Codex AutoRAG Metric Plugin for Retrieval Precision Fixtures
Implement an AutoRAG-Research metric using MetricInput and its decorators. Verify single-item or batch scoring with known inputs.
中文Install in Codex
Codex with an AutoRAG-Research checkout, Python, and synthetic retrieval data with expected scores.
- Open the pinned GitHub original and retain the complete `.agents/skills/create-metric-plugin/` directory in the checkout.
- Open Codex in that checkout and invoke `$create-metric-plugin` with the example task and desired output.
Example task
Invoke `$create-metric-plugin` with two fictional retrieval queries. Give one relevant result and one empty result list, then establish expected precision values.What to check
The tests check the metric against fixed labels and outputs, including empty retrieval results.
Original SKILL.md on GitHub · Updated
Frequently asked question
When should I use the batch decorator?
Choose the batch entry when scoring depends on several inputs. A per-item metric can use the single-input entry.