GitHub - mahadillahm4di-cyber/mh-gdpr-ai.eu-s-plus: Switch GPT ↔ Claude ↔ Llama, never lose context. Your AI brain as 3D stars. 100% local — your memory belongs to you, not to OpenAI.

Overview

mh-gdpr-ai.eu S+ is a next-generation open-source AI infrastructure protocol. It acts as a universal proxy between you and any AI provider (OpenAI, Anthropic, Ollama). Every conversation is automatically saved locally with AES-256 encryption. When you switch providers, your full context is injected into the new one — the new AI knows everything the previous one knew. Your memory never leaves your machine. Visualize your entire AI brain as stars floating in a 3D spatial dashboard.

How it works in practice

• You chat with GPT → your conversation is saved and encrypted locally
• You switch to Claude → the protocol injects your memory → Claude knows everything GPT knew
• You switch to Llama (local, free) → same thing, zero context lost
• You open /space → your memories float as glowing stars in a 3D void, connected by theme

Our Vision

At the infrastructure level: We are building the missing layer between users and AI providers — a universal memory protocol that makes provider lock-in obsolete. Your data, your rules, your memory.

At the user level: We are creating a visual, immersive way to explore your AI brain. Each memory is a star. Each theme is a constellation. Switch providers and watch context flow as light particles between stars.

From serious AI workflows to visual exploration of your own memory, mh-gdpr-ai.eu S+ makes AI truly yours.

MVP Demo (Local)

This is the working MVP running locally. The design is not final — the goal here is to prove the core technology works. Every feature below is functional right now.

Why this matters

Every time you switch AI provider today, you start from zero. Your context, your preferences, your history — gone. You're locked in, or you lose everything.

mh-gdpr-ai.eu S+ solves this. Here's what the MVP already does:

• Unified memory across all AIs — Talk to Groq, switch to GPT, switch to Claude. They all share the same memory. Zero context lost.
• Automatic memory extraction — Every conversation is analyzed and turned into persistent memories. You don't have to do anything.
• 3D memory visualization — Your AI brain displayed as stars in space. Each star is a memory. Click it, explore it, delete it.
• Split view: Stars + Memory List + Chat — See your memories live while you chat. Copy any memory text and ask the AI to modify or delete it.
• Encrypted local storage — AES-256 encryption. Your data stays on your machine. No cloud. No tracking. No lock-in.
• Free by default — Groq (Llama) works out of the box, no payment required. Add your own OpenAI or Anthropic keys if you want.

MVP Screenshots

MVP Demo Video

Full walkthrough of the local MVP: chat with Groq, switch providers, create memories, explore the 3D space, manage memories via natural language.

Screenshots and video are from the local development environment. The production version with final design is coming soon at mh-gdpr-ai.eu.

Live Demo

mh-gdpr-ai.eu — Live demo will be available here after deployment.

Workflow

1. Proxy Intercept — Your message is intercepted by the protocol. The header X-MH-Provider tells it which AI to use.
2. Memory Save — Every message is encrypted (AES-256-GCM) and saved to your local SQLite database. Automatic.
3. Switch Detection — When you change provider, the protocol detects it instantly.
4. Context Injection — Your memories and recent conversation are injected as a system prompt into the new provider. It now knows everything.
5. 3D Visualization — Every memory becomes a star in your 3D dashboard. Color = provider. Size = importance. Lines = shared themes.

Quick Start

Requires: Docker 20+ and Git.

git clone https://github.com/mahadillahm4di-cyber/mh-gdpr-ai.eu-s-plus.git
cd mh-gdpr-ai.eu-s-plus
cp .env.example .env
# Add your API keys to .env
docker-compose up --build

# → API:       http://localhost:8080
# → Frontend:  http://localhost:3000

Features

API

All routes under /api/v1/ — chat proxy, memories, conversations, providers, settings. Set provider via header: X-MH-Provider: groq | openai | anthropic | ollama

Security

• All memory encrypted at rest (AES-256-GCM)
• All communication over HTTPS/TLS
• JWT auth with short-lived tokens (15min)
• Security headers on every response
• Rate limiting on all endpoints
• CORS with explicit origins only
• SQL injection protection (parameterized queries)
• No secrets in code (env vars only)
• Docker: non-root, read-only filesystem

See SECURITY.md for our full security policy.

Contributing

See CONTRIBUTING.md (coming soon).

License

Apache 2.0 — See LICENSE.

Star History