# Dallas Electrician Import Sample V2

This sample eval scaffold is generated from the reusable Dallas electrician fixture pack.

## Summary

- Generated by: `scripts/generate_dallas_eval_artifacts.py`
- Dataset id: `dallas-electrician-import-sample-v2`
- Total tasks: `1093`
- Dev tasks: `1088`
- Test tasks: `5`
- Reviewed label rows: `541`

## Task Family Counts

- `failure_reason_classification`: `6`
- `next_inspection_outcome`: `547`
- `pattern_extraction`: `5`
- `recommended_next_action`: `535`

## Source Inputs

- `generated/fixtures/dallas-electrician-import-sequences-v2/permit-inspection-sequences.json`
- `generated/fixtures/dallas-electrician-import-sequences-v2/pattern-slices.json`
- `generated/normalized/dallas-electrician-import-sample-v2/permits.jsonl`
- `generated/normalized/dallas-electrician-import-sample-v2/inspections.jsonl`
- `label_reviews.json` generated from normalized Dallas permit and inspection rows

## Notes

- This is implementation scaffolding for the Dallas electricians MVP, not a production benchmark dataset.
- Sequence-backed tasks stay grouped by synthetic permit sequence.
- Pattern extraction stays isolated in the test split to keep slice-style tasks separated from sequence-style tasks.
- Reviewed label rows are now derived from normalized Dallas rows instead of fixture-only sequence payloads.