Getting Started

Deskly is a Next.js app for embeddable, multi-tenant AI chat widgets: each account runs unlimited bots, each bot has its own public embed key, domain allowlist, and document-scoped knowledge base. Follow the steps below until you have a bot embedded and answering questions on a real page.

Quick Start

1. Clone and install

   git clone https://github.com/boilerlykit/deskly.git
   cd deskly
   pnpm install
   

2. Set up environment variables

   • Copy .env.local.example to .env.local
   • Follow the Configuration Guide to fill in Supabase, AI provider, and (optionally) Resend/CRM keys

3. Set up the database

   • Run the SQL schema from src/db/schema.sql in your Supabase SQL editor
   • Create a storage bucket named uploads in Supabase Storage

4. Start the dev server

   pnpm dev
   

5. Run the built-in setup checklist (recommended)

   • Visit /dashboard/setup to validate your environment variables and database schema

6. Create your first bot

   • Sign in, open /dashboard/bots, create a bot, and add at least one domain to its allowed-domains list (a bot with no allowed domains never renders — secure by default)
   • Upload a document so the bot has something to answer from

7. Embed it

   <script src="https://your-deploy-url.com/widget.js" data-bot-key="pk_..." async></script>
   

   Drop this before </body> on any page. It injects a sandboxed iframe bubble in the bottom-right corner — see Embeddable Widget for how the isolation and resize behavior work.

What ships

• Multi-tenant bots — a bots table, not a single hard-coded assistant; see Architecture
• Embeddable iframe widget — one script tag, CSS-isolated from the host page
• Per-bot RAG — uploads scoped to a bot, not the whole account; see RAG Pipeline
• Abuse protection — visitor, hourly, and daily rate limits plus a real owner credit check; see Abuse Protection
• Lead capture — email/phone extraction and owner email alerts; see Lead Capture
• Same foundation as the base template — Supabase Auth, Postgres RLS, Stripe credits, unchanged

Documentation

Next Steps

1. Configure your environment variables
2. Create a bot in /dashboard/bots and upload a document
3. Embed the widget on a test page
4. Deploy to production