Recommendations only from the catalog
For each query the plugin searches the real shop catalog. The AI may only recommend from the matches found - no invented products, no hallucinations.
AI-powered pre-sales chat for the Shopware storefront - advises customers from the real catalog, recommends matching products and reduces return rates.
The AI Product Advisor is a pre-sales chat in the storefront that answers customer questions and recommends only real products from the shop. Fast to install, granularly configurable, GDPR-compliant.
For each query the plugin searches the real shop catalog. The AI may only recommend from the matches found - no invented products, no hallucinations.
Freely selectable, with a per-provider API key and model. Anyone who already has an account with OpenAI, Anthropic (Claude), Mistral or Google can use it directly.
A fast pre-call extracts search keywords plus price and property constraints (colour, material, size) - with a local fallback and a 10-second timeout.
Recommended products appear directly in the chat as a quickview - with a variant configurator. The selection resolves the matching variant, with correct cart and product links.
The answer is streamed live via SSE, optionally with a typing effect - no waiting for the full block, the customer sees something happening immediately.
Clicks, cart adds and cart value per recommendation are tracked. Statistics with top terms and products, CSV export in the admin.
Product advice is the strongest conversion lever in physical retail - online it is usually missing entirely. That is exactly what the plugin addresses.
In brick-and-mortar retail the customer asks an employee. Online they stand alone in front of 5,000 items and bounce. The AI Product Advisor closes this gap: it understands the question in natural language, searches the catalog and recommends matching items - increasing conversion and, through better fit, reducing return rates.
Instead of clicking through filters, the customer describes their need - and gets concrete products immediately. Fewer bounces, more purchases.
When the AI recommends the right product, expectation matches the item. This reduces wrong purchases and the expensive returns that follow.
The advisor works around the clock, even at night and on weekends - without additional customer-service staff costs.
Beyond pure advice: features that turn a good conversation into an order - continuously refined from real shop projects.
Search results are weighted by keyword coverage: matches in the product name count most, then category/manufacturer, then properties and description. A product matching more search terms ranks higher.
If the customer names a property (colour, material, size) and no product matches it exactly, no mismatching items are slipped in - the advisor honestly points to close alternatives. An optional toggle enforces exact matches only.
After each answer the AI suggests context-aware follow-up questions as tappable chips (e.g. "Patterned" | "Striped" | "Rather plain"). A tap sends the reply directly - the customer doesn't have to type.
Optional "add to cart" button directly on the recommendation card for products without variants. Variants still go through the detail modal with the configurator.
Optional thumbs feedback under each answer. Ratings feed anonymously into the statistics as a "satisfaction" metric - without any extra database migration.
Recommended products are stored with the answer in the chat history and restored when the chat is reopened, minimised or after navigation - including click, detail modal and cart.
From provider choice via persona to position - all through the admin UI under Marketing → AI Product Advisor. No code changes required.
Active switch, AI provider, API key, model, display area, smart search on/off and a separate extraction model for the pre-search.
Advisor name and avatar, greeting text, quick-start suggestions, consent text and a freely definable persona (tonality, role).
Eight placement positions, offsets, accent colour, proactive opening after a delay and an attention pulse on the button.
Logging on/off, retention in days, analysis dashboard and CSV export - all in the admin under Marketing → AI Product Advisor.
Position, persona, quick-start suggestions and accent colour can be configured separately per sales channel - B2C different from B2B.
Optionally the chat opens automatically after a configurable delay - or pulses subtly to draw attention without being intrusive.
AI advice and data protection are not mutually exclusive. The plugin redacts personal data, offers a consent gate and cleans up logs automatically.
Before sending to the AI provider, e-mail addresses and phone numbers are automatically removed from the message - the provider never sees personal contact data.
If desired, the customer must actively consent before the chat starts - with freely configurable consent text for GDPR compliance.
Chat logs are kept only as long as configured. A daily scheduled task automatically deletes older entries.
Storefront component, provider registry, SSE streaming, anti-hallucination and automatic log cleanup - without a core patch.
Dedicated storefront component staw-advisor as a floating widget, included via base.html.twig. Works with default and custom themes.
Provider registry with interchangeable providers (OpenAI, Anthropic, Mistral, Gemini). Answer streaming via Server-Sent Events, separate extraction model for the pre-search.
The ProductContextBuilder passes only actually found catalog matches to the AI. The prompt restricts recommendations to these matches.
A built-in RateLimiter protects against abuse and uncontrolled API costs. Limitable per session/IP.
Dedicated tables for logs and conversion events. A daily scheduled task CleanupLogsTask removes expired entries automatically.
Two routes, the same feature set - pick whichever suits your setup.
All prices net, plus VAT.
You buy the licence directly from us and get an invoice from us. Install by upload in the admin without SSH, updates through our licensing platform, support straight from the developer.
The plugin will be listed in the official Shopware Community Store. Installation and automatic updates then run through the plugin manager, billing through your Shopware account. We will link it here as soon as it is live.