Turn any Shopify store into an AI sales assistant in under 10 seconds — just paste your store URL.
This project ingests a Shopify store, understands its products and business information, and deploys an AI agent that can answer questions, recommend products, manage a cart, and guide users toward checkout — all powered by Retrieval-Augmented Generation (RAG).
This system automatically:
- Ingests a Shopify store using native Shopify endpoints
- Extracts products, prices, availability, and descriptions
- Extracts business information (policies, contact info, about pages, FAQs, etc.)
- Chunks and embeds all content into a vector search index
- Runs an agentic AI layer that:
- Answers product & business questions
- Recommends products with confidence scoring
- Manages a cart (add / remove / view / clear)
- Maintains conversational context per user session
All of this is driven by one input:
👉 the Shopify store URL
Most ecommerce traffic never reaches checkout because customers ask questions before buying:
- “What’s the difference between these two products?”
- “Do you ship to Ireland?”
- “Is this suitable for sensitive skin?”
- “What’s your return policy?”
This project gives Shopify merchants an always-on AI sales rep that understands their store as well as a human would — without manual setup, tagging, or training.
- Paste a Shopify store URL
- Products and pages are auto-discovered
- No Shopify app install required
- Uses
/products.jsonand/collections.json - Structured product embeddings (not scraped HTML)
- Accurate pricing, availability, and descriptions
- Extracts:
- About pages
- Policies (shipping, returns, privacy)
- Contact information
- FAQ-style content
- Supports footer + homepage link discovery for non-standard pages
- All content is chunked and embedded
- Queries are answered using retrieval + reasoning
- Clear separation between:
- Product knowledge
- Business / informational knowledge
The AI can:
- Add items to cart
- Remove items
- Change quantities
- Show cart state
- Remember recently discussed products
Each stage is isolated and testable:
- Product extraction
- Info page processing
- Chunking
- Embedding
- Querying
- Agent routing
flowchart TD
A[Shopify Store URL] --> B[Ingestion Pipeline]
B --> B1[Products]
B --> B2[Collections]
B --> B3[Pages / Policies]
B --> B4[Footer & Homepage Links]
B --> C[Chunking & Embeddings]
C --> D[Vector Store - per business]
D --> E[Agent Layer]
E --> E1[Enquiry]
E --> E2[Cart]
E --> E3[Payments - stubbed]
E --> E4[Chitchat]
-
POST /login
- Authenticates user
- Ensures
business_idexists
-
POST /ingest
- Input:
{ store_url, business_id } - Extracts products + info pages
- Builds RAG index
- Returns
upload_batch_id
- Input:
-
POST /query
- Input:
{ business_id, question } - AI answers using store-specific knowledge
- Input:
- Embedded Shopify chat widget
- Telegram / WhatsApp sales assistant
- Internal sales support bot
- Headless commerce AI layer
- Conversational checkout assistant
- ❌ Not a generic chatbot
- ❌ Not scraping random HTML for products
- ❌ Not a single-prompt RAG demo
- ❌ Not Shopify-theme dependent
This is a store-aware commerce agent, built around structured data and long-term extensibility.
- Python
- FastAPI
- MongoDB
- OpenAI embeddings
- LangGraph / LangChain
- httpx
- Shopify native JSON endpoints
- ✅ Product ingestion working
- ✅ Business page extraction working
- ✅ Chunking + embeddings working
- ✅ Agent routing working
- ✅ Cart management working
- 🚧 Payments temporarily disabled
- 🚧 Social links extraction (planned)
- 🚧 Telegram / WhatsApp deployment (planned)
The long-term goal is to make AI-powered sales assistants a default layer for ecommerce — where merchants don’t “build bots”, they just turn one on.
Paste your Shopify URL.
Your AI sales agent is live in seconds.
