README.md

---
title: Lightweight LLM with MCP Integration
emoji: 🧠
colorFrom: green
colorTo: blue
sdk: docker
app_port: 7860
pinned: false
license: mit
tags:
  - llm
  - mcp-integrated
  - inference
  - fastapi
  - prestashop
short_description: Lightweight LLM with integrated PrestaShop MCP tools
---

# Lightweight LLM with Integrated MCP

A lightweight Large Language Model inference endpoint with **integrated** MCP (Model Context Protocol) server for enhanced PrestaShop capabilities. Everything runs in a single container for maximum efficiency and simplicity.

## 🚀 Features

- **Lightweight Architecture**: Single container with minimal dependencies
- **Hugging Face Compatible**: Standard HF API format support
- **Integrated MCP**: Built-in PrestaShop MCP server (no external dependencies)
- **Fast Inference**: Optimized for quick response times with local MCP calls
- **PrestaShop Ready**: Direct integration with PrestaShop APIs
- **Docker Single Container**: Everything in one container for easy deployment

## 📡 API Endpoints

### Main Inference Endpoint: `POST /`
**Hugging Face compatible format**

Request:
```json
{
  "inputs": "What are the features of product 123?",
  "parameters": {
    "max_new_tokens": 150,
    "temperature": 0.7,
    "prestashop_api_url": "https://your-shop.com"
  }
}
```

Response:
```json
[
  {
    "generated_text": "Product 123 is a premium item with advanced features..."
  }
]
```

### Health Check: `GET /health`
Service health and configuration status

### Model Info: `GET /info`
Model and configuration information

## 🔧 Configuration

### Environment Variables
- `PRESTASHOP_API_URL`: Default PrestaShop API URL 
- `MODEL_NAME`: Hugging Face model to use (default: lightweight model)
- `MAX_NEW_TOKENS`: Default max tokens (default: 256)

### Integrated MCP
The LLM includes a built-in MCP server with PrestaShop tools:
- **Direct Integration**: No network overhead, fastest possible responses
- **Auto-Detection**: Automatically detects when to use PrestaShop tools
- **Fallback**: Works without PrestaShop (direct responses)
- **Single Container**: Everything runs together for simplicity

## 🛠️ Usage

### Direct Inference
```bash
curl -X POST https://your-inference-endpoint.hf.space/ \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": "Hello, how are you?",
    "parameters": {"max_new_tokens": 100}
  }'
```

### Integrated MCP Inference
```bash
curl -X POST https://your-inference-endpoint.hf.space/ \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": "Tell me about product 123",
    "parameters": {
      "max_new_tokens": 200,
      "prestashop_api_url": "https://your-shop.com"
    }
  }'
```

## 🏗️ Architecture

```
HTTP Request → Lightweight LLM → Integrated MCP Server → Enhanced Response
```

1. **Request Processing**: Analyze user input for PrestaShop tool requirements
2. **Tool Decision**: Determine if PrestaShop tools are needed
3. **MCP Integration**: Call integrated MCP server directly (no network overhead)
4. **Response Generation**: Generate final response with enriched PrestaShop data
5. **HF Format**: Return in standard Hugging Face format

## 🚀 Deployment

### Hugging Face Spaces
1. Create new Space with Docker SDK
2. Upload all files from `inference_point/` directory
3. Set app_port to 7860 (already configured in README.md header)
4. Configure PRESTASHOP_API_URL environment variable

### Hugging Face Inference Endpoints
1. Use this Docker setup for custom inference endpoints
2. Configure scaling and instance types as needed
3. Set environment variables for PrestaShop integration

## 📊 Performance

- **Startup Time**: < 30 seconds
- **Inference Speed**: 1-3 seconds for direct responses
- **MCP-Enhanced**: 2-5 seconds (local integration, no network overhead)
- **Memory Usage**: Optimized for single container deployment
- **Concurrent Requests**: Supports multiple simultaneous requests

## 🔗 Integrated MCP Tools

This service includes built-in PrestaShop MCP tools:
- **get_product_details**: Get complete product information
- **get_product_features**: Get product specifications and features  
- **get_product_images**: Get product images and media
- **search_products**: Search the product catalog

**Keywords that trigger MCP usage:**
- product, price, stock, inventory
- order, customer, database
- search, find, get, fetch, check
- prestashop, shop, cart, purchase

## 📝 License

MIT License - Feel free to use and modify for your projects.

---

⚡ **Lightweight & Fast**: Optimized for production inference workloads.