LetterMint AI – Browser-Native AI Content Writer React App

LetterMint AI is a complete AI-powered content creation and productivity suite that runs full Large Language Models entirely inside the user’s browser. No backend servers, no API keys, no monthly subscription fees. Powered by WebLLM and WebGPU hardware acceleration, it delivers real AI inference using models like Llama 3.2, Phi 3.5, and Gemma 2 on hardware users already own.This is not a wrapper around a cloud API. This is not a UI mockup with fake responses. LetterMint AI downloads real LLM weights into the browser cache and runs genuine on-device inference with streaming output. Completely offline-capable once models are cached.
Why LetterMint AI is Different
Zero API costs, forever. Models run locally via WebGPU. You pay $0.00 for inference, no matter how many users you scale to. No OpenAI bills, no Anthropic invoices, no token metering.
Absolute privacy. No data ever leaves the device. No telemetry, no cloud processing. Ideal for enterprise compliance, legal writing, medical notes, or personal journaling.
No backend required. Pure front-end architecture built with React + Vite. Deploy to any static host (Vercel, Netlify, Firebase Hosting, GitHub Pages) at zero cost.
Real AI, not a demo. On supported browsers (Chrome 113+ / Edge 113+), the AI Writer, Summarizer, Code Generator, and all 13 tool categories produce genuine model-generated output, not pre-written placeholder text.
Core Features
AI Writer and Editor. Rich text editing with inline AI generation, rewriting, grammar correction, and tone adjustment. Supports markdown and real-time streaming output from the local model.
13 AI-Powered Tools. SQL Generator, Code Writer, Stripe API Generator, Python Generator, Summarizer, Translator, Regex Builder, JSON Formatter, Text Classifier, Data Extractor, and more. Each tool uses dedicated system prompts for accurate, context-aware generation.
Knowledge Base and RAG. Upload local documents and ask the AI questions grounded in your content. Real retrieval-augmented generation running entirely in-browser.
30+ Writing Templates. Blog posts, marketing emails, social media copy, technical documentation, sales pitches, cover letters, and more. Each template includes a tuned system prompt that drives actual AI output. Users can also create and save custom templates.
Document and Project Management. Create, organize, tag, search, and manage documents and projects. Full CRUD with persistent storage across browser sessions.
Export. Download documents as PDF, Markdown, or plain text. Fully client-side, no server processing required.
Dashboard with Activity Tracking. Documents created, words generated, time saved, and AI requests. Tracked from real user activity and updated live.
Global Search (Ctrl+K). Instant search across all documents, tools, templates, and pages.
Model Selection and Configuration. Choose between Llama 3.2 (1B / 3B), Phi 3.5 Mini, or Gemma 2 (2B). Adjust temperature, token limits, and generation parameters per session.
Light and Dark Mode. Fully integrated theme switching with smooth transitions.
Premium UI/UX. Built with Tailwind CSS, Framer Motion animations, and Shadcn UI components for a polished, responsive, production-grade interface.
Built for Developers
Clean, modern, and highly extensible codebase designed for easy customization:

Layer
Technology

Framework
React 19 + Vite 6

Styling
Tailwind CSS + Radix UI Primitives + Shadcn UI

AI Engine
@mlc-ai/web-llm (WebGPU accelerated)

Routing
React Router v7

Storage
IndexedDB for model caching, localStorage for app data

Animations
Framer Motion

Testing
Vitest

What’s Fully Working vs. Demo Scaffolding
LetterMint AI is a working product, not a static mockup. The table below clearly separates genuinely functional features from the lightweight scaffolding included so the template runs standalone without a backend.