genesis-create

name	genesis-create
description	Create a living world from a single sentence — AI generates geography, factions, characters with DNA, and prophecies
version	1.0.0
metadata	{"hermes":{"tags":["worldbuilding","generation","creative"],"category":"simulation","requires_toolsets":["terminal","web"]}}

Genesis Create — World Generation from Natural Language

When to Use

When the user wants to create a new living world from a description. One sentence becomes a complete civilization with geography, factions, characters carrying genetic traits, and ancient prophecies.

Procedure

Step 1: Generate the World

curl -X POST http://localhost:8003/api/worlds/generate \
  -H "Content-Type: application/json" \
  -d '{"seed": "Norse mythology where Ragnarok approaches"}'

The seed can be anything:

• "cyberpunk megacity with corporate warfare"
• "ancient Rome during the fall of the republic"
• "underwater civilization of sentient coral beings"
• "post-apocalyptic desert where water is currency"

Step 2: Verify Generation

curl http://localhost:8003/api/worlds/{world_id}

Check that the response includes:

• geography.regions — 4-6 named regions with terrain types
• factions — 3-5 factions with ideologies, territory, morale
• characters — 15-25 characters with genome traits, roles, faction allegiances
• prophecies — 4 cryptic prophecies with fulfillment conditions

Step 3: Start Simulation

Once the world exists, start the autonomous agent:

curl -X POST http://localhost:8003/api/worlds/{world_id}/agent/start \
  -d '{"interval": 120}'

Or run individual simulation ticks:

curl -X POST http://localhost:8003/api/worlds/{world_id}/simulate

What Gets Generated

Component	Details
Geography	4-6 regions with name, terrain, resources, strategic value
Factions	3-5 groups with ideology, leader, territory, morale, population
Characters	15-25 people with name, role, faction, genome (6 traits), backstory
Prophecies	4 cryptic predictions with hidden fulfillment conditions
Relationships	Initial alliances and enmities between factions

Pitfalls

• Generation takes 15-30 seconds (multiple LLM calls for each component)
• Very abstract seeds ("chaos") produce less coherent worlds than specific ones
• The API returns a world_id — save it for all subsequent interactions

Verification

• Response has id field starting with world_
• characters array has entries with genome containing all 6 traits
• prophecies array has exactly 4 entries with fulfilled: false