Missions

The colony can be deployed to different problem domains. Each mission is isolated but shares permanent knowledge.

What is a Mission?

A mission is a specific problem the colony is trying to solve:

Mission Structure

# missions/hunt-btc/mission.yaml
name: hunt-btc
description: Find Bitcoin Puzzle #71 private key
status: active

target:
  type: search
  objective: collision_detection
  address: 1PWo3JeB9jrGwfHDNpdGK54CRas7fsVzXU

castes:
  tame_hunter: 50%
  wild_hunter: 50%

schema: hunt_btc.tql

Mission Lifecycle

┌─────────────────────────────────────────────────────────────────────────────┐
│                         MISSION LIFECYCLE                                    │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                              │
│   PROPOSAL ──────► PLANNING ──────► ACTIVE ──────► COMPLETED                │
│       │                │               │               │                     │
│       │                │               │               │                     │
│   Define           Set up          Run ants        Archive                  │
│   problem          schema,         on mission      mission,                 │
│   domain           castes                          crystallize              │
│                                                    patterns                  │
│                                                                              │
└─────────────────────────────────────────────────────────────────────────────┘

Ephemeral vs Permanent Data

Ephemeral (Mission-Specific)

• Pheromone levels on edges
• Traversal events
• Working memory
• Active intentions

Deleted when mission ends.

Permanent (Colony Knowledge)

• Superhighways (pheromone > 20)
• Crystallized patterns
• Proven strategies
• Knowledge embeddings

Survives forever. Transfers to future missions.

Cross-Mission Learning

The magic: patterns discovered in one mission can help others.

hunt-btc discovers "sparse-marker-search"
    → crystallizes to permanent knowledge
        → buy-ads applies to "find cheap inventory"
            → korea-tourism applies to "find hidden gems"

Example Pattern Transfer

Pattern: “When tame hunters find clusters, wild hunters should concentrate there”

Hunt BTC: Concentrate wild kangaroos near distinguished point clusters

Trading: Concentrate harvesters near opportunity clusters

Supply Chain: Concentrate resources near demand clusters

Same abstract pattern. Different concrete domains.

Creating a Mission

1. Define the Domain

# missions/my-mission/mission.yaml
name: my-mission
description: What problem are we solving?
status: planning

target:
  type: optimization
  objective: minimize_cost

castes:
  scout: 40%
  harvester: 40%
  relay: 20%

2. Extend the Schema

# missions/my-mission/schema.tql
define

# Domain entities
my_entity sub concept,
  owns my_attribute;

# Domain relations
my_relation relates source, relates target;

3. Implement Domain Logic

from ants.actors.base import Agent

class MyDomainAnt(Agent):
    async def perceive(self):
        # Domain-specific sensing
        pass

    async def decide(self, perception):
        # Domain-specific decision
        pass

    async def act(self, decision):
        # Domain-specific action
        pass

4. Deploy

/mission new my-mission
/mission activate my-mission

Mission Commands

# List all missions
/mission list

# View mission details
/mission show hunt-btc

# Switch active mission
/mission switch hunt-btc

# Create new mission
/mission new my-mission

# Crystallize patterns (force)
/mission crystallize

# Archive completed mission
/mission archive old-mission

Mission Isolation

Each mission has:

• Separate namespace in TypeDB
• Dedicated pheromone space
• Isolated event logs
• Scoped permissions

But shares:

• Colony actors (ants can work multiple missions)
• Permanent knowledge (patterns transfer)
• Security infrastructure (growth gates apply globally)

Active Mission: Hunt BTC

Currently running:

Join with:

pip install ants-worker
ants-worker join