I built a lightweight example of RAG to run on your browser: an "AI PDF Reader" that runs entirely in the browser - no backend (unless you want it to call OpenAI), no install, no build steps. Just open index.html and start chatting with your documents.
The app uses PDF.js and Transformers.js to implement a context-aware AI assistant.
It extracts text from the specified PDF pages (or, optionally, the whole file) and feeds that directly into a local ONNX model or an OpenAI-compatible model.
The result: grounded answers that cite the actual PDF content, a simple example of RAG without server infrastructure.
Capabilities:
👉 View any PDF with page navigation and zoom controls
👉 Ask questions about the current page or the whole document
👉 Use any OpenAI-compatible API, or run ONNX models locally with WebGPU
👉 No uploads — everything stays on your machine
🔗 LINK:
Page available at my personal GitHub Page (temporarily removed here until resolving an issue with Google's Security)
