Analysis updated 2026-05-18
Extract text from scanned documents and receipts without sending data to a server.
Transcribe meetings or dictate voice-to-email locally in the browser.
Convert a wireframe sketch into working HTML.
Parse contracts and financial statements for risky clauses privately.
| stratoslab/stratos-office | abhishek-akkal/finova | adan-shahid/ecommerce_website | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | JavaScript | JavaScript | JavaScript |
| Setup difficulty | easy | easy | moderate |
| Complexity | 2/5 | 1/5 | 3/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Chrome or Edge 113+ with a WebGPU-capable GPU and at least 4 GB of RAM.
Stratos Office is a browser-based AI assistant that runs entirely on your own machine, no server, no API keys, and no data sent anywhere externally. It loads Google's Gemma 4 E2B model directly in Chrome or Edge using WebGPU, a technology that lets browsers use your computer's graphics card for heavy computation. The model downloads once on first launch and is cached for all future visits. The tool covers over 30 tasks in six categories. Document tasks include extracting text from scanned images (OCR), parsing receipts into structured data, transcribing handwriting, and analyzing contracts for risk and key clauses. Visual tasks include reading chart data, converting wireframe sketches to working HTML, and detecting objects in images. Audio tasks cover meeting transcription and minutes, voice-to-email dictation, and multilingual transcription across 140+ languages. Writing tools handle email drafting, tone rewriting, summarization, and code review. A research mode uses the optional TinyFish web search API, the only feature contacting an external service. Privacy-focused tasks cover medical records, legal documents, and financial statement parsing. Results export as plain text, JSON, Markdown, or HTML. A 200-entry task history persists across sessions via IndexedDB storage. An offline mode disables all web features for fully air-gapped use. The frontend is built with React 18, TypeScript, Vite, and Tailwind CSS v4. Inference runs in a background thread to keep the interface responsive. Requirements: Chrome or Edge 113 or later, a WebGPU-capable GPU, and at least 4 GB of RAM.
A browser-based AI assistant running fully offline in Chrome or Edge via WebGPU, covering over 30 document, visual, audio, and writing tasks.
Mainly JavaScript. The stack also includes React, TypeScript, Vite.
No license file was found in the README.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.