Run a neuro‑symbolic AI entirely on your laptop – no API keys, no hallucinations
TL;DR: I built CORTEX – a local‑first AI framework that combines deterministic symbolic reasoning (math, logic, physics) with local LLMs (GGUF). It includes a security scanner (Mythos), a BeastMode performance booster, and a formal logic prover. All runs offline.
👉 Star on GitHub: github.com/vimalspaceton618-afk/CORTEX
The problem with today’s AI agents
Most “intelligent” systems rely on black‑box APIs:
- ❌ Hallucinations on simple math (
2+3*4→ sometimes 20, sometimes 14) - ❌ Leaking proprietary code to third‑party servers
- ❌ High latency and unpredictable costs
What if you could keep the power of LLMs for language understanding, but switch to a deterministic engine for everything that requires precision?
That’s exactly what CORTEX does.
What is CORTEX?
CORTEX is a neuro‑symbolic hybrid:
| Domain | How CORTEX handles it |
|---|---|
| Natural language / chat | Local SLM (GGUF – e.g., Llama 3, Mistral) |
| Math & physics | Symbolic engine – exact, no drift |
| Logic & code verification | Formal logic prover (deterministic) |
| Security auditing | Mythos Threat Ontology (12 domains) |
| Long‑term memory | Episodic memory with vector + TF‑IDF |
Everything runs in‑process, zero HTTP calls. Your data never leaves your machine.
Quick demo: security scanning
Let’s scan a dummy project for vulnerabilities – without sending a single file to the cloud.
1. Install CORTEX
bash
git clone https://github.com/vimalspaceton618-afk/CORTEX
cd CORTEX
npm install
npm run build
Top comments (1)
Actually, we wanted a local not Big frontier models and support for Pewdiepie🫡