├── Dockerfile
├── .dockerignore
└── README.md

Notes

• The first startup downloads the model and can take longer than later runs.
• The app is configured for CPU by default. Change device=-1 to device=0 in app.py if you want GPU inference.