Code Review · AI SaaS

Your REXAK,
reviewed.

An AI chatbot builder — train it on your content, embed it anywhere, bill for it. Here's what works, what to improve, and the path to public-safe.
10 edge functions14 tables~12K linesRAG + StripeRLS secured
Scroll
01 — What it is

A real AI chatbot SaaS.

Create a bot, train it on your own content (files, web pages, text, Q&A), embed a branded widget, watch analytics, and charge with Stripe. Same space as Chatbase & SiteGPT.

Train (RAG)

  • File upload
  • Web scraping
  • Text & Q&A
  • Embeddings + search

Deploy

  • Embeddable widget
  • Theme customizer
  • Live demo page

Measure

  • Analytics events
  • Dashboard
  • Conversations

Monetise

  • Stripe checkout
  • Customer portal
  • Free / Pro / Business
02 — Real-time usability

Use it now? Yes — with guardrails.

The full loop works: sign up → train → embed → answer → analytics → upgrade. Before real public traffic, add abuse protection.

✅ Ready today

  • Build & train a chatbot on your own content
  • Embed a themeable widget on any website
  • Analytics + full admin panel
  • Real Stripe subscriptions (Free / Pro / Business)
  • Great for habash.design — use it, or sell "AI chatbot setup" to clients

✕ Not ready yet

  • Heavy public traffic without rate limiting (cost risk)
  • Enterprise (no SSO, no audit logs)
  • Human handoff / live agent fallback
  • Multi-channel (WhatsApp / Slack / Messenger)
  • Automated tests before scaling
03 — Production benchmark

How far is it, really?

Scored from the real code. Click a card for detail.

0/100 · v1 readiness

The strongest of the code projects — a complete, monetisable AI SaaS with a real RAG pipeline, embeddable widget, analytics, admin and Stripe billing. About 74% shippable. The remaining work is hardening for public traffic (rate limiting & usage caps), not building missing features.

04 — Roadmap

What to work on later.

Recommendations only — nothing changed. Prioritised for a public-safe, sellable product.

P0 · Make it public-safe (before real traffic)

Protect the money & the model

  • Rate limiting + usage metering on ai-chat / rag-search / generate-embeddings (per chatbot, IP, plan)
  • Origin / domain allowlist per chatbot for the widget
  • Server-side plan enforcement (message & training caps)
P1 · Product depth

Make it stickier

  • Human handoff / live-chat fallback + email capture
  • Conversation review UI (read, rate, turn into training)
  • Second LLM provider as fallback + model choice per plan
P2 · Growth

Reach your market

  • WhatsApp / Slack / Messenger / Instagram channels
  • Multi-language answers + RTL Arabic widget (your differentiator)
  • Lead capture in chat → CRM
← Habash Design