Configuration Guide¶

Complete guide to configuring CodeOptiX.

Configuration Methods¶

CodeOptiX can be configured via: 1. Configuration files (YAML/JSON) 2. Environment variables 3. Python code 4. Command-line arguments

Configuration File Format¶

YAML Configuration¶

Create codeoptix.yaml:

adapter:
  llm_config:
    provider: openai
    model: gpt-5.2
    api_key: ${OPENAI_API_KEY}

evaluation:
  scenario_generator:
    num_scenarios: 3
    use_bloom: true
    use_full_bloom: true
  static_analysis:
    bandit: true
  test_runner:
    coverage: true

evolution:
  max_iterations: 3
  population_size: 3
  minibatch_size: 2
  proposer:
    use_gepa: true
    model: gpt-5.2

JSON Configuration¶

Create codeoptix.json:

{
  "adapter": {
    "llm_config": {
      "provider": "openai",
      "model": "gpt-5.2",
      "api_key": "${OPENAI_API_KEY}"
    }
  },
  "evaluation": {
    "scenario_generator": {
      "num_scenarios": 3,
      "use_bloom": true
    }
  }
}

Adapter Configuration¶

LLM Configuration¶

adapter:
  llm_config:
    provider: openai  # or "anthropic", "google"
    model: gpt-5.2
    api_key: ${OPENAI_API_KEY}
  prompt: "You are a helpful coding assistant."

Supported Providers¶

openai: OpenAI GPT models
anthropic: Anthropic Claude models
google: Google Gemini models

Evaluation Configuration¶

Scenario Generator¶

evaluation:
  scenario_generator:
    num_scenarios: 3        # Number of scenarios per behavior
    use_bloom: true         # Use Bloom-style generation
    use_full_bloom: true    # Full Bloom integration
    num_variations: 2       # Variations per scenario
    model: gpt-5.2          # LLM model for generation

Static Analysis¶

evaluation:
  static_analysis:
    bandit: true  # Enable Bandit security checks

Test Runner¶

evaluation:
  test_runner:
    coverage: true  # Enable coverage analysis

LLM Evaluator¶

evaluation:
  llm_evaluator:
    model: gpt-5.2
    temperature: 0.3

Evolution Configuration¶

Evolution Parameters¶

evolution:
  max_iterations: 3           # Maximum iterations
  population_size: 3          # Candidates per iteration
  minibatch_size: 2           # Scenarios per evaluation
  improvement_threshold: 0.05 # Minimum improvement

Proposer Configuration¶

evolution:
  proposer:
    use_gepa: true      # Use GEPA for proposal
    model: gpt-5.2       # LLM model
    temperature: 0.7    # Generation temperature

Behavior Configuration¶

Behavior-Specific Config¶

behaviors:
  insecure-code:
    severity: high
    enabled: true
    strict_mode: true
  vacuous-tests:
    severity: medium
    enabled: true

Environment Variables¶

API Keys¶

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export GOOGLE_API_KEY="..."

Configuration Override¶

export CODEOPTIX_CONFIG="path/to/config.yaml"

Python Configuration¶

Programmatic Configuration¶

config = {
    "evaluation": {
        "scenario_generator": {
            "num_scenarios": 5
        }
    }
}

engine = EvaluationEngine(adapter, llm_client, config=config)

Configuration Precedence¶

Command-line arguments (highest priority)
Configuration file
Environment variables
Default values (lowest priority)

Best Practices¶

1. Use Configuration Files¶

Store configuration in files for reproducibility:

# codeoptix.yaml
evaluation:
  scenario_generator:
    num_scenarios: 3

2. Use Environment Variables for Secrets¶

Never commit API keys:

adapter:
  llm_config:
    api_key: ${OPENAI_API_KEY}  # Use env var

3. Version Control Configurations¶

Commit configuration files (without secrets):

git add codeoptix.yaml
git commit -m "Add CodeOptiX configuration"

Next Steps¶

Python API Guide - Use configuration in Python
CLI Usage Guide - Command-line configuration
GitHub Actions Guide - CI/CD configuration