# Algeria Telecom 2026 — AI Agent Demo
A React web app showcasing the DPO fine-tuned Llama-3-8B model for Algeria Telecom customer support.
## Features
- 💬 Live chat with the HuggingFace-hosted model
- 📊 Training metrics dashboard (loss curve, ReAct compliance, language distribution)
- ℹ️ Model info page
- 🌍 Multilingual: Arabic, French, English
## Setup
### 1. Install dependencies
```bash
npm install
```
### 2. Add your HuggingFace token
Edit `.env` and replace `your_hf_token_here` with your HF read token:
```
REACT_APP_HF_TOKEN=hf_xxxxxxxxxxxxxxxxxxxx
```
### 3. Run locally
```bash
npm start
```
Opens at
localhost
### 4. Build for production
```bash
npm run build
```
## Deploy (free options)
- **Vercel**: `npm i -g vercel && vercel` (add HF_TOKEN as env variable in dashboard)
- **Netlify**: drag the `build/` folder to netlify.com
- **GitHub Pages**: push to GitHub, enable Pages
## Model
- Base: `meta-llama/Meta-Llama-3-8B-Instruct`
- Fine-tuned: `RaaNiaaa/algeria-telecom-dpo-merged`
- Method: DPO with LoRA (r=16)
- Best checkpoint: Step 10 (loss: 0.396)