[OSS/free chatbot kit · v1.0]

The free AI chatbot for business websites.

One HTML file. Streaming replies. Server-held API key. MIT-licensed. Drop it on any site — WordPress, Webflow, Next.js, plain Apache — and customize the system prompt with your prices, hours, and policies. No SaaS subscription, no usage cap, no branding.

Download v1.0.0 (zip · ~70 KB)Try the live demo View on GitHub

Want it installed and customized for you? See chatbot install packages — flat fee, 2-week delivery.

[00/privacy]

What the model sees — and what we prompt it to do.

The API sends your conversation transcript plus a fixed system prompt to Groq so it can answer. That is inference only — not a separate “training” product on publishd.app. The help bot's system prompt tells the model to treat visitor messages as confidential, not to ask for passwords or API keys, and not to invent Publishd prices or client identities. If you fork the kit, put your own rules in api/system-prompt.ts (or equivalent) and match them to your privacy policy.

Full policy: Privacy Policy.

[01/what's in the box]

A real, production widget — not a teaser.

One HTML file

Drop chat-demo.html on any host. No bundler, no React, no framework. The iPhone bezel is embedded as base64 — true single-file delivery.

Streaming, server-held key

Tokens stream straight from Groq through your /api/chat proxy. The HTML file contains zero secrets — your API key never touches the browser.

MIT, fork it, own it

Public GitHub repo. No usage cap, no telemetry, no upsell modal. Fork, rename, restyle, ship under your brand.

[02/install in 4 steps]

From zip to live chat in five minutes.

01
Download the v1.0 kit
Grab the zip below. It contains the widget, a Next.js-ready API route, system-prompt template, .env.example, and the MIT license — about 70 KB total.
02
Drop it into your project
Copy chat-demo.html into public/. Copy api/chat.ts → app/api/chat/route.ts and api/system-prompt.ts next to it. Edit the system prompt with your business facts (prices, hours, links).
03
Add a free Groq key
Create one at console.groq.com and put it in .env.local as GROQ_API_KEY. Do not commit it. Same key works in production once you set it on your host.
04
Run npm run dev
Open /chat-demo.html. Streaming replies, no key prompts. When you ship to production, the widget auto-uses same-origin /api/chat — no CORS wiring needed.

[03/embed it anywhere]

Hosted? One iframe.

Don't want to host the proxy yourself? Embed the publishd.app-hosted demo on your site. Inference still routes through this site's server-held key.

<iframe
  src="https://publishd.app/chat-demo.html"
  style="border:0;width:100%;height:760px;max-width:420px"
  title="AI chat assistant"
></iframe>

[04/zip contents]

Five files. That's it.

chat-demo.html
The widget. Drop on any static host.
api/chat.ts
Streaming proxy — drop into Next.js as app/api/chat/route.ts.
api/system-prompt.ts
Edit this with your business facts.
.env.example
Documents GROQ_API_KEY + optional CORS origins.
LICENSE
MIT — fork, rename, ship under your brand.

[05/faq]

Common questions about the free chatbot kit.

Yes. MIT licensed, no usage cap, no branding, no telemetry. Inference uses your own free Groq API key. Hosting is your own (Cloudflare's free Worker tier handles a small business site easily).