Quickstart

from evolverl.evolution import Evolution, EvolutionConfig
from evolverl.llm import LLMConfig
from evolverl.agent import Agent, AgentConfig

# Configure LLM backend
llm_config = LLMConfig(
    model_name="gpt-4o-mini",
    model_type="openai",  # or "anthropic"
    openai_api_key="your-api-key"  # or anthropic_api_key for Claude
)

# Create agent with system prompt
agent_config = AgentConfig(llm_config=llm_config)
agent = Agent(agent_config)
agent.set_default_prompt("""You are an expert AI agent specialized in mathematics.
You break down complex problems step by step and show your work clearly.""")

# Configure evolution process
config = EvolutionConfig(
    population_size=5,
    generations=10,
    mutation_rate=0.1,
    crossover_rate=0.8,
    output_dir="agents"
)

# Create evolution instance
evolution = Evolution(config, experiment_id="math_solver")

# Run evolution process
await evolution.evolve(
    domain="mathematics",
    description="Solve complex math problems with detailed explanations"
)

# Basic usage with OpenAI
python train_agent.py --domain math --description "Solve math problems" -v

# Use Anthropic's Claude
python train_agent.py --provider anthropic --domain math --description "Solve math problems"

# Load domain from file
python train_agent.py --domain-file domains/math_solver.json

# Custom output directory
python train_agent.py --domain math --description "..." --output-dir ./my_agents

# Increase verbosity (up to -vvvvv)
python train_agent.py --domain math --description "..." -vvv

agents/
├── {experiment_id}_gen0.json           # Best agent from generation 0
├── {experiment_id}_gen0_full.json      # All variants from generation 0
├── {experiment_id}_gen1.json           # Best agent from generation 1
├── {experiment_id}_gen1_full.json      # All variants from generation 1
└── {experiment_id}_best.json           # Best agent overall