Skip to content

Configuration Reference

This section provides exhaustive reference documentation for all configuration objects in karenina. For tutorial-style introductions with examples, see Configuration and Running Verification.

Reference Pages

Page What It Covers Fields
Environment Variables All environment variables recognized by karenina 15+ vars
VerificationConfig Complete verification pipeline configuration 33 fields
ModelConfig Model identity, parameters, MCP, and middleware 19 fields
PromptConfig Custom instruction injection into pipeline LLM calls 6 fields
Preset Schema JSON format for saved configuration presets Metadata + config
DBConfig Database connection for auto-saving results 8 fields

Quick Lookup

Which reference page do I need?

I want to... See
Set an API key or path Environment Variables
Configure answering or parsing models ModelConfig
Choose evaluation mode or feature flags VerificationConfig
Customize LLM prompts for parsing or rubric evaluation PromptConfig
Understand preset file format Preset Schema
Configure MCP servers or agent middleware ModelConfig
Set up async execution VerificationConfig or Environment Variables
Enable deep judgment or embedding checks VerificationConfig

Configuration Hierarchy

Configuration values are resolved in this order (highest priority first):

CLI arguments / explicit Python args
        ↓ (override)
Preset values
        ↓ (override)
Environment variables
        ↓ (override)
Built-in defaults

See Configuration Hierarchy for details.

Import Paths

from karenina.schemas import VerificationConfig, ModelConfig
from karenina.schemas.verification import PromptConfig