🤖 Sarcasm Detector Bot

An MLOps-ready sarcasm detection API using a HuggingFace T5 model, FastAPI for serving predictions, Docker for containerization, and GitHub Actions for CI/CD. This project automatically trains the model, packages it, serves it with FastAPI, and tests its behavior using a sample input.

---

📌 Features

• 🔍 Detects sarcasm using mrm8488/t5-base-finetuned-sarcasm-twitter
• 🚀 Dockerized FastAPI app for deployment anywhere
• 🔄 GitHub Actions CI/CD pipeline that:
  • Trains and serializes the model
  • Uploads the model as an artifact
  • Builds and runs the API in a container
  • Sends a test input and asserts expected output

---

🧱 Project Structure

sarcasm-detector/
├── app/
│   ├── main.py                # FastAPI application
│   ├── model/                 # Contains saved model (.pkl)
│   └── requirements.txt       # Python dependencies
├── train_model.py            # Loads and saves the sarcasm detection model
├── test_model.py             # Sends test request to API and validates output
├── Dockerfile                # Dockerfile for containerizing the backend
├── requirements.txt          # Root-level dependencies
├── .github/workflows/
│   └── docker-ci.yml         # GitHub Actions pipeline

---

🐳 Local Setup

1. Clone the Repo

git clone https://github.com/ShouryaAswal/Sarcasm_detector.git
cd Sarcasm_detector

2. Build Docker Image

docker build -t sarcasm-detector .

3. Run the Container

docker run -p 8000:8000 sarcasm-detector

4. Test the API

curl -X POST http://localhost:8000/predict \
     -H "Content-Type: application/json" \
     -d '{"text": "Oh wow, this is just *perfect*!"}'

---

🧮 GitHub Actions CI/CD Workflow

• Job 1: build-model

  • Sets up Python
  • Installs dependencies from requirements.txt
  • Runs train_model.py
  • Uploads .pkl model as an artifact

• Job 2: build-and-test

  • Downloads artifact
  • Builds Docker image
  • Runs container exposing API
  • Installs test dependencies
  • Runs test_model.py which:
    • Sends a request to /predict
    • Compares the output against expected value ("derison")
    • Fails the build if the output doesn’t match

✅ Automatically triggered on every push to main

---

📄 API Endpoint

POST /predict

Request Body:

{
  "text": "Oh fantastic, another Monday meeting that could’ve been an email."
}

Response:

{
  "input": "Oh fantastic, another Monday meeting that could’ve been an email.",
  "prediction": {
    "sarcasm": "derison"
  }
}

"derison" means the model found sarcasm; "normal" means it didn't.

---