SmartSiteChat

Overview

SmartSiteChat is a SaaS platform that lets any business add an AI assistant to its website and to WhatsApp in minutes. The assistant trains itself on the customer's own site content through an automated crawler, then answers visitor questions accurately, replies to WhatsApp messages automatically based on each query, captures leads, and surfaces analytics. It ships as a drop-in embeddable widget, a WhatsApp integration, and a self-serve dashboard for configuration and monitoring.

The challenge

Most businesses cannot staff live chat around the clock, and generic chatbots know nothing about the specific company they are dropped onto. Building and training a useful bot is normally a manual, technical effort, and ungrounded models hallucinate answers that erode trust. The product needed to remove that setup work entirely while keeping answers accurate and on-brand.

What we built

• A modular Express backend organized into clean domain modules: auth, workspace, site, chatbot, scraper, training, conversation, lead, analytics, billing, onboarding, webhooks, and the public widget API.
• An automated content-ingestion pipeline with a pluggable scraper factory supporting both Firecrawl and a Playwright-based crawler, feeding a training service that builds each bot's knowledge base.
• OpenAI-backed answer generation grounded in the crawled content, with tracking of unanswered questions so knowledge gaps can be identified and closed.
• A multi-tenant, scale-conscious design: Redis caching with automatic in-memory fallback, MongoDB indexing, JWT access/refresh auth, rate limiting and request slow-down, and Helmet hardening.
• A WhatsApp channel that receives customer messages and generates automatic AI replies from the same grounded knowledge base, so businesses answer on the channel customers already use.
• An embeddable JavaScript widget that installs on any external site, a Nuxt 3 dashboard for config, conversations, leads and analytics, and a separate React marketing site.

Results

SmartSiteChat lets non-technical site owners stand up a knowledgeable, on-brand chatbot quickly by crawling their site rather than training it by hand. Visitors get round-the-clock answers grounded in real company content, and conversations convert into captured leads instead of lost inquiries. The caching, indexing and rate-limiting architecture keeps responses fast under concurrent traffic across many tenant sites, while unanswered-question tracking gives operators a feedback loop to keep improving answer quality.