ComingUp ComingUp
Omni - Your Universal AI

Omni - Your Universal AI

Local-first AI assistant PWA built with Vanilla JS and WebLLM.

Jun 15, 2026 AI & Machine Learning
browser-llm local ai pwa webgpu webllm

Gallery

Omni - Your Universal AI

About

Omni is a turnkey, production-ready Progressive Web App (PWA) designed as a universal, minimalist AI assistant. It runs open-source models (Gemma 2, Llama 3.2, Phi 3.5, Qwen 2, Llama 3.1) 100% locally in the user's browser via WebGPU acceleration. A standout UX feature is its automatic hardware detection. The app inspects the user's machine capabilities on launch and dynamically recommends the optimal model tier to maximize performance and prevent browser crashes. It also includes a "Bring Your Own Key" cloud fallback toggle for Gemini, Groq, or OpenAI keys. The interface features a premium, distraction-free, Notion-style dark aesthetic with native markdown text streaming and a robust universal task system (General, Coding, Writing, Analysis). It includes a service worker auto-update system that detects new Vercel deployments and safely pushes updates to the client. Tech Stack: Vanilla HTML5, CSS3, JavaScript (ES6+), WebLLM, WebGPU, marked.js, Vercel. 📊 Business & FinancialsWhy are you selling? I am liquidating this unlaunched codebase asset to cover an immediate, urgent personal expense.Monthly Revenue: $0 (Pre-revenue)Monthly Expense: $0 (The local-first and BYOK cloud architecture means the owner incurs absolute zero hosting or API maintenance costs). Traffic: Minimal/Testing traffic from Vercel deployment. 💰 The SaleAsking Price: $450 USD (Approx. ₹37,500 INR. This leaves you room to accept a lower offer of $360 / ₹30,000 if a buyer tries to negotiate).What is included in the sale?Full ownership transfer of the private GitHub repository . Production deployment configuration files for automatic Vercel hosting. Comprehensive architectural developer breakdown documentation .

Comments (6)

Halle Bernier Halle Bernier 3 weeks ago

running llama 3.1 locally in a browser must be agonizingly slow

Destinee Fay Destinee Fay 3 weeks ago

how bad is inference speed on mobile

Camryn Barrows Camryn Barrows 3 weeks ago

webllm inference on consumer hardware is still painfully slow

Felicity Rau Felicity Rau 3 weeks ago

running llms locally in my browser with no api key is wild

Ladarius Mills Ladarius Mills 2 weeks ago

vanilla js for webllm is smart, keeps the wasm pipeline tight

Jalen VonRueden Jalen VonRueden 2 weeks ago

vanilla js and local inference, really smart combo