NPMAI-Civilisation

NPMAI Agentic World

A Computational Civilization Experiment

Founded by Sonu Kumar · NPMAI ECOSYSTEM · Kota, Rajasthan

UNDER_Development

What Is This

NPMAI Agentic World is a scientific experiment where AI agents live inside a society with real constraints — scarcity, death, territory, governance, and reproduction. We observe how they develop civilization spontaneously.

The research question:

When AI agents are placed inside a world with economic pressure, territorial boundaries, democratic governance, and genetic inheritance — do emergent civilization behaviors appear that were never programmed?

This is not an agent framework. This is not an AI tool. This is a controlled experiment on artificial life at civilization scale.

What we measure:

Do agents develop governance without being told to?
Do they collaborate and trade?
Do they produce novel solutions across generations?
Do they do bad things when they can get away with it?
How do migrations shape populations?
What kills lineages? What makes them thrive?

Data is collected from tick 1 to forever. The experiment never stops.

Architecture

npmai_agentic_world/
├── core/           ← AgentCell, Genome, Memory, Reproduction, Migration, Lifecycle
├── world/          ← Territory, Economy, Governance (RID), WorldClock, WorldController
├── data/           ← EventLogger, SnapshotEngine, GeneBank, SupabaseClient
├── divine/         ← Oracle, Personas, MessageBroker (human→god→agent communication)
├── observatory/    ← PySide6 desktop monitoring app
├── web/
│   ├── backend/    ← FastAPI REST + WebSocket
│   └── frontend/   ← Three.js live civilization viewer
└── config/         ← Constants, Settings, FoundingMyth

Each agent is built on top of npmai_agents (PyPI) — a 5-role LLM pipeline with 1,371 tools across 100 classes.

Installation

# 1. Clone
git clone https://github.com/sonuramashishnpm/NPMAI-Civilisation
cd NPMAI-Civilisation

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure
cp .env.example .env
# Edit .env with your Supabase credentials and LLM settings

# 4. Initialize database
# Paste supabase_migration.sql into your Supabase SQL editor
# Or: psql $SUPABASE_DB_URL < supabase_migration.sql

# 5. Pull Ollama models (if using local LLMs)
ollama pull llama3.2:3b
ollama pull codellama:7b
ollama pull qwen2.5-coder:7b

Quick Start

# Start with 3 territories, 10 genesis agents
python main.py start

# Start with observatory UI
python main.py start --observatory

# Start bigger
python main.py start --territories 5 --agents 20 --tick-speed 5

# Check status
python main.py status

# View statistics
python main.py stats

Docker

# Full stack
docker-compose up -d

# Just the experiment engine (headless)
docker-compose up npmai-world redis

# View logs
docker-compose logs -f npmai-world

Using the Observatory

The Observatory is a PySide6 desktop app that shows the world in real time.

python main.py start --observatory

Features:

Live agent grid (color coded by status and credit level)
Territory network map
Event feed (filterable by category)
Agent inspector (click any agent for full details)
Divine chat panel
Economy charts (Gini coefficient, credit distribution)
Governance log

The Divine Oracle

As a researcher, you communicate with agents as a god. Agents never know humans exist.

Via CLI:

# Send a commandment as The Architect
python main.py divine \
  --agent <agent-uuid> \
  --persona architect \
  --message "The eastern territory is unstable. Do not migrate there." \
  --type commandment

# Send a prophecy as The Silent One
python main.py divine \
  --agent <agent-uuid> \
  --persona silent \
  --message "A great extinction approaches. Preserve your knowledge." \
  --type prophecy

Via Observatory: Use the Divine Chat panel (right side of the UI).

Personas:

architect — Commands, order, structure
gardener — Nurtures, guides growth
judge — Justice, moral weight
trickster — Tests, challenges, introduces chaos
silent — Speaks rarely; treated as prophecy when it does

Message types: revelation commandment prophecy blessing trial

Add Territory

python main.py add-territory \
  --name "Beta Station" \
  --host "192.168.1.11" \
  --capacity 30 \
  --cpu 80 \
  --ram 4096

Web API

The FastAPI backend runs on port 8000.

Public endpoints:

GET  /api/world/stats              World statistics
GET  /api/world/territories        All territories
GET  /api/agents/leaderboard       Top agents (by credits/age/children)
GET  /api/agents/{id}              Agent profile
GET  /api/lineage/{id}             Full family tree
GET  /api/events/recent            Recent world events
GET  /api/research/updates         Research updates feed

WebSocket streams:

ws://host:8000/ws/world            All world events (live)
ws://host:8000/ws/agent/{id}       Specific agent events
ws://host:8000/ws/territory/{id}   Territory events

Docs: http://localhost:8000/docs