Configuration

Configuration options for this package.

Overview

Configuration is loaded from YAML, environment variables, or programmatically via EvaluationConfig.

Basic Example

from lexigram.ai.evaluation.config import EvaluationConfig

config = EvaluationConfig(
    enabled=True,
    default_threshold=0.8,
    embedding_model="text-embedding-3-small",
)
EvaluationModule.configure(config)

Options

Option	Type	Default	Description
`enabled`	bool	`True`	Enable the evaluation subsystem
`default_threshold`	float	`0.8`	Default pass threshold
`embedding_model`	str	`text-embedding-3-small`	Model for embeddings
`include_metadata`	bool	`True`	Include metadata in reports
`max_samples`	int	`None`	Max samples per run
`max_retries`	int	`3`	Max evaluation retries
`timeout_seconds`	int	`30`	Execution timeout

Environment Variables

Variable	Description
`LEX_AI_EVALUATION__ENABLED`	Enable/disable subsystem
`LEX_AI_EVALUATION__DEFAULT_THRESHOLD`	Default threshold
`LEX_AI_EVALUATION__EMBEDDING_MODEL`	Embedding model
`LEX_AI_EVALUATION__INCLUDE_METADATA`	Include metadata
`LEX_AI_EVALUATION__MAX_SAMPLES`	Max samples
`LEX_AI_EVALUATION__MAX_RETRIES`	Max retries
`LEX_AI_EVALUATION__TIMEOUT_SECONDS`	Timeout

Advanced Configuration

from lexigram.ai.evaluation.config import EvaluationConfig

config = EvaluationConfig(
    enabled=True,
    default_threshold=0.95,
    embedding_model="text-embedding-3-large",
    max_samples=100,
    max_retries=5,
    timeout_seconds=60,
)

Best Practices

set thresholds based on your accuracy requirements
choose embedding model based on your quality/speed needs
configure timeouts appropriately for batch runs