LangDetect API

NLP-powered language detection service — trained on a custom dataset, served via FastAPI, containerized with Docker, deployed on Railway.

Built end-to-end: from raw data and model training in Jupyter to a live production endpoint with a web UI.

Why it exists: My first complete ML project shipped to production — not a notebook, not a tutorial result. A running service.

Live Demo 🚀

• UI: https://full-ml-pipeline-production.up.railway.app/ui
• Health check: https://full-ml-pipeline-production.up.railway.app/
• API docs (Swagger): https://full-ml-pipeline-production.up.railway.app/docs

Features ✨

• Language prediction endpoint for user text
• Clean web UI for manual testing
• FastAPI OpenAPI documentation
• Dockerized deployment flow
• Railway hosting support

Project Structure

.
├── Dockerfile
├── README.md
└── app
    ├── main.py
    ├── requirements.txt
    ├── static
    │   └── index.html
    └── model
        ├── model.py
        ├── trained_pipeline-0.1.0.pkl
        └── LangDetetctionModel.ipynb

API 🤖

GET /

Returns service status and model version.

Example response:

{
  "health_check": "OK",
  "model_version": "0.1.0"
}

GET /ui

Serves the web interface for submitting text and viewing predictions.

POST /predict

Predicts the language of the provided text.

Request body:

{
  "text": "Bonjour, comment allez-vous aujourd'hui?"
}

Response:

{
  "language": "French"
}

Run Locally (Docker)

Build image:

docker build -t language-detection-app .

Run container:

docker run --rm -p 80:80 language-detection-app

Open in browser:

• http://localhost/ui
• http://localhost/docs

Quick Test with curl

curl -X POST "http://localhost/predict" \
  -H "Content-Type: application/json" \
  -d '{"text":"Hello, how are you today?"}'

Deployment Notes

• Hosted on Railway using the repository Dockerfile.
• For lower startup overhead on small instances, worker settings can be tuned with:
  • WEB_CONCURRENCY=1
  • MAX_WORKERS=1

Tech Stack

• Python 3.9
• FastAPI
• scikit-learn
• NumPy
• Docker
• Railway

Author

Adam Vakar