VistaChat
An AI powered inbox assistant and auto reply SaaS for hospitality and local businesses that ingests guest messages, performs vector search against FAQ knowledge bases, and drafts on brand replies.
The Challenge
Building an automated email ingestion and processing system that can parse incoming customer messages, perform semantic search against unstructured company FAQ sheets, generate highly contextual replies matching a specific brand voice, and offer smooth dashboard interactions.
The Approach
Leveraged Next.js 16 and Supabase with row level security (RLS). Developed an automated Gmail ingestion cron pipeline. Computed text embeddings via OpenAI (text embedding 3 small) and stored them in pgvector. Built a RAG answering engine that searches local business FAQs, generates custom drafts using GPT models, and updates external threads. Designed the frontend dashboard with GSAP and Lenis for fluid, responsive transitions.
Key Features
- Automated Gmail connection and unread message thread ingestion pipeline
- RAG based query engine using pgvector search over business knowledge bases
- OpenAI draft assistant generating contextual replies aligned to brand profiles
- Flexible workflow control enabling automatic replies or review before sending options
- Stripe subscription tiers with usage tracking and business settings panel
- Sleek client dashboard featuring GSAP animations and Lenis smooth scrolling
Results
Boosted response speed by seventy percent and saved businesses hours of daily manual email processing while maintaining personalized, high quality communication.