Skip to content

YASSERRMD/barq-web-rag

Repository files navigation

Barq RAG

A high-performance, browser-native Retrieval-Augmented Generation (RAG) platform. This application enables local document processing and private AI interaction without external API dependencies or data exfiltration.

Core Technology Stack

The platform utilizes a multi-layered WASM and WebGPU architecture for client-side inference:

  • RAG Engine: barq-mesh-browser – Native WebAssembly mesh orchestrating parallel workers for high-speed document ingestion and hybrid retrieval.
  • Vector Storage: barq-vweb – High-performance HNSW and BM25 indexing with local persistence.
  • Compute Runtime: barq-wasm – SIMD-accelerated compute kernels for vector operations and similarity search.
  • AI Inference: transformers.js / WebGPU – Running local LLMs (e.g., LFM2.5) directly on the client hardware.

Features

  • Private & Local: 100% of data processing, embedding, and inference occurs on-device.
  • Hybrid Search: Unified BM25 and semantic search with RRF reranking for superior context relevance.
  • Universal Ingestion: Parallel processing of PDF, DOCX, Markdown, and TXT files.
  • State-of-the-Art Retrieval: Leverages barq-mesh-browser for near-instantaneous indexing of large document sets.

Getting Started

  1. Install dependencies: npm install
  2. Launch development server: npm run dev
  3. Build for production: npm run build

Prerequisites

  • Modern browser with WebGPU and WebAssembly Threading support (Chrome/Edge 113+, Safari 17+).

Powered by the Barq Mesh ecosystem.

About

A fully browser-native RAG application for document Q&A, powered by Rust and WebAssembly with local vector search, embeddings, and in-browser LLM inference.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors