YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Nekomata

NEKOMATA 猫又

The Autonomous Container Engine

Infrastructure that rewrites itself.


 β–ˆβ–ˆβ–ˆβ•—   β–ˆβ–ˆβ•—β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•—β–ˆβ–ˆβ•—  β–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ•—   β–ˆβ–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•—
 β–ˆβ–ˆβ–ˆβ–ˆβ•—  β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•”β•β•β•β•β•β–ˆβ–ˆβ•‘ β–ˆβ–ˆβ•”β•β–ˆβ–ˆβ•”β•β•β•β–ˆβ–ˆβ•—β–ˆβ–ˆβ–ˆβ–ˆβ•— β–ˆβ–ˆβ–ˆβ–ˆβ•‘β–ˆβ–ˆβ•”β•β•β–ˆβ–ˆβ•—β•šβ•β•β–ˆβ–ˆβ•”β•β•β•β–ˆβ–ˆβ•”β•β•β–ˆβ–ˆβ•—
 β–ˆβ–ˆβ•”β–ˆβ–ˆβ•— β–ˆβ–ˆβ•‘β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•—  β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•”β• β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•”β–ˆβ–ˆβ–ˆβ–ˆβ•”β–ˆβ–ˆβ•‘β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•‘
 β–ˆβ–ˆβ•‘β•šβ–ˆβ–ˆβ•—β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•”β•β•β•  β–ˆβ–ˆβ•”β•β–ˆβ–ˆβ•— β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•‘β•šβ–ˆβ–ˆβ•”β•β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•”β•β•β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•”β•β•β–ˆβ–ˆβ•‘
 β–ˆβ–ˆβ•‘ β•šβ–ˆβ–ˆβ–ˆβ–ˆβ•‘β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•—β–ˆβ–ˆβ•‘  β–ˆβ–ˆβ•—β•šβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ•”β•β–ˆβ–ˆβ•‘ β•šβ•β• β–ˆβ–ˆβ•‘β–ˆβ–ˆβ•‘  β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘   β–ˆβ–ˆβ•‘  β–ˆβ–ˆβ•‘
 β•šβ•β•  β•šβ•β•β•β•β•šβ•β•β•β•β•β•β•β•šβ•β•  β•šβ•β• β•šβ•β•β•β•β•β• β•šβ•β•     β•šβ•β•β•šβ•β•  β•šβ•β•   β•šβ•β•   β•šβ•β•  β•šβ•β•

                    Infrastructure is code.
                    Code is data.
                    Data is alive.

What Is This?

Nekomata is the first AI-native container engine built from first principles in Common Lisp. Not a Kubernetes alternative. A Kubernetes replacement.

The Nekomata (猫又) is a supernatural cat with two tails: code and data. In Common Lisp, these are the same thing. Infrastructure is s-expressions. Containers evaluate. They transform. They rewrite themselves.

╔══════════════════════════════════════════════════════════════════════╗
β•‘                                                                      β•‘
β•‘   Kubernetes:  50B of YAML debt on a distributed key-value store     β•‘
β•‘   Nekomata:    s-expressions that evaluate their own infrastructure  β•‘
β•‘                                                                      β•‘
β•‘   Kubernetes:  Reactive watch loops waiting for state drift           β•‘
║   Nekomata:    Predictive LLM→ASM compilation of intent              ║
β•‘                                                                      β•‘
β•‘   Kubernetes:  Monolithic control plane (etcd + api-server + ...)    β•‘
β•‘   Nekomata:    Sparse Mixture of Experts β€” 6 specialists, Top-K=2   β•‘
β•‘                                                                      β•‘
β•‘   Kubernetes:  Immutable containers, no runtime modification         β•‘
β•‘   Nekomata:    Live REPL injection + genetic self-modification       β•‘
β•‘                                                                      β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Setup

# One command β€” installs Docker + SBCL + Quicklisp
bash setup.sh

Supports Ubuntu/Debian, Fedora/RHEL, Arch Linux, and macOS (via Homebrew).

Or manually:

# 1. Docker
curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER && newgrp docker

# 2. SBCL
sudo apt install sbcl   # or: brew install sbcl

# 3. Quicklisp
curl -O https://beta.quicklisp.org/quicklisp.lisp
sbcl --load quicklisp.lisp --eval '(quicklisp-quickstart:install)'

# 4. Load Nekomata
cd nekomata && sbcl --load nekod.asd --eval '(ql:quickload :nekod)'

Quick Start

;; Load the system
(ql:quickload :nekod)

;; Start the daemon
(nekod:nekomata-main)

;; Define a container β€” infrastructure as code, literally
(nekod:define-container my-app
  :image "nginx:latest"
  :ports '((8080 . 80))
  :env '((:NODE_ENV . "production"))
  :restart-policy :always)

;; Self-healing failover chain
(nekod:with-failover my-app my-app-replica my-app-standby)

;; Drop into the live REPL
(nekod:nekod-repl)

Architecture

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚                         NEKOMATA ENGINE                              β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚  CLI      β”‚   β”‚  EVAL LOOP   β”‚   β”‚  CONTAINER REGISTRY      β”‚  β”‚
β”‚  β”‚  neko.lisp│──>β”‚  nekod-repl  │──>β”‚  define-container        β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β”‚  with-failover           β”‚  β”‚
β”‚                                      β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚                   SPARSE MoE ROUTER                           β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   Input(8) ──> Gate Network ──> Top-K(2) ──> Softmax         β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β” β”‚  β”‚
β”‚  β”‚   β”‚SCHEDULINGβ”‚NETWORKINGβ”‚ STORAGE β”‚ RECOVERY β”‚ SECURITY β”‚SCβ”‚ β”‚  β”‚
β”‚  β”‚   β”‚ cron     β”‚ socket   β”‚ disk    β”‚ failover β”‚ auth     β”‚alβ”‚ β”‚  β”‚
β”‚  β”‚   β”‚ deploy   β”‚ http     β”‚ persist β”‚ rollback β”‚ encrypt  β”‚e β”‚ β”‚  β”‚
β”‚  β”‚   β”‚ queue    β”‚ route    β”‚ worm    β”‚ backup   β”‚ seal     β”‚  β”‚ β”‚  β”‚
β”‚  β”‚   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”˜ β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚                   LLM β†’ ASM COMPILER                          β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   "scale nginx to 5"                                          β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚  β”‚
β”‚  β”‚   β”‚ TOKENIZE │──>β”‚ AST      │──>β”‚ CODEGEN  │──>β”‚ VERIFY  β”‚ β”‚  β”‚
β”‚  β”‚   β”‚ intent   β”‚   β”‚ extract  β”‚   β”‚ x86/ARM  β”‚   β”‚ bytes   β”‚ β”‚  β”‚
β”‚  β”‚   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚                   SELF-MODIFICATION                            β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚  β”‚
β”‚  β”‚   β”‚ HOT-SWAP β”‚   β”‚ GENETIC  β”‚   β”‚ STATE    β”‚   β”‚ROLLBACK β”‚ β”‚  β”‚
β”‚  β”‚   β”‚ live fn  β”‚   β”‚ evolution β”‚   β”‚ capture  β”‚   β”‚ points  β”‚ β”‚  β”‚
β”‚  β”‚   β”‚ redefine β”‚   β”‚ crossoverβ”‚   β”‚ snapshot β”‚   β”‚ restore β”‚ β”‚  β”‚
β”‚  β”‚   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚                   DOCKER SUBSTRATE                             β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   Unix Socket ──> HTTP/1.1 Protocol ──> Container Lifecycle   β”‚  β”‚
β”‚  β”‚   /var/run/docker.sock                                        β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   create | start | stop | kill | list | inspect | stats       β”‚  β”‚
β”‚  β”‚   pull   | tag   | prune | network | volume                   β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β”‚                                                                     β”‚
β”‚  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”‚
β”‚  β”‚                   SOCKET INGRESS PIPELINE                      β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   /var/run/docker.sock                                        β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   Unix Domain Socket (SBCL/CCL/IOlib portable)                β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   Incremental HTTP Parser (chunked + content-length)          β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   Docker /events Stream (infinite chunked consumption)        β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   Compiled DFA Classifier (Thompson + subset construction)    β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   8-Feature Routing Vector ──> Top-K=2 Sparse MoE             β”‚  β”‚
β”‚  β”‚         β”‚                                                     β”‚  β”‚
β”‚  β”‚         v                                                     β”‚  β”‚
β”‚  β”‚   Expert Dispatch + Deterministic Receipt                     β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   Security: capability-based policy, path validation,         β”‚  β”‚
β”‚  β”‚             DENY-by-default, no raw socket to untrusted agent β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β”‚   Backpressure: high/low watermarks, queue depth cap,         β”‚  β”‚
β”‚  β”‚                 event size limits, zero unbounded growth       β”‚  β”‚
β”‚  β”‚                                                               β”‚  β”‚
β”‚  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

The 5-Layer Stack

        LAYER 5: SELF-MODIFICATION
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚  Genetic evolution loop                     β”‚
        β”‚  Hot-swap live function redefinition        β”‚
        β”‚  State capture + rollback points            β”‚
        β”‚  Tournament / roulette selection            β”‚
        β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                          β”‚
        LAYER 4: LLM β†’ MACHINE CODE
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚  Intent tokenization                        β”‚
        β”‚  AST extraction (infra-ast struct)          β”‚
        β”‚  x86-64 + ARM64 code generation            β”‚
        β”‚  Byte-level verification                    β”‚
        β”‚  Constraint algebra (cpu/mem/latency)       β”‚
        β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                          β”‚
        LAYER 3: SPARSE MoE ROUTING
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚  6 domain experts                           β”‚
        β”‚  Noisy Top-K gating (k=2)                  β”‚
        β”‚  Load-balance loss                          β”‚
        β”‚  Xavier/Glorot trained weights (R pipeline) β”‚
        β”‚  WORM-sealed gradient history               β”‚
        β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                          β”‚
        LAYER 2: REGEX MATHEMATICS
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚  Kleene algebra (concat/union/star)         β”‚
        β”‚  DFA compilation + socket event filtering   β”‚
        β”‚  NL intent β†’ regex pattern mapping          β”‚
        β”‚  Formal algebra at the wire level           β”‚
        β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                          β”‚
        LAYER 1: LISP CORE
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚  s-expression infrastructure DSL            β”‚
        β”‚  Container lifecycle management             β”‚
        β”‚  Docker Engine API (Unix socket)            β”‚
        β”‚  Condition system + restarts                β”‚
        β”‚  Image persistence (save/restore world)     β”‚
        β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

User Guide

Container Management

╔══════════════════════════════════════════════════════════════╗
β•‘  COMMAND              DESCRIPTION                            β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘  neko run <spec>      Create and start a container          β•‘
β•‘  neko stop <name>     Stop a running container              β•‘
β•‘  neko kill <name>     Force-kill a container                β•‘
β•‘  neko list            List all managed containers           β•‘
β•‘  neko status          Show system status + expert load      β•‘
β•‘  neko evolve          Run one generation of evolution       β•‘
β•‘  neko repl            Drop into the Nekomata REPL          β•‘
β•‘  neko save            Snapshot current world state          β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Container DSL Examples

;; ─── Basic Container ────────────────────────────────────────────
(define-container web-server
  :image "nginx:alpine"
  :ports '((80 . 80) (443 . 443))
  :env '((:WORKER_PROCESSES . "auto"))
  :restart-policy :always)

;; ─── Database with Persistent Volume ────────────────────────────
(define-container postgres-db
  :image "postgres:16"
  :ports '((5432 . 5432))
  :env '((:POSTGRES_PASSWORD . "sovereign")
         (:POSTGRES_DB . "nekod"))
  :volumes '(("/data/pg" . "/var/lib/postgresql/data"))
  :restart-policy :unless-stopped)

;; ─── Failover Chain (automatic promotion) ───────────────────────
(with-failover postgres-db postgres-replica postgres-standby)

;; ─── Natural Language Intent ────────────────────────────────────
(nekod-eval "scale web-server to 5 replicas with 512mb memory limit")
;; => Compiles to ASM, verifies bytes, executes container operations

The REPL

╔══════════════════════════════════════════════════════════════╗
β•‘                    NEKOMATA REPL v1.0                        β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘                                                              β•‘
β•‘  nekod> (list-containers)                                    β•‘
β•‘  => (#<CONTAINER web-server :running>                        β•‘
β•‘      #<CONTAINER postgres-db :running>                       β•‘
β•‘      #<CONTAINER redis-cache :running>)                      β•‘
β•‘                                                              β•‘
β•‘  nekod> (route-intent "deploy new redis cluster")            β•‘
β•‘  => Expert: SCHEDULING (0.72)                                β•‘
β•‘     Expert: STORAGE (0.28)                                   β•‘
β•‘     Action: create-cluster redis 3 replicas                  β•‘
β•‘                                                              β•‘
β•‘  nekod> (evolve-generation :population 20 :rounds 5)         β•‘
β•‘  => Generation 1: best-fitness 0.847                         β•‘
β•‘     Generation 2: best-fitness 0.891                         β•‘
β•‘     Generation 3: best-fitness 0.923                         β•‘
β•‘     ...                                                      β•‘
β•‘     Winner: scheduling-expert-v3 (fitness: 0.961)            β•‘
β•‘     Hot-swapping into live router...                         β•‘
β•‘     Done.                                                    β•‘
β•‘                                                              β•‘
β•‘  nekod> (save-nekod-image "checkpoint-2026-08-08")           β•‘
β•‘  => World saved. 847 objects, 12.3MB.                        β•‘
β•‘     WORM sealed: a8f3c91d...                                β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

MoE Router β€” Sparse Expert Selection

         β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
         β”‚            INPUT VECTOR (8 features)         β”‚
         β”‚                                             β”‚
         β”‚  [is_proof, is_auth, is_code, is_system,    β”‚
         β”‚   complexity, entropy, confidence, urgency]  β”‚
         β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                β”‚
                                v
         β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
         β”‚           GATING NETWORK                     β”‚
         β”‚                                             β”‚
         β”‚   s_i = dot(W_i, h) + b_i + noise_i        β”‚
         β”‚   noise_i ~ N(0, softplus(W_noise Β· h))     β”‚
         β”‚                                             β”‚
         β”‚   Top-K selection (K=2)                      β”‚
         β”‚   Softmax over selected ONLY                 β”‚
         β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                β”‚
               β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
               v                v                v
    β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
    β”‚  EXPERT A    β”‚  β”‚  EXPERT B    β”‚  β”‚  (inactive)  β”‚
    β”‚  prob: 0.72  β”‚  β”‚  prob: 0.28  β”‚  β”‚  prob: 0.00  β”‚
    β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
               β”‚                β”‚
               v                v
         β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
         β”‚   OUTPUT = p_A * expert_A + p_B * expert_B  β”‚
         β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

The 6 Experts

Expert Domain Handles
SCHEDULING Time + orchestration cron, deploy, queue, scaling events
NETWORKING Connectivity socket, http, routing, port management
STORAGE Persistence disk, WORM chains, volumes, snapshots
RECOVERY Fault tolerance failover, rollback, restart, backup
SECURITY Auth + crypto certificates, encryption, sealing
SCALING Load management replicas, sharding, capacity planning

Weight Training Pipeline

WORM Chain ──> R Feature Extraction ──> Xavier/Glorot Init
    (.jsonl)       (8 features)           (W1: 8x16, W2: 16x6)
                                                β”‚
                                                v
                                    Forward Pass + Backprop
                                    Binary Cross-Entropy Loss
                                                β”‚
                                                v
                                    WORM Seal (every 10 epochs)
                                    SHA-256 weight hash
                                                β”‚
                                                v
                                    Export: router_weights.json
                                                β”‚
                                                v
                                    Lisp: load-trained-weights
                                    Hot-loaded into live router

LLM β†’ Machine Code Compilation

╔══════════════════════════════════════════════════════════════╗
β•‘                                                              β•‘
β•‘  "scale nginx to 5 with 256mb memory limit"                  β•‘
β•‘                                                              β•‘
β•‘       β”‚  TOKENIZE                                            β•‘
β•‘       v                                                      β•‘
β•‘  [:scale :nginx :to :5 :with :256mb :memory :limit]         β•‘
β•‘                                                              β•‘
β•‘       β”‚  AST EXTRACT                                         β•‘
β•‘       v                                                      β•‘
β•‘  #S(INFRA-AST                                                β•‘
β•‘     :TYPE :SCALE                                             β•‘
β•‘     :TARGET "nginx"                                          β•‘
β•‘     :PARAMS (:COUNT 5 :MEMORY 268435456))                    β•‘
β•‘                                                              β•‘
β•‘       β”‚  CODEGEN (x86-64)                                    β•‘
β•‘       v                                                      β•‘
β•‘  #(#x48 #x89 #xF8    ; mov rax, rdi                         β•‘
β•‘    #x48 #x6B #xC0 5  ; imul rax, 5                          β•‘
β•‘    #xC3)              ; ret                                   β•‘
β•‘                                                              β•‘
β•‘       β”‚  VERIFY                                              β•‘
β•‘       v                                                      β•‘
β•‘  [OK] First byte valid (0x48 = REX.W prefix)                β•‘
β•‘  [OK] Last byte = 0xC3 (RET)                                β•‘
β•‘  [OK] No privileged instructions                             β•‘
β•‘  [OK] Length within bounds                                   β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Supported Backends

Backend Status Instruction Set
x86-64 Production REX prefix, MOV, ADD, IMUL, XOR, PUSH/POP, CALL, RET
ARM64 Production Fixed-width 32-bit, MOV/MOVK, ADD, SUB, MUL, BR, RET

Self-Modification Engine

╔══════════════════════════════════════════════════════════════╗
β•‘                   GENETIC EVOLUTION LOOP                      β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘                                                              β•‘
β•‘  1. POPULATION                                               β•‘
β•‘     Generate N candidate expert functions                    β•‘
β•‘                                                              β•‘
β•‘  2. EVALUATE                                                 β•‘
β•‘     Run each candidate against real container workloads      β•‘
β•‘     Fitness = success_rate * latency_inverse * memory_eff    β•‘
β•‘                                                              β•‘
β•‘  3. SELECT                                                   β•‘
β•‘     Tournament selection (pool=4) OR roulette wheel          β•‘
β•‘                                                              β•‘
β•‘  4. CROSSOVER                                                β•‘
β•‘     Swap s-expression subtrees between parents               β•‘
β•‘     Code IS data β€” crossover is tree surgery                 β•‘
β•‘                                                              β•‘
β•‘  5. MUTATE                                                   β•‘
β•‘     Random subtree replacement, constant perturbation        β•‘
β•‘                                                              β•‘
β•‘  6. HOT-SWAP                                                 β•‘
β•‘     Winner replaces live expert function β€” zero downtime     β•‘
β•‘     Automatic rollback if fitness drops below threshold      β•‘
β•‘                                                              β•‘
β•‘  7. SEAL                                                     β•‘
β•‘     State captured, rollback point created, WORM sealed      β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Regex Mathematics β€” Formal Algebra at the Wire

╔══════════════════════════════════════════════════════════════╗
β•‘                  KLEENE ALGEBRA PRIMITIVES                    β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘                                                              β•‘
β•‘  LITERAL(c)        Match single character c                  β•‘
β•‘  CONCAT(r1, r2)    Sequence: r1 then r2                      β•‘
β•‘  UNION(r1, r2)     Choice: r1 or r2                          β•‘
β•‘  KLEENE(r)         Zero or more: r*                          β•‘
β•‘  PLUS(r)           One or more: r+                           β•‘
β•‘  OPTIONAL(r)       Zero or one: r?                           β•‘
β•‘                                                              β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘                                                              β•‘
β•‘  DFA COMPILATION:                                            β•‘
β•‘                                                              β•‘
β•‘    regex-algebra.lisp  ──>  pattern-compiler.lisp            β•‘
β•‘    (algebraic form)         (DFA states + transitions)       β•‘
β•‘                                                              β•‘
β•‘    dfa-step(state, char)  =>  next-state                     β•‘
β•‘    dfa-run(dfa, string)   =>  accept | reject                β•‘
β•‘                                                              β•‘
β•‘  SOCKET FILTERING:                                           β•‘
β•‘                                                              β•‘
β•‘    Every Docker event passes through DFA filter              β•‘
β•‘    Pattern: container.(start|stop|die|kill)                  β•‘
β•‘    Only matching events reach the MoE router                 β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Project Layout

nekod/
β”œβ”€β”€ nekod.asd                    ASDF system definition
β”œβ”€β”€ setup.sh                     One-command Docker + SBCL installer
β”œβ”€β”€ README.md                    This file
β”œβ”€β”€ LICENSE                      AGPL-3.0 + cloud provider poison pill
β”œβ”€β”€ assets/
β”‚   └── nekomata-avatar.png      Production model avatar
β”‚
β”œβ”€β”€ core/                        ─── LAYER 1: LISP CORE ───
β”‚   β”œβ”€β”€ package.lisp             Package definitions (nekod + nekod.docker)
β”‚   β”œβ”€β”€ conditions.lisp          Error hierarchy + restart system
β”‚   β”œβ”€β”€ sexpr-infra.lisp         Container DSL (define-container, with-failover)
β”‚   β”œβ”€β”€ eval-loop.lisp           Nekomata REPL (nekod-eval, nekod-repl)
β”‚   β”œβ”€β”€ image-persist.lisp       World state save/restore
β”‚   └── nekomata.lisp            Entry point + container registry
β”‚
β”œβ”€β”€ docker/                      ─── DOCKER SUBSTRATE ───
β”‚   β”œβ”€β”€ unix-sock.lisp           Real Unix socket I/O (SBCL/CCL/IOlib)
β”‚   β”œβ”€β”€ protocol.lisp            Incremental HTTP/1.1 parser (chunked + CL)
β”‚   β”œβ”€β”€ events.lisp              Streaming /events consumer (infinite chunked)
β”‚   β”œβ”€β”€ container.lisp           Container lifecycle (create/start/stop/kill)
β”‚   β”œβ”€β”€ image.lisp               Image management (pull/list/tag/prune)
β”‚   β”œβ”€β”€ network.lisp             Network create/connect/disconnect
β”‚   β”œβ”€β”€ volume.lisp              Volume create/mount/prune
β”‚   └── inspect.lisp             Container + stats inspection
β”‚
β”œβ”€β”€ moe/                         ─── LAYER 3: SPARSE MoE ───
β”‚   β”œβ”€β”€ router.lisp              Gating network (score, top-k, softmax)
β”‚   β”œβ”€β”€ sparse-activate.lisp     Noisy Top-K + load-balance loss
β”‚   β”œβ”€β”€ expert-train.lisp        Online weight updates from telemetry
β”‚   β”œβ”€β”€ load-balance.lisp        Expert utilization balancing
β”‚   β”œβ”€β”€ load-trained-weights.lisp  R-trained weight loader (JSON import)
β”‚   └── experts/
β”‚       β”œβ”€β”€ scheduling.lisp      Cron, deploy, queue management
β”‚       β”œβ”€β”€ networking.lisp      Socket, HTTP, port routing
β”‚       β”œβ”€β”€ storage.lisp         Disk, WORM, volume persistence
β”‚       β”œβ”€β”€ recovery.lisp        Failover, rollback, restart
β”‚       β”œβ”€β”€ security.lisp        Auth, encryption, sealing
β”‚       └── scaling.lisp         Replicas, shards, capacity
β”‚
β”œβ”€β”€ regex-math/                  ─── LAYER 2: KLEENE ALGEBRA ───
β”‚   β”œβ”€β”€ regex-algebra.lisp       RE structs (literal/concat/union/kleene)
β”‚   β”œβ”€β”€ pattern-compiler.lisp    Thompson NFA β†’ subset DFA compilation
β”‚   β”œβ”€β”€ socket-filter.lisp       Compiled DFA event classifier + features
β”‚   └── intent-parse.lisp        NL intent β†’ expert keyword mapping
β”‚
β”œβ”€β”€ socket/                      ─── SOCKET INGRESS PIPELINE ───
β”‚   β”œβ”€β”€ backpressure.lisp        Watermarks, queue depth, size limits
β”‚   β”œβ”€β”€ policy.lisp              Capability-based security boundary
β”‚   β”œβ”€β”€ receipt.lisp             Deterministic operation receipts
β”‚   └── agent-ingress.lisp       Event β†’ classify β†’ route β†’ dispatch
β”‚
β”œβ”€β”€ llm2asm/                     ─── LAYER 4: LLM β†’ MACHINE CODE ───
β”‚   β”œβ”€β”€ tokenizer.lisp           Intent tokenization + kernel map
β”‚   β”œβ”€β”€ ast-extract.lisp         infra-ast struct, intent type detection
β”‚   β”œβ”€β”€ codegen.lisp             AST β†’ byte vector compilation
β”‚   β”œβ”€β”€ verify.lisp              Byte-level safety verification
β”‚   β”œβ”€β”€ jit-link.lisp            JIT stub (mmap + mprotect reference)
β”‚   β”œβ”€β”€ constraint-build.lisp    Constraint algebra (cpu/mem/latency)
β”‚   └── backends/
β”‚       β”œβ”€β”€ x86-64.lisp          x86-64 instruction encoders
β”‚       └── arm64.lisp           ARM64 fixed-width instruction encoders
β”‚
β”œβ”€β”€ selfmod/                     ─── LAYER 5: SELF-MODIFICATION ───
β”‚   β”œβ”€β”€ gen-loop.lisp            Genetic evolution loop
β”‚   β”œβ”€β”€ hot-swap.lisp            Live function redefinition + rollback
β”‚   β”œβ”€β”€ state-capture.lisp       State snapshots
β”‚   β”œβ”€β”€ evolve.lisp              Crossover + selection algorithms
β”‚   └── rollback.lisp            Rollback points, with-rollback macro
β”‚
β”œβ”€β”€ cli/                         ─── USER INTERFACE ───
β”‚   └── neko.lisp                8 CLI commands (run/stop/kill/list/...)
β”‚
β”œβ”€β”€ tests/
β”‚   β”œβ”€β”€ test-core.lisp           Core test suite
β”‚   β”œβ”€β”€ test-edge-cases.lisp     Edge case coverage
β”‚   └── socket-tests.lisp        Socket layer: 28 unit + 4 integration tests
β”‚
└── docs/                        ─── DOCUMENTATION ───
    β”œβ”€β”€ manifesto.org            Why Nekomata exists
    β”œβ”€β”€ architecture.org         5-layer system design
    β”œβ”€β”€ moe-theory.org           Sparse MoE mathematics
    β”œβ”€β”€ regex-calculus.org        Kleene algebra formalism
    └── migration.org            Kubernetes escape route

Why Common Lisp?

╔══════════════════════════════════════════════════════════════╗
β•‘                                                              β•‘
β•‘  CODE = DATA                                                 β•‘
β•‘  An s-expression describing a container IS that container.   β•‘
β•‘  No YAML. No templating. No string interpolation.            β•‘
β•‘  The description evaluates to the thing itself.              β•‘
β•‘                                                              β•‘
β•‘  CONDITION SYSTEM                                            β•‘
β•‘  Not try/catch. Restarts. When a container dies, the         β•‘
β•‘  condition system offers: restart, promote-replica,          β•‘
β•‘  rollback, or escalate. Decided at runtime, not design-time. β•‘
β•‘                                                              β•‘
β•‘  IMAGE-BASED PERSISTENCE                                     β•‘
β•‘  Save the entire world to disk. Restore it later.            β•‘
β•‘  The running state IS the deployment artifact.               β•‘
β•‘  No container image layers. No registries. Just the world.   β•‘
β•‘                                                              β•‘
β•‘  MACROS                                                      β•‘
β•‘  define-container is a macro. with-failover is a macro.      β•‘
β•‘  The language extends itself to become the domain.            β•‘
β•‘  Infrastructure IS language. Language IS infrastructure.      β•‘
β•‘                                                              β•‘
β•‘  HOT RELOAD                                                  β•‘
β•‘  Redefine any function while the system runs.                β•‘
β•‘  No restart. No redeployment. The new code IS the system.    β•‘
β•‘  Genetic evolution swaps expert functions live.               β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Migration from Kubernetes

╔══════════════════════════════════════════════════════════════╗
β•‘              K8S β†’ NEKOMATA MIGRATION PATH                   β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘                                                              β•‘
β•‘  Step 1: Install nekod alongside existing cluster            β•‘
β•‘          (nekod speaks Docker socket directly β€” no conflict) β•‘
β•‘                                                              β•‘
β•‘  Step 2: Convert Deployment YAMLs β†’ define-container forms   β•‘
β•‘          One YAML = one s-expression. Mechanical.            β•‘
β•‘                                                              β•‘
β•‘  Step 3: Wire failover chains                                β•‘
β•‘          (with-failover primary replica standby)             β•‘
β•‘                                                              β•‘
β•‘  Step 4: Enable MoE routing                                  β•‘
β•‘          Expert weights train on YOUR workload patterns       β•‘
β•‘                                                              β•‘
β•‘  Step 5: Enable self-modification                            β•‘
β•‘          The system evolves its own expert functions          β•‘
β•‘                                                              β•‘
β•‘  Step 6: Decommission K8s control plane                      β•‘
β•‘          etcd, kube-apiserver, kube-scheduler β€” gone.         β•‘
β•‘          Nekomata IS the control plane now.                   β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Performance Characteristics

╔══════════════════════════════════════════════════════════════╗
β•‘  METRIC                    VALUE                             β•‘
╠══════════════════════════════════════════════════════════════╣
β•‘  Memory per expert         ~2.3 KB (sparse activation)      β•‘
β•‘  Router decision latency   <1ms (dot product + softmax)     β•‘
β•‘  Active experts per call   2 (Top-K=2, out of 6)            β•‘
β•‘  Agents at 2.3KB each      500+ sustainable                 β•‘
β•‘  Code generation            x86-64 + ARM64 native           β•‘
β•‘  Hot-swap latency          ~0ms (defun redefinition)        β•‘
β•‘  World snapshot             Full state in single file        β•‘
β•‘  Container start            Direct Docker API (no etcd)     β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

License

AGPL-3.0 with cloud provider poison pill.

The engine is free. Sovereignty is not optional.

Any cloud provider deploying Nekomata as a hosted service MUST:

  1. Release all modifications under AGPL-3.0
  2. Provide users a prominent link to modified source
  3. Not remove attribution to SnapKitty Collective

Credits

╔══════════════════════════════════════════════════════════════╗
β•‘                                                              β•‘
β•‘  Architecture    Ahmad Ali Parr                              β•‘
β•‘  Engineering     SnapKitty Collective                        β•‘
β•‘  Model Avatar    Nekomata β€” the two-tailed sovereign cat     β•‘
β•‘  Execution       WORM-sealed, receipt-generating             β•‘
β•‘                                                              β•‘
β•‘  "Infrastructure that rewrites itself."                      β•‘
β•‘                                                              β•‘
β•šβ•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•β•

Nekomata
猫又 β€” The two-tailed cat that never dies.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Space using Snapkitty/nekomata 1

Collection including Snapkitty/nekomata