Answers only from your documents.
Never made up.
Drop an AI chat widget onto your site with one line of code. It replies to visitors using only the documents you upload — and runs entirely on your own server, so your data never leaves it.
Acme Assistant
Online · powered by your docs
Live in minutes, grounded from day one
From your documents to answered questions — without your data ever leaving your infrastructure.
-
1
Upload your documents
Drop in PDF, DOCX, or TXT — policies, manuals, FAQs, anything you want answered.
-
2
Indexed privately
Documents are processed and indexed on your own server. No third party ever sees them.
-
3
Drop in one script
Paste a single line into your site. The widget appears, styled to match your brand.
-
4
Instant, grounded answers
Streaming responses in each visitor's own language, around the clock.
Everything a trustworthy assistant needs
Accurate, private, and built to match your brand — out of the box.
No hallucinations
Answers come only from your documents. If it's not there, it says so — never invents.
Self-hosted & private
Runs on your server. Data never leaves it and is never used to train third-party AI.
Multilingual replies
Detects each visitor's language and answers in it — from one set of documents.
Assistant role & voice
Set its persona, tone, and brand voice so it sounds like part of your team.
Multi-tenant isolation
Every client's documents and conversations are walled off — strictly separated by design.
Knowledge-gap alerts
When it can't answer, it logs the question so you know exactly what to add next.
Streaming answers
Replies stream token-by-token, so it feels like chatting with a real live agent.
Customizable widget
Set the bot name, greeting, color, and screen position to fit your site perfectly.
Paste one tag. You're live.
Add this single script to your site's HTML. The widget loads, reads your settings, and starts answering — no build step, no dependencies.
Live in ~2 minutes
<!-- Add before </body> -->
<script src="https://yourdomain.com/widget.js"
data-site-key="site_xxx"></script>Runs entirely on your own server
An open, swappable stack — local LLMs, no third-party AI, and your data stays exactly where it is. Deploy it all with one command.
Node.js + Express
Web application layer
Python + FastAPI
AI service layer
Ollama · Qwen2.5
Local LLM — no third parties
Qdrant
Vector search
MySQL
Application data
Stripe
Billing
nginx
HTTPS & routing
Docker
One-command deploy
Open and swappable — bring your own model, database, and server. Your data stays put.
One plan. Everything included.
No tiers, no usage meters, no per-message fees. Just everything ChatWidget does, for one flat price.
No per-message fees. Cancel anytime.
- Unlimited document uploads
- The embeddable chat widget
- Every feature — multilingual, personas, streaming
- Fully self-hosted on your own server
- Knowledge-gap alerts & logs
Self-hosted · your data never leaves your server
Questions, answered
Give your visitors answers they can trust
Self-hosted, private, and grounded only in your documents. Add it to your site in minutes.