Save Logs

Scenario

You want a human-readable record of one evaluation run that you can inspect later or attach to an experiment artifact.

Runnable example

from pydantic import BaseModel

from extraction_testing import (
    FeatureRule,
    RunConfig,
    RunLogger,
    TaskType,
    build_run_context,
    evaluate,
)


class ArticleLabel(BaseModel):
    row_identifier: int
    topic_label: str


predicted_records = [ArticleLabel(row_identifier=1, topic_label="business")]
gold_records = [ArticleLabel(row_identifier=1, topic_label="business")]

run_config = RunConfig(
    task_type=TaskType.SINGLE_FEATURE,
    feature_rules=[FeatureRule(feature_name="topic_label", feature_type="category")],
    index_key_name="row_identifier",
    log_directory_path="./logs",
)

result_bundle = evaluate(predicted_records, gold_records, run_config)
run_context = build_run_context(run_config)

logger = RunLogger(run_config.log_directory_path)
log_path = logger.write_log(
    run_context,
    result_bundle,
    run_config,
    note_message="Saved from save-logs guide",
)

print(log_path)

What to expect

• the ./logs directory is created automatically if it does not exist
• the file name follows the pattern test_run_<run_identifier>.txt
• the file includes total metrics, per-feature metrics, row accuracy, and metadata about the run

When to use build_run_context()

Use build_run_context(run_config) when you want:

• a timestamp-based run identifier
• a configuration hash for reproducibility
• the metadata expected by RunLogger.write_log()

Related concepts

• Logging
• Run Config