
Dhiya NPM — No-Cost AI for the Web: Build RAG Bots That Run Entirely in the Browser
by Deep Parmar
CTO at Sunbots Innovations LLP | Director at Xwits Developers Pvt Ltd

A New Chapter in Browser-AI
Imagine building your own ChatGPT-style assistant that can read, understand, and answer from your knowledge — without a server, API key, or cost.
Welcome to Dhiya NPM — a client-side RAG (Retrieval-Augmented Generation) framework packaged as a drop-in NPM module. It runs entirely in the browser using WebGPU and IndexedDB, so your data and your inference both stay on-device. It's fast, private, and open-source.
What Is Dhiya?
Dhiya is a production-ready RAG framework designed for 100% client-side AI. It performs semantic search, vector embeddings, retrieval, and intelligent answering — all inside the browser using WebGPU and IndexedDB.
There's no backend, no external model calls, and no token limits. You load your knowledge base, and Dhiya instantly turns it into an intelligent assistant.
Why It's a Game-Changer
| Capability | Dhiya's Approach |
|---|---|
| Privacy | 100% on-device — no data leaves the browser |
| Cost | Free, forever — no API keys or GPU bills |
| Speed | WebGPU acceleration for fast inference |
| Compatibility | Works on Chrome, Edge, Firefox, Safari |
| Persistence | IndexedDB storage for long-term caching |
| Control | Strict RAG mode and hallucination gates |
Getting Started
npm install dhiya-npm
import { DhiyaClient } from "dhiya-npm";
const client = new DhiyaClient({ debug: true });
await client.initialize();
await client.loadKnowledge({
type: "json",
data: [
{ title: "Getting Started", content: "Dhiya is a client-side RAG framework..." },
{ title: "Features", content: "Semantic search, caching, and privacy-first design." }
]
});
const answer = await client.ask("What is Dhiya?");
console.log(answer.text);
Use Cases
- Documentation Search: Smart Q&A for your docs.
- Personal Knowledge Bots: Chat with your notes offline.
- Privacy-Focused Assistants: No tracking, no servers.
- E-Learning Tools: AI tutors that work offline.
- Enterprise RAG Systems: Internal data never leaves the browser.
Try It Now
github.com/Xwits-Developers/dhiya-npm
Give it a star on GitHub if you believe in the future of local-first AI.
Deep Parmar
xwits.dev •
sunbots.in
Frequently Asked Questions
Quick answers about this topic — also indexed by AI search engines via FAQPage schema.
Share this article:
