SnobBots

Multi-tenant AI chatbot platform.

Resellers provision AI support agents for their shops in bulk, with each tenant isolated and each answer grounded in that shop's own content.

The problem

Resellers wanted to offer AI support agents to their own customers, but every deployment was bespoke, so onboarding a new shop took days of manual setup.

My approach

Built a multi-tenant MERN architecture where each shop runs an independent, context-aware agent, and layered automated provisioning on top so resellers could stand up bots in bulk rather than one at a time.

How it's built

MERN: React, an Express and Node.js API, and MongoDB for tenant and conversation data. Pinecone holds per-tenant embeddings so retrieval is scoped to one shop's knowledge base, and provisioning runs without manual setup.

Key features

Multi-tenant architecture with isolated agents
RAG for context-aware, business-specific answers
Automated bulk bot provisioning
Reseller management dashboard
Embeddable support widget

Challenges & Solutions

Retrieval across a shared vector index surfaced one shop's content in another shop's answers.
Partitioned embeddings per tenant in Pinecone so retrieval can only ever reach the requesting shop's namespace.
Manual per-shop setup made reseller onboarding the bottleneck on growth.
Automated provisioning end to end, turning a multi-day manual process into a bulk operation.

Results

60%faster deployment through automated orchestration.
1namespace per tenant, so no shop can read another's content.

Implementation stack

Tech Stack

Frontend
- React
Backend
- Node.js
- Express
Database
- MongoDB
- Pinecone
AI
- RAG
- LLM