RDAI-AI-in-Production

Project: Text Summarisation with FastAPI backend

Overview

This project implements a text summarisation API using FastAPI and the facebook/bart-large-cnn model from Hugging Face’s transformers library. The API enables users to submit text input and receive a concise summary based on configurable parameters for summary length. The application is containerized using Docker. This allows it to run consistently across different environments. Once deployed, the container exposes the API on port 8000, allowing the user to interact via RESTful requests or use the Swagger UI for testing.

Model

The summarisation model is implemented in model/model.py. It uses the facebook/bart-large-cnn model from Hugging Face to generate text summaries. The model takes in input text and produces a summary within the specified length range.

Key Functions:

• summarise_text(text: str, max_length: int, min_length: int) -> str: Summarises the given text based on length constraints.
• main(): Command-line interface to summarise text.

Backend Application

The FastAPI backend is defined in app/main.py. It exposes an API endpoint to serve summarisation requests.

Endpoints:

• GET / - Returns a welcome message.
• POST /summarise/ - Accepts text input and returns a summarised version.

Request Example:

{
    "text": "Long input text that needs summarisation...",
    "max_length": 150,
    "min_length": 50
}

Response Example:

{
    "summary": "Shortened version of the input text."
}

Dockerfile

The application is containerized using Docker, with the setup defined in Dockerfile.

Key Steps:

1. Uses python:3.12-slim as the base image.
2. Copies dependencies from requirements.txt and installs them.
3. Copies application code into the container.
4. Exposes port 8000 and runs uvicorn to serve the FastAPI app.

Required Packages

Dependencies are listed in requirements.txt:

fastapi
uvicorn
transformers
torch
pydantic

Running the Containerized Application

The script run.sh automates building and running the container.

Steps to Run:

1. Build the Docker image:

   bash run.sh

2. The API will be accessible at http://localhost:8000.

Sending a Request

After running the container, you can test the API using curl:

curl -X POST "http://localhost:8000/summarise/" \
     -H "Content-Type: application/json" \
     -d '{"text": "Your long text here", "max_length": 150, "min_length": 50}'

Or use the interactive Swagger UI at http://localhost:8000/docs.