Client Orbit Support AI

AI-powered production-style, zero-budget Support Assistant with automated replies and lead extraction.

Live Demo

• Public demo: https://client-orbit-kohl.vercel.app/

Overview

This application provides:

• Policy-aware support responses with escalation behavior.
• Lead capture from user conversations (name, email, need, urgency).
• API validation and request rate limiting.
• Persistent storage for conversations and leads.

Tech Stack

• Next.js (App Router, TypeScript)
• Google Gemini API
• Supabase (Postgres)
• Upstash Redis (rate limiting)

Project Structure

• app/api/chat/route.ts: Chat endpoint and request orchestration
• app/api/health/route.ts: Health check endpoint
• app/api/leads/route.ts: Lead listing and manual lead insert endpoint
• lib/ai/provider.ts: Gemini integration and fallback logic
• lib/ai/prompts.ts: System and extraction prompts
• lib/db/queries.ts: Persistence layer for leads and conversations
• lib/rateLimit/upstash.ts: Rate limiter implementation
• lib/utils/validation.ts: Input validation schemas

Getting Started

1. Install dependencies: npm install
2. Configure environment variables: copy .env.example .env.local
3. Start the development server: npm run dev
4. Run checks: npm run lint npm run test

Environment Variables

Use these variables in local and deployment environments:

• GEMINI_API_KEY
• GEMINI_MODEL
• SUPABASE_URL
• SUPABASE_SECRET_API_KEY (preferred) or SUPABASE_SERVICE_ROLE_KEY (fallback)
• UPSTASH_REDIS_REST_URL
• UPSTASH_REDIS_REST_TOKEN
• ADMIN_TOKEN

API Endpoints

• GET /api/health
• GET /api/chat (usage helper)
• POST /api/chat
• GET /api/leads (requires x-admin-token)
• POST /api/leads

Deployment

Deploy to Vercel and set all required environment variables in project settings.

Test Case Screenshots

1. Design Overview

2. Simple Hi Test

3. Hardcoded Test

4. Confusing Realistic Prompt

5. Lead Extraction Test

6. Ambiguous Prompt