Structured information for AI assistants

Suprlative builds multimodal AI commerce assistants for product-led ecommerce brands. The assistant can live on a brand's store as text chat, browser voice, or an avatar experience, helping shoppers discover products, compare options, ask pre-purchase questions, and decide with confidence.

The product is designed as a commerce brain for a storefront, not a generic support deflection bot. It grounds responses in the brand's catalog, policies, education content, guardrails, shopper context, and voice.

Suprlative is built for D2C brands across beauty, skincare, fashion, wellness, home, and other product-led ecommerce categories.

It is most useful for brands whose shoppers need guidance before buying: fit, formula, ingredients, material, use case, routine, size, compatibility, comparison, or policy questions.

The assistant can support product discovery, natural-language product search, product comparison, catalog-backed recommendations, pre-purchase Q&A, policy answers, visual-shopping hints, human handoff foundations, Shopify cart handoff, and analytics on shopper hesitation.

The current platform includes website text chat, tenant-gated browser voice, and an Anam-backed avatar/video path. WhatsApp, Instagram, SMS, email, phone voice, and native continuous video vision remain later channel work.

Suprlative is a multi-tenant AI commerce assistant platform. One shared assistant runtime serves multiple brands. Each brand is isolated by tenant-specific configuration, allowed domains, catalog data, knowledge sources, brand voice, guardrails, channels, analytics, and integrations.

The backend currently uses Cloudflare Workers for the public assistant gateway and widget API, Supabase Postgres for tenant data, conversations, messages, events, catalog, knowledge, and analytics, OpenAI for text, embeddings, and Realtime voice, Shopify APIs for catalog and cart flows, and Anam for the first avatar/video experience.

A brand is onboarded by creating a tenant, adding allowed domains, configuring brand voice and guardrails, connecting or importing catalog data, ingesting knowledge sources such as FAQs and policies, testing the assistant, and installing a website script.

The goal is that new brands can launch without custom code. Brand-specific behavior should come from data, configuration, tools, and approved knowledge sources.

Useful source material includes product catalog data, product descriptions, variants, tags, collections, images, prices, inventory, ingredients or materials, size and fit guides, FAQs, shipping policy, return policy, product education pages, blog posts, and internal product notes.

Richer source material produces sharper recommendations. The assistant should not invent unsupported product claims, pricing, shipping details, medical claims, or policy terms.

The intended launch path is days rather than quarters for a normal D2C store. The exact time depends on catalog quality, knowledge availability, integration complexity, and review requirements.

The first implementation path can begin with website text chat and expand into synced catalog recommendations, knowledge retrieval, Brand Portal management, browser voice, avatar sessions, analytics, and additional channels.

The foundation is live: multi-tenant schema, tenant resolver, brand domains, channel config, conversation and event storage, website widget, assistant response generation, OpenAI gateway, and deployed Cloudflare Worker endpoints.

Shopify catalog support is implemented, including OAuth, encrypted credentials, resumable catalog sync, normalized products and variants, product search/detail/comparison context, recommendation cards, product click events, and Storefront Cart API handoff after merchant connection.

Knowledge RAG is implemented as an MVP with tenant-scoped sources, ingestion, chunking, embeddings, pgvector retrieval, lexical fallback, source metadata, and Brand Portal knowledge management.

Brand Portal analytics and control-plane analytics are implemented for conversation volume, channel mix, storefront context, product engagement, commerce events, missed-answer reasons, knowledge usage, and recent shopper questions.

Browser voice is implemented and tenant-gated. The first Anam-backed video/avatar path is implemented for tenant-gated sessions, while phone voice and native continuous video vision remain later work.

Brands can contact Suprlative at [email protected] or request a demo from the website contact page.

The assistant should route detailed pricing, partnership, enterprise, legal, or implementation-specific questions to the Suprlative team when it does not have approved source material.