---
title: MLP Safety Classifier
emoji: 🛡️
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: "4.44.0"
app_file: app.py
pinned: false
license: mit
---


# 🛡️ Aegis Safety Classifier

This Space hosts a text classifier trained on the [NVIDIA Aegis 2.0 dataset](https://huggingface.co/datasets/nvidia/Aegis-AI-Content-Safety-Dataset-2.0).  
It predicts whether a piece of text is **safe** or **unsafe**.

The model is a simple **TF-IDF + MLP** pipeline implemented in scikit-learn.

---

## Features

- Paste any text into the textbox and get a prediction.
- Adjustable **threshold** for the probability of being classified as "safe".
- JSON output with:
  - `prediction`: `1` = safe, `0` = unsafe
  - `probabilities`: [p(unsafe), p(safe)] if available

---

## How to Use

### 
- Open this Space.
- Enter some text in the input field.
- Adjust the threshold (default = 0.5).
- Press **Submit** to get results.