Remix

Make your own version of AI Business Assistant

Get an exact, working copy in your account, then change anything with chat.

Original by Banny Singh

What would you change?

Start free with 40 credits. No card needed.

Start fresh from this prompt

Builds a brand new app from the first prompt only, so later changes to the original are not included.

Role: You are an expert Full-Stack Developer and AI Architect. Your task is to build a Multi-Tenant "AI Business Assistant" SaaS platform. The platform allows clients to log in, input their business details (onboarding), and deploy an AI assistant that handles customer support, lead generation, and automation.

Tech Stack: Next.js (App Router), Tailwind CSS, PostgreSQL (Supabase), Vector DB (pgvector for RAG), OpenAI API, and Twilio (for WhatsApp).

Here are the 12 core modules to build. Please execute this in phases.

---

PHASE 1: Auth, Database & Multi-Tenancy (Layer 1 & 3)

1. Client Login & Auth (Part 2): Implement Supabase/Clerk authentication (Email/Password + Google OAuth). Each user is a "Tenant".
2. Database (Part 7): Create tables with Row Level Security (RLS). Tables: tenants (business info), knowledge_base (docs/chunks), chats, leads, workflows. Ensure data isolation (Tenant A cannot see Tenant B's data).
3. Security & Compliance (Part 11): Add basic PII masking on chat inputs. Setup audit logs for admin actions.

PHASE 2: Onboarding & Knowledge Base (Layer 5 & 3)

4. Onboarding Wizard (Part 10): Create a multi-step form for the client to fill out:
   · Step 1: Business Name, Industry, Language (Hindi/English/Hinglish).
   · Step 2: Upload Docs (PDF/Text), Paste FAQs, Add Pricing.
   · Step 3: Set AI Persona (Formal/Friendly) & Fallback message ("I don't know").
   · Step 4: Connect Channels (WhatsApp API key / Website Widget).
5. Knowledge Base + RAG (Part 6): Write a server-side function to chunk uploaded documents, generate embeddings via OpenAI, and store them in pgvector. The AI must query this before answering.

PHASE 3: The Brain & Chat Interface (Layer 2 & 1)

6. LLM / AI Brain (Part 4): Integrate OpenAI API.
7. Orchestrator (Part 5): Create an intent classifier. If the user says "book an appointment", route to a booking tool. If they ask a question, route to RAG.
8. User Interface (Part 1): Build a floating chat widget for websites and a WhatsApp webhook endpoint. The UI must show typing indicators and support bilingual responses.

PHASE 4: Actions, Dashboard & Analytics (Layer 4 & 5)

9. Tools / Integrations (Part 8): Create backend functions to push leads to a CRM (via webhook/API) and send emails using Resend/SendGrid.
10. Automation Engine (Part 9): Build a simple workflow builder UI. Example: "IF intent = pricing, THEN send pricing PDF AND create lead in DB."
11. Admin Panel / Dashboard (Part 3): A UI where the client can see all active chats, view captured leads, and edit the Knowledge Base.
12. Analytics & Human Handover (Part 12): Dashboard charts for Daily Chats, Leads, and Top Questions. Add a "Talk to Human" button that pauses the AI and sends an alert to the admin.

Instructions for AI Builder:

· Start by generating the Database Schema first.
· Build the Onboarding UI and Admin Dashboard.
· Finally, integrate the AI logic and chat widget.
· Do not use mock data for Auth; use real Supabase/Clerk integration.

Start fresh from this prompt

Builds a brand new app from the first prompt only, so later changes to the original are not included.