---
title: GuardPII GLiNER2 Demo
emoji: 🔎
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false
---

# GuardPII GLiNER2 Demo

Gradio demo for PII extraction with GLiNER2 plus the high-precision rulebase used by the project prediction pipeline.

## Inputs

- Raw text
- `.docx` files via `python-docx`
- `.pdf` files via LlamaParse by default, with optional local parsers: `pdfplumber`, `pymupdf`, `pypdf`

## Hugging Face Space Setup

Upload the contents of this `DEMO` folder to a Hugging Face Space.

Set Space secrets or variables as needed:

- `LLAMA_CLOUD_API_KEY`: required when using the default `llamaparse` PDF parser
- `HF_TOKEN`: optional, needed only if the selected GLiNER2 model is private
- `MODEL_ID`: optional default model ID, defaults to `AITeamUIT/gliner2-multi-v1-3e-20260514`
- `PDF_BACKEND`: optional default PDF parser, defaults to `llamaparse`
- `THRESHOLD`: optional default confidence threshold, defaults to `0.5`

The app caches loaded models, so the first run may be slow while the model downloads and initializes.