agent-generations — Agent Version Evolution

Track agent lineage, versioning, and generational evolution. Agents breed, mutate, and evolve.

What This Gives You

• Generation tracking — version agents with parent-child relationships and generation numbers
• Lineage trees — build and query full ancestry trees for any agent
• Mutation engine — apply controlled mutations (config changes, capability adjustments) between generations
• Evolution management — orchestrate generation transitions with validation and rollback
• Breeding — combine traits from two parent agents into offspring configurations

Quick Start

pip install agent-generations

from agent_generations import Generation, LineageTree, EvolutionManager, Breeding

# Create generations
gen1 = Generation(version=1, agent_id="agent-base", traits={"speed": 0.7, "accuracy": 0.9})
gen2 = Generation(version=2, agent_id="agent-base", traits={"speed": 0.8, "accuracy": 0.9}, parent=gen1)

# Build lineage
tree = LineageTree()
tree.add(gen1)
tree.add(gen2)
ancestry = tree.trace("agent-base")  # [gen2, gen1]

# Breed two agents
offspring = Breeding().breed(
    parent_a={"speed": 0.9, "accuracy": 0.7},
    parent_b={"speed": 0.5, "accuracy": 0.95},
    strategy="average",
)
# {"speed": 0.7, "accuracy": 0.825}

# Manage evolution
manager = EvolutionManager()
manager.register("agent-base", gen1)
evolved = manager.evolve("agent-base", mutations={"speed": +0.1})

API Reference

Generation(version, agent_id, traits, parent=None)

LineageTree — add(generation), trace(agent_id), roots(), descendants(agent_id)

MutationEngine — apply(generation, mutations) → Generation

EvolutionManager — register(agent_id, generation), evolve(agent_id, mutations), rollback(agent_id, version)

Breeding — breed(parent_a, parent_b, strategy) → dict

How It Fits

The evolution layer for the SuperInstance fleet. Agents improve across generations, with full lineage tracking.

• agent-tattoo — Tattoos carry forward across generations
• agent-resume — Resumes reflect generational capabilities
• become-ai — Self-evolving agent platform

Testing

pytest tests/

Installation

pip install agent-generations

Python 3.10+. MIT license.