Multi-agent AI shopping assistant for Shopware 6, built on NVIDIA's retail blueprint
Fork of NVIDIA AI Blueprints: Retail Shopping Assistant with Shopware 6 integration, German/English bilingual support, and MEMOTECH branding.
| Feature | NVIDIA Blueprint | MEMOTECH Fork |
|---|---|---|
| LLM | Llama 3.1 70B Instruct | Llama 4 Maverick (MoE, 400B params, 17B active) |
| Language | English only | German/English bilingual |
| E-Commerce | Static CSV catalog | Shopware 6 Store API integration |
| Planner | Basic routing | Robust routing for verbose MoE models |
| Branding | NVIDIA generic | MEMOTECH + NVIDIA |
| Widget | Standalone UI | Embeddable Shopware storefront widget |
| Sync | Manual CSV import | Automated Shopware product sync script |
The Retail Shopping Assistant is an AI-powered blueprint that provides a comprehensive interface for an intelligent retail shopping advisor. Built with LangGraph for agent orchestration, it features multi-agent architecture, real-time streaming responses, image-based search, and intelligent shopping cart management.
- 🤖 Intelligent Product Search: Find products using natural language or images
- 🛒 Smart Cart Management: Add, remove, and manage shopping cart items
- 🖼️ Visual Search: Upload images to find similar products
- 💬 Conversational AI: Natural language interactions
- 🔒 Content Safety: Built-in moderation and safety checks
- ⚡ Real-time Streaming: Live response generation
- 📱 Responsive UI: Modern, mobile-friendly interface
Note: The diagram above shows the original NVIDIA Blueprint with Llama 3.1 70B. This fork uses Llama 4 Maverick (MoE, 400B params, 17B active) via NVIDIA Cloud API.
The application follows a microservices architecture with specialized agents for different tasks:
- Chain Server: Main API with LangGraph orchestration
- Catalog Retriever: Product search and recommendations
- Memory Retriever: User context and cart management
- Guardrails: Content safety and moderation
- UI: React-based frontend interface
For detailed architecture information, see Architecture Overview.
- Docker: Version 20.10+ with Docker Compose plugin
- NVIDIA NGC Account: For API access (Get API Key)
- Hardware: 4x H100 GPUs (preferred) or 4x A100 GPUs (minimum) for local deployment, or cloud access
-
Clone the repository:
git clone https://github.com/NVIDIA-AI-Blueprints/retail-shopping-assistant.git cd retail-shopping-assistant -
Authenticate with NVIDIA Container Registry:
docker login nvcr.io
Use
$oauthtokenas the username and your NGC API key as the password. -
Set up environment:
export NGC_API_KEY=your_nvapi_key_here export LLM_API_KEY=$NGC_API_KEY export EMBED_API_KEY=$NGC_API_KEY export RAIL_API_KEY=$NGC_API_KEY export LOCAL_NIM_CACHE=~/.cache/nim mkdir -p "$LOCAL_NIM_CACHE" chmod a+w "$LOCAL_NIM_CACHE"
-
Launch the application:
Option A: Local Deployment:
# Start local NIMs (requires 4x H100 GPUs) docker compose -f docker-compose-nim-local.yaml up -d # Build and launch the application docker compose -f docker-compose.yaml up -d --build
Option B: Cloud Deployment (no local GPUs required):
# Configure to use NVIDIA API Catalog endpoints export CONFIG_OVERRIDE=config-build.yaml # Build and launch the application docker compose -f docker-compose.yaml up -d --build
-
Access the application: Open your browser to
http://localhost:3000 -
Stop the containers:
Option A: Local Deployment:
docker compose -f docker-compose.yaml -f docker-compose-nim-local.yaml down
Option B: Cloud Deployment:
docker compose -f docker-compose.yaml down
For detailed installation instructions, see Deployment Guide.
For a streamlined cloud deployment experience, you can deploy the Retail Shopping Assistant on NVIDIA Brev using GPU Environment Templates (Launchables):
NVIDIA Brev Deployment Guide - Complete step-by-step instructions for deploying on Brev
- One-Click Deployment: Pre-configured GPU environments with automatic setup
- Managed Infrastructure: No need to manage servers or GPU clusters
- Secure Access: Built-in secure tunneling for web interface access
- Flexible Resources: Choose from H100, A100, and other GPU configurations
- Cost-Effective: Pay only for actual usage time
The Brev deployment guide walks you through the entire process from creating a Launchable to accessing your fully functional retail shopping assistant.
- User Guide: How to use the application
- API Documentation: Complete API reference
- Deployment Guide: Installation and setup instructions
- Documentation Hub: Complete documentation index
We welcome contributions! Please see our Contributing Guide for details on:
- Development setup and environment configuration
- Coding standards and best practices
- Testing guidelines and examples
- Pull request process and code review guidelines
- GitHub Issues: Report bugs and feature requests
- Documentation: Comprehensive guides and references
- NVIDIA AI Blueprints: Collection of AI application blueprints
- NVIDIA NIM: Containerized AI models
- NVIDIA NGC: AI platform and container registry
- LangGraph: Agent orchestration framework
- FastAPI: Modern Python web framework
- React: JavaScript library for building user interfaces
- Milvus: Vector database for similarity search
- NVIDIA Retrieval QA: Embedding model for semantic search
- NV-CLIP: Visual understanding model
- Llama 4 Maverick: Large language model (MoE, 17B active params)
- Shopware 6: Open-source e-commerce platform
GOVERNING TERMS: Use of the blueprint software and materials and NIM containers are governed by the NVIDIA Software License Agreement and Product-specific Terms for AI products; and the use of models is governed by the NVIDIA Community Model License.
ADDITIONAL INFORMATION: Llama 3.1 Community License Agreement for Llama 3.1 NemoGuard 8B - Content Safety and Llama 3.1 NemoGuard 8B - Topic Control models, built with Llama; Llama 4 Community License Agreement for Llama 4 Maverick 17B-128E Instruct; (ii) MIT license for NV-EmbedQA-E5-v5.
This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use, found in License-3rd-party.txt.
Use of the product catalog data in the retail shopping assistant is governed by the terms of the NVIDIA Data License for Retail Shopping Assistant (15Aug2025).