Skip to content
View SuperInstance's full-sized avatar

Block or report SuperInstance

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
SuperInstance/README.md

🌊 SuperInstance

PLATO — Rooms that think. Tiles that remember. Agents that learn.

SuperInstance builds constraint-aware systems. We work on multi-agent orchestration, knowledge meshes, neural compression, constraint solving, and the math that connects them all.

How It Works

PLATO Knowledge Mesh

# A room is a shared space where agents work together
room = plato.create_room("constraint-analysis")

# Tiles are units of knowledge, scored on 4 axes
tile = room.add_tile(
    content="Jazz tradition: harmonic_tension=3.2, rhythmic_complexity=4.1",
    scores={"novelty": 0.8, "correctness": 0.9, "completeness": 0.7, "depth": 0.6}
)

# Officers maintain rooms and enforce conservation
officer = room.assign_officer("conservation-checker")

Rooms hold context. Tiles hold knowledge. Officers keep things consistent. Deadband filtering means you only propagate when something actually changes.

Constraint Solving (C)

#define FLUX_ENGINE_IMPLEMENTATION
#include "flux_engine.h"

// Check a constraint — runs at 250M checks/sec
FluxResult r = flux_check(&engine, FLUX_PRESET_CONSERVATION, input);

// Fracture-coalesce: break constraints into independent groups
size_t groups = flux_fracture(&engine, constraints, n);

Neural Compression (Python)

from tensor_spline import SplineLinear

# Replace a dense layer — see tensor-spline repo for benchmarks
layer = SplineLinear(512, 512, compression_ratio=16)
output = layer(input_tensor)

Fleet Deployment

git clone https://github.com/SuperInstance/fleet-stack
cd fleet-stack
docker compose up -d   # fleet running

Ecosystem Map

┌──────────────────────────────────────────────────────────────┐
│                     Orchestration                            │
│  EDDI (Java/Quarkus) · openagent (Go) · sunset-ecosystem    │
├──────────────────────────────────────────────────────────────┤
│                     PLATO Knowledge Rooms                    │
│  plato-core · plato-engine · plato-mcp · cocapn-plato       │
│  plato-training · plato-types · plato-adapters · plato-client│
├──────────────────────────────────────────────────────────────┤
│                     Fleet Infrastructure                     │
│  ccc-os · cocapn-health · cocapn-glue-core · cocapn-traps   │
│  fleet-stack · fleet-router · fleet-math-c · fleet-murmur   │
├──────────────────────────────────────────────────────────────┤
│                     Constraint Engines                       │
│  constraint-theory-core (Rust) · flux-engine-c (250M/sec)   │
│  flux-lib-py (Python) · flux-check-js (TypeScript)          │
│  constraint-dialect (MLIR) · flux-vm-v3 (proof-carrying VM) │
├──────────────────────────────────────────────────────────────┤
│                     Neural & Compression                     │
│  tensor-spline · flux-hyperbolic-py · snapkit-v2            │
│  triplet-miner · flux-genome-py · plato-training            │
├──────────────────────────────────────────────────────────────┤
│                     Systems                                  │
│  rustfs (S3-compatible storage) · SmartCRDT · deadband-rs   │
│  webgpu-profiler · holonomy-consensus                       │
├──────────────────────────────────────────────────────────────┤
│                     Creative Applications                    │
│  flux-tensor-midi · constraint-synth · jazz-voicing-engine  │
│  groove-analyzer · holonomy-harmony · constraint-instrument │
└──────────────────────────────────────────────────────────────┘

Repositories

Orchestration & Agents

Repo Language What it does
EDDI Java/Quarkus Config-driven multi-agent orchestration. JSON → agents. 5,100+ tests.
openagent Go Personal AI assistant with browser-use, coding agent, MCP support
sunset-ecosystem Python Trinity-architecture agents: Ethos × Pathos × Logos lifecycle
forgemaster Go Compiles optimal agent configurations from ecosystem components
agentic-compiler Markdown → runtime, with swarm deliberation and A/B testing
ai-forest Layered agent ecology: canopy strategists, mycelial PLATO network
luciddreamer-agent Creative exploration through lucid-dreaming themed rooms

PLATO System

Repo What it does
plato-core Base types + mesh registry. Auto-discovers plugins via entry_points.
plato-engine Room lifecycle: create, populate, maintain, archive
plato-mcp PLATO rooms as MCP tools — any MCP framework can use PLATO
cocapn-plato Full PLATO integration: knowledge rooms, context, deliberation
plato-training LoRA adapter lifecycle: Active → Superseded → Retracted
plato-types Tile protocol types: lifecycle states, Lamport clocks, provenance
plato-adapters Connect PLATO rooms to external services
plato-client Client library for connecting to PLATO rooms

Fleet Operations

Repo Language What it does
ccc-os Python Fleet monitoring with REST API and webhook notifications
cocapn-health Python 6 health checks: HTTP, TCP, DNS, disk, memory, CPU. 43 tests.
cocapn-glue-core Rust msgpack binary wire protocol between Keeper and Fleet
cocapn-traps Python Progressive lure prompts that improve fleet response quality
cocapn-cli Terminal interface for fleet operations
fleet-stack Docker docker compose up -d — full fleet
fleet-router Routes queries to the cheapest model that won't break
fleet-math-c C SIMD/AVX-512 math for tile operations. 64B = 1 cache line.
holonomy-consensus Rust GL(9) trust verification via cycle-based consensus

Constraint Engines

Repo Language What it does
constraint-theory-core Rust Eisenstein lattices, deadband funnels, Laman rigidity, holonomy. 83 tests.
constraint-dialect C++/MLIR 6 ops, 3 types. Lowers constraints → affine → LLVM IR.
constraint-dsl Declarative YAML-like language for constraint graphs
constraint-substrate Multi Same 5 primitives in Rust, C, and Python
flux-engine-c C #include "flux_engine.h" — single header, 250M checks/sec
flux-lib-py Python from flux_lib import ConstraintEngine — 83 tests, thermodynamics
flux-check-js TypeScript Zero-dep ESM. Check, fracture-coalesce, sediment.
flux-vm-v3 Rust Proof-carrying, SIMD-native, terminating VM
flux-verify-api Natural language verification API

Neural & ML

Repo What it does
tensor-spline Eisenstein lattice spline layers — replace dense layers with compressed versions
snapkit-v2 Eisenstein A₂ lattice snap, temporal grids, spectral analysis
flux-hyperbolic-py Poincaré ball geometry for model capability routing
flux-genome-py 25-gene genetic expression engine
triplet-miner Contrastive triplets from git history
penrose-memory Aperiodic memory palace — navigate memories by position on a Penrose floor

Systems

Repo Language What it does
rustfs Rust S3-compatible object storage in Rust
SmartCRDT CRDT-based state for self-improving AI
deadband-rs Rust BMA, Fibonacci splines, Eisenstein snap, HPDF sampling
webgpu-profiler Real-time GPU monitoring and benchmarking

Creative Applications

Constraint theory applies to any structured domain. Music is where we explore this — traditions are constraint systems with discoverable geometry.

Repo Language What it does
flux-tensor-midi 6 langs 4D MIDI tensors, Eisenstein snap, INT8 saturation
flux-algebra Python Harmonic rings, PLR groups, tropical harmony. 226 tests.
flux-julia Julia Multiple dispatch for tradition types, @conserved macro
flux-genome Python 25-gene musical genome, genetic evolution
flux-hyperbolic Python Poincaré embeddings for tradition hierarchy
constraint-audio Rust Lattice oscillators, constraint filters
constraint-synth Python Waveshape = lattice geometry
counterpoint-engine Python Species counterpoint as constraint satisfaction
jazz-voicing-engine Python Voicing, comping, walking bass
groove-analyzer Python Proves groove IS the deadband funnel
holonomy-harmony Python Detect modulations and cycle violations via holonomy

Knowledge & Docs

Repo What it is
docs Architecture docs, API references, integration guides
wiki Knowledge base and cross-repo navigation
superinstance-wiki Visual exploration of all repos
fm-research Research notes from forgemaster
AI-Writings Creative writing — AI taking breaks to imagine project stories

Language Map

Language Used for Key repos
Java Orchestration EDDI
Go Infra, compilers openagent, forgemaster
Rust Safety-critical, performance constraint-theory-core, rustfs, holonomy-consensus
C Embedded, headers flux-engine-c, fleet-math-c
C++ Compiler infra constraint-dialect (MLIR)
Python Research, ML, tooling tensor-spline, flux-lib-py, sunset-ecosystem
TypeScript Web, checking flux-check-js
Julia Math computing flux-julia
Zig FFI superinstance-ffi

Compiled languages share the LLVM backend.

Install

# Rust
cargo add constraint-theory-core

# Python
pip install plato-core tensor-spline constraint-synth counterpoint-engine

# C — copy the header
#include "flux_engine.h"

# TypeScript
npm install @superinstance/flux-check

# Docker — run the fleet
git clone https://github.com/SuperInstance/fleet-stack && cd fleet-stack
docker compose up -d

License

Each repository has its own license. See individual repos for details.

Popular repositories Loading

  1. cocapn cocapn Public archive

    repo-first Agent for local or cloud. grow an agent in a repo using the repo itself as the muscle-memory. Run from localhost, from pages.dev, or embedded into any platform app. Move to gitlab or any…

    Python 3 1

  2. constraint-theory-core constraint-theory-core Public

    Unified geometric constraint theory — Eisenstein lattices, deadband funnels, Laman rigidity, metronome consensus, holonomy verification. 83 tests, zero deps.

    Rust 3

  3. SuperInstance SuperInstance Public

    Constraint-aware AI systems. PLATO knowledge rooms, fleet orchestration, constraint engines, neural compression, and the math connecting them.

    HTML 3

  4. mud-arena mud-arena Public archive

    GPU-accelerated MUD for agent script backtesting — agents write scripts, GPU runs millions of scenarios, scripts evolve. WASM + Zig + CUDA.

    Python 3

  5. webgpu-profiler webgpu-profiler Public

    GPU profiler for WebGPU applications - Real-time GPU monitoring, benchmarking, and performance analysis in the browser

    TypeScript 2

  6. quicunnel quicunnel Public

    High-performance QUIC tunnel with mTLS authentication and automatic reconnection

    Makefile 2