◈ SummarAI

AI-Powered Text Summarization Tool

Paste any article. Get a sharp, accurate summary in seconds.
Powered by state-of-the-art NLP — completely free, runs locally.

🌐 Live Demo →   |   📖 API Docs   |   🐛 Report a Bug

---

📌 Table of Contents

• About the Project
• Features
• Live Demo
• Tech Stack
• Project Structure
• Getting Started
• API Reference
• How It Works
• Running Tests
• Roadmap
• Contributing
• License
• Author

---

📖 About the Project

SummarAI is a full-stack web application that uses Natural Language Processing (NLP) to automatically summarize long-form text — articles, research papers, blog posts, news — into concise, readable summaries.

The project implements two distinct summarization approaches:

• Abstractive — uses the facebook/bart-large-cnn transformer model (fine-tuned on 300,000+ CNN/DailyMail articles) to understand the text and write a fresh summary in its own words, just like a human would.
• Extractive — uses a custom-built TF-IDF scoring algorithm to identify and return the most statistically important sentences from the original text.

This was built as a beginner-to-intermediate learning project to understand NLP pipelines, REST API design, and full-stack web development — from model inference all the way to a production-ready interface.

---

✨ Features

Feature	Description
🧠 Abstractive Summarization	facebook/bart-large-cnn rewrites text in its own words — natural, human-sounding output
✂️ Extractive Summarization	Custom TF-IDF engine selects the most important original sentences
📏 3 Summary Lengths	Short, Medium, Long — control exactly how detailed the output is
📊 Live Statistics	Compression ratio, word counts, and method used — shown after every summary
📋 One-Click Copy	Copy the generated summary to clipboard instantly
🎯 Built-in Sample Article	Demo article included — works immediately with no setup
🌐 Clean REST API	JSON API with 3 endpoints — can be integrated into any app
💯 100% Free & Local	No paid APIs, no subscriptions, no data sent to the cloud

---

🌐 Live Demo

🚀 The app is deployed and accessible here:

https://summari-84uc.onrender.com

⚠️ Note: The app runs on Render's free tier — it may take 30–60 seconds to wake up on first visit.

---

🛠️ Tech Stack

Layer	Technology	Why
Frontend	HTML5, CSS3, Vanilla JavaScript	Lightweight, no framework overhead
Backend	Python 3.10, Flask 3.0	Simple, fast REST API server
AI Model	HuggingFace Transformers + BART	State-of-the-art abstractive summarization
NLP Algorithm	Custom TF-IDF (no libraries)	Fast extractive summarization, zero dependencies
Deployment	Render	Free cloud hosting for Python apps
Testing	pytest	Unit + integration test coverage

---

🗂️ Project Structure

Text-Summarizer-Tool/
│
├── 📂 backend/
│   ├── app.py              # Flask app — all API routes and server config
│   └── summarizer.py       # Core NLP engine (BART abstractive + TF-IDF extractive)
│
├── 📂 frontend/
│   ├── templates/
│   │   └── index.html      # Main UI — single page app
│   └── static/
│       ├── css/
│       │   └── style.css   # Dark editorial theme, fully responsive
│       └── js/
│           └── main.js     # UI logic, API calls, state management
│
├── 📂 tests/
│   └── test_summarizer.py  # 15 unit + integration tests (pytest)
│
├── 📂 docs/
│   └── API.md              # Full API endpoint reference
│
├── requirements.txt        # All Python dependencies with pinned versions
├── .gitignore              # Excludes venv, cache, model files from Git
├── LICENSE                 # MIT License
└── README.md               # This file

---

🚀 Getting Started

Prerequisites

• Python 3.10 or higher — Download
• pip (bundled with Python)
• ~2GB free disk space (for the BART model cache)
• Git — Download

---

Installation

1. Clone the repository

git clone https://github.com/TUSHARTAMRAKAR/Text-Summarizer-Tool.git
cd Text-Summarizer-Tool

2. Create a virtual environment

python -m venv venv

3. Activate the virtual environment

# Windows
venv\Scripts\activate

# macOS / Linux
source venv/bin/activate

You'll see (venv) at the start of your terminal prompt. ✅

4. Install dependencies

pip install -r requirements.txt

⚠️ The first run will download the facebook/bart-large-cnn model (~1.6GB).
This only happens once — it's cached locally afterwards.

5. Start the server

cd backend
python app.py

6. Open in browser

http://localhost:5000

The app is running. 🎉

---

🔌 API Reference

Full documentation → docs/API.md

Base URL: http://localhost:5000

---

POST /api/summarize

Summarize a piece of text.

Request body:

{
  "text": "Paste your article text here...",
  "length": "medium",
  "method": "abstractive"
}

Parameter	Type	Options	Default
text	string	50 – 50,000 characters	required
length	string	short · medium · long	medium
method	string	abstractive · extractive	abstractive

Success response 200:

{
  "success": true,
  "summary": "Artificial intelligence is reshaping industries worldwide...",
  "original_word_count": 500,
  "summary_word_count": 82,
  "compression_ratio": "84%",
  "method_used": "abstractive"
}

Error response 400:

{
  "success": false,
  "error": "Text is too short. Please provide at least 50 characters."
}

---

GET /api/health

Returns server and model status.

{
  "status": "healthy",
  "model_loaded": true,
  "version": "1.0.0"
}

---

GET /api/sample

Returns a built-in demo article for testing.

---

🧠 How It Works

Abstractive Summarization (BART Transformer)

Your Text
    ↓
BART Encoder  →  Reads and builds deep understanding of the text
    ↓
BART Decoder  →  Generates a brand new summary, word by word
    ↓
Fresh Summary (written in the model's own words)

BART (Bidirectional and Auto-Regressive Transformer) was fine-tuned by Facebook AI on the CNN/DailyMail dataset — 300,000+ news articles with human-written summaries. It learns to compress information the same way a journalist would.

---

Extractive Summarization (TF-IDF)

Your Text
    ↓
Split into individual sentences
    ↓
Score each sentence using TF-IDF weights
(words rare in the document but frequent in the sentence = more important)
    ↓
Rank sentences by score
    ↓
Return top N sentences in original order

TF-IDF = Term Frequency × Inverse Document Frequency. It mathematically identifies which words — and therefore which sentences — carry the most unique information in the document. No AI model needed, works fully offline, instant results.

---

🧪 Running Tests

# From project root with venv active
python -m pytest tests/ -v

Expected output:

tests/test_summarizer.py::TestExtractive::test_extractive_returns_dict     PASSED
tests/test_summarizer.py::TestExtractive::test_extractive_success_flag     PASSED
tests/test_summarizer.py::TestExtractive::test_extractive_has_summary      PASSED
tests/test_summarizer.py::TestExtractive::test_extractive_method_label     PASSED
tests/test_summarizer.py::TestExtractive::test_extractive_word_counts      PASSED
tests/test_summarizer.py::TestExtractive::test_extractive_short_length     PASSED
tests/test_summarizer.py::TestWordCounting::test_compression_ratio         PASSED
tests/test_summarizer.py::TestEdgeCases::test_handles_multiline_text       PASSED
tests/test_summarizer.py::TestEdgeCases::test_all_length_options           PASSED
tests/test_summarizer.py::TestAPIEndpoints::test_health_endpoint           PASSED
tests/test_summarizer.py::TestAPIEndpoints::test_summarize_endpoint_valid  PASSED
...

15 passed in 3.42s ✅

---

🗺️ Roadmap

• Abstractive summarization (BART model)
• Extractive summarization (TF-IDF)
• 3 summary length options
• REST API with JSON responses
• Beautiful responsive web UI
• Unit + integration test coverage
• Deployed to Render
• PDF file upload and summarization
• Summarize from a URL (paste a link, get a summary)
• Export summary as .txt / .pdf
• Multi-language support
• Summary history (save and revisit past summaries)
• Browser extension

---

🤝 Contributing

Contributions are welcome and appreciated!

1. Fork the repository
2. Create your feature branch

git checkout -b feature/your-feature-name

3. Commit your changes using Conventional Commits

git commit -m "feat: add PDF upload support"

4. Push to your branch

git push origin feature/your-feature-name

5. Open a Pull Request and describe what you changed and why

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

You're free to use, copy, modify, and distribute this project, as long as the original license is included.

---

👤 Author

Tushar Tamrakar

Built from scratch as a learning project — NLP, Flask, full-stack development, and Git workflow.

---

⭐ Found this useful? Give it a star — it helps others discover the project!

---

_{Made with ❤️ by Tushar Tamrakar}