Skip to content

Latest commit

 

History

History
43 lines (30 loc) · 2.11 KB

File metadata and controls

43 lines (30 loc) · 2.11 KB

Cortex 1.1.2: System Requirements

This document outlines the hardware and software specifications required to run Cortex 1.1.2 reliably and efficiently.

💻 Software Prerequisites

Requirement Version Notes
Operating System Linux (Ubuntu 22.04+ recommended) MacOS and Windows support (Best Effort)
Python 3.10 - 3.12 Required for agent execution via PyO3
Pip 23.0+ Required for automatic dependency installation
Nvidia Driver 535+ Required only for CUDA/GPU acceleration
OpenSSL 1.1.1+ Required for secure bundle encryption/decryption

⚙️ Hardware Specifications

Cortex 1.1 implements dynamic Memory Threshold Guards to prevent system lockups. Requirements scale based on the complexity of the agents and models being executed.

Tier 1: Small Agents & APIs (e.g., Flask, Scrappers)

  • CPU: 2+ Cores (AVX2 support recommended)
  • RAM: 4GB Baseline (1GB required per active bundle)
  • Storage: 500MB for runtime + bundle size

Tier 2: Local LLM Inference (e.g., Llama-3 8B, Mistral)

  • CPU: 4+ Cores (AVX2/AVX512)
  • RAM: 16GB+
  • GPU: 8GB VRAM+ (Recommended for low latency)
  • Storage: 10GB+ (For model weights and KV Cache)

Tier 3: Large Scale / BLOOM Models (176B+)

  • CPU: 16+ Cores
  • RAM: 55GB+ (Enforced via Memory Guard)
  • GPU: Multiple A100/H100 recommended for BLOOM Full
  • Storage: 350GB+ (For model weights)

🛡️ Production Security Requirements

  • Environment Variables: Root access to set CORTEX_BUNDLE_PASSWORD if using encrypted bundles.
  • Disk Permissions: Write access to the /tmp directory (used for isolated bundle inflation).
  • Networking: Outbound access to registry if agents require pip install during runtime init.