🧠 LLM Word Predictor

This is a demo. It contains no confidential data/IP.

An interactive web-based application to explore how Language Models (LLMs) predict missing words in sentences.
Users can enter a sentence, select a word to blank out, and let the chosen LLM predict the most likely word(s) to fill in.
The app also visualizes token probabilities in a structured table for deeper insights.

---

🚀 Features

• Model Playground – Select from multiple OpenAI/LLM models via OpenRouter.
• Interactive Sentence Editing – Click any word in the sentence to blank it out (____) and trigger predictions.
• Top Predictions Table – View the top 20 model predictions with probability scores.
• API Key Support – Securely enter your OpenRouter API Key to authenticate.
• Responsive UI – Built with Bootstrap 5 and Bootstrap Icons for a modern, mobile-friendly experience.

---

🗂️ Project Structure

LLM-FILL/
├── index.html    # Main application UI (Bootstrap 5)
├── scripts.js    # Core logic: tokenization, API calls, rendering
└── README.md     # Project documentation

---

⚡ How It Works

flowchart TD
    A[User enters sentence] --> B[Click Process button]
    B --> C[Tokenize sentence into words]
    C --> D[Render interactive sentence as buttons]

    D -->|User clicks token| E[Replace token with ____]
    E --> F[Send modified sentence to OpenRouter API]
    F --> G[Model predicts top tokens + logprobs]
    G --> H[Process response JSON]
    H --> I[Visualize top 20 tokens in probability table]

    I --> J{User interaction}
    J -->|Click another token| E
    J -->|Reset blank| D

Loading

1. Enter a sentence in the input box.
2. Click Process to tokenize the sentence.
3. Click on any token (word) to blank it out (____).
4. The app sends the modified sentence to the selected LLM model via OpenRouter API.
5. The model predicts the most likely token(s) to fill the blank.
6. A probability table shows the top predictions with their likelihoods.

---

🛠️ Setup & Usage

1. Clone the repository

git clone https://github.com/Nitin399-maker/LLM-FILL.git
cd LLM-FILL

2. Open in browser

Open index.html directly in any modern browser — no build step or server required.

3. Configure

• Enter your OpenRouter API Key in the key field.
• Select an LLM model (GPT-4.1 Nano, Mini, or Full).
• Type a sentence, click Process, then click any word to see predictions.

---

🔑 API

This app uses the OpenRouter API to access OpenAI models with logprobs support. You will need a valid OpenRouter API key to use the prediction feature.

---

🧰 Tech Stack

Technology	Purpose
HTML5 / CSS3	Structure & styling
JavaScript (Vanilla)	Application logic
Bootstrap 5	Responsive UI components
Bootstrap Icons	Icon set
OpenRouter API	LLM inference with logprobs

---

⚠️ Disclaimer

This is a demo. It contains no confidential data/IP.