QWE Solutions EMEA
QWE-BrAIn: Data-Provenance-First Enterprise AI
A self-hosted, sovereign, and defense-grade AI assistant platform built to prevent model collapse and safeguard your organization's proprietary intellectual property.
Core Technical Features
Data Provenance Engine
Track the exact lineage of your knowledge. Classify data into PRIMARY (authoritative human-generated sources like textbooks, PDFs, and Obsidian vaults) and SECONDARY (synthesized AI records, scratchpads, and agent logs) to prevent model collapse.
Commit to Brain Curation
Incorporate a secure human-in-the-loop review system. Curation loops allow administrators and users to explicitly validate AI insights and write them back into the local Obsidian workspace, upgrading them to PRIMARY status.
Local-First RAG Hierarchy
Prioritize local primary sources in retrieval and search. All external web-search tools are strictly gated behind explicit user approval, keeping your private context safe from external leakage.
Multi-Gateway Operation
Deploy secure Telegram bots, Slack Socket Mode gateways, and multi-profile embeddable web widgets with granular RAG directory scoping. Ground LLMs locally using LanceDB vector stores.
Enterprise Security & Design
Engineered to guarantee strict data containment and defense-grade infrastructure protection.
Zero Hardcoded Credentials
A secure 'First Sign-On' bootstrapping initialization flow removes static admin passwords. Features an offline, cryptographically generated Master Recovery Key to rescue admin accounts without SSH.
Application-Level Encryption
All sensitive database fields, user keys, and configuration secrets are encrypted at rest with AES-256-GCM. Native Node.js APIs (e.g. crypto) are leveraged to eliminate native compiled exploits.
Sovereign Monorepo Stack
Built as a modular TypeScript monorepo combining React, Prisma ORM, and LanceDB. Connects to Ollama, Llama.cpp, or OpenRouter, keeping your private corporate context 100% local.