Configure, embed, converse: anAI advisor widget forany website.
- Type
- Web app, AI SaaS
- Category
- AI & LLM
- Stack
- React / Next.js, FastAPI, RAG with hosted & self-hosted LLMs
- Live site
- custombot.store
The challenge
Businesses want an assistant that answers from their own content, but building one means wiring up document indexing, language models, voice and per-customer data isolation. The goal was to make that a single script tag.
CustomBot (advisorAI) is a multi-tenant SaaS that puts an AI advisor on any website. The site owner sets the brand, personality, FAQs and opening questions in a console, with a live preview beside the settings and no deploy step.
What it does
- 01
Console with live preview
Brand, personality, FAQs and opening questions are edited next to a live preview, with no deploy step.
- 02
Answers from your documents
Documents of up to 20 MB each are indexed so the advisor answers from them, using a RAG architecture.
- 03
One-line install
The whole integration is one script tag containing a widget id.
- 04
Text and live voice
Visitors can type, or start a live voice call, inside the same conversation.
- 05
Multi-tenant by design
Each widget has its own identity mode and allowed origins, and tenants' data is kept isolated.
Built with
- React / Next.js
- FastAPI
- RAG with hosted & self-hosted LLMs
- FAISS / Qdrant / Weaviate
- Supabase (PostgreSQL)
- Azure Blob Storage
- Docker
- GitHub Actions
At a glance
- 1
- Script tag to embed the widget
- 20 MB
- Per uploaded document
- Text + voice
- In the same conversation
- Isolated
- Per-tenant data and allowed origins
