Text Summarizer

A small full-stack app that accepts unstructured text and returns a structured summary using an LLM API.

Built as part of the AI Developer Intern take-home assignment.

Live url : https://assignment-summarizer.netlify.app/

Tech Stack

• Frontend: React + Vite
• Backend: Node.js + Express
• LLM: Groq API (llama3-8b-8192)
• Other: dotenv, cors

---

Setup & Running Locally

1. Clone the repository

git clone <your-repo-url>
cd assignment-summarizer

2. Setup the backend

cd server
npm install
cp .env.example .env

Open .env and add your Groq API key:

GROQ_API_KEY=your_groq_api_key_here

Start the backend:

npm run dev

Server runs on http://localhost:3000

3. Setup the frontend

Open a new terminal:

cd client
npm install
npm run dev

Frontend runs on http://localhost:5173

---

Which LLM API and Why

I used Groq with the llama3-8b-8192 model for two reasons:

1. It has a generous free tier with high rate limits — reliable for demos without hitting quota issues.
2. Its API is OpenAI-compatible, which keeps the integration simple and easy to swap out if needed.

---

Prompt Design

The prompt instructs the model to act as a strict JSON extractor, not a conversational assistant. It defines the exact output shape, constrains the sentiment to three allowed values, and explicitly forbids markdown or extra keys.

This reduces the chance of malformed output without needing complex post-processing. A lower temperature (0.3) further keeps the output consistent.

---

Example Output

Input:

NASA's Perseverance rover has been exploring Mars since February 2021. 
It has collected rock samples, studied the Martian atmosphere, and 
helped test oxygen production on the planet. Scientists are excited 
about the potential clues these samples may offer about ancient life on Mars.

Output:

{
  "summary": "NASA's Perseverance rover has been actively exploring Mars since 2021, collecting samples and conducting experiments that could reveal evidence of ancient life.",
  "keyPoints": [
    "Perseverance has collected rock samples and studied the Martian atmosphere",
    "The rover successfully tested oxygen production on Mars",
    "Scientists believe the samples may contain clues about ancient Martian life"
  ],
  "sentiment": "positive"
}

Known Trade-offs

• Used a simple fixed output schema instead of supporting user-defined fields — sufficient for the assignment scope.
• Kept a single backend route rather than splitting into service layers — easier to read and explain.
• No authentication — not needed for a local assignment demo.
• No test coverage — kept within the 1–2 hour time budget.

---

What I Would Improve With More Time

• Add file upload support so users can summarize .txt files directly
• Stream the LLM response for faster perceived performance
• Add schema validation on the LLM output using zod
• Support batch processing of multiple texts