# Robotaxi Price Index Public Dataset

Version 1.0.1; data updated 2026-09-16.

This release contains 14 deduplicated observations and 21 cited sources. It is generated deterministically from the site's `data/*.json` files by `export_dataset.py` using only Python's standard library.

## Files

- `observations.csv`: one row per deduplicated observation.
- `sources.csv`: one row per unique source URL.
- `observation_sources.csv`: many-to-many citation relationships and `source_role`.
- `robotaxi-price-index.json`: complete JSON form containing metadata and all three tables.
- `field-dictionary.json`: field definitions and keys.
- `dataset-metadata.json`: release-level metadata.
- `LICENSE-DATA.txt`, `NOTICE.txt`, and `CITATION.cff`: reuse, exclusions, and citation guidance.

## Deduplication and dates

Known duplicate representations in `fares.json` and `cities.json` share a single stable `observation_id`; all contributing JSON paths remain in `source_data_paths`. Dates are never padded with invented month/day values: `date_precision` records whether the supplied date is a year, month, or day. `source_role` identifies lead versus supporting evidence and does not change the separate editorial `source_tier`.

## Rebuild locally

Run `python export_dataset.py`, then `python -m unittest discover -s tests -v`. Generated bytes are stable while source JSON and exporter code are unchanged.

## Scope

Observations are dated evidence snapshots, not live quotes. Single trips are not rate cards; estimates and targets are not measured current costs. Follow source URLs and verify operator apps before using fare information.

## Public mirrors

- Interactive dataset: https://robotaxiprice.com/dataset/
- GitHub repository: https://github.com/leown/robotaxi-price-data
- Kaggle dataset: https://www.kaggle.com/datasets/leowatsonleo/robotaxi-fares-and-cost-benchmarks
- Hugging Face dataset: https://huggingface.co/datasets/wojiushiwn/robotaxi-price-index
- Zenodo archive (DOI): https://doi.org/10.5281/zenodo.21701387

## Citation

See `CITATION.cff`. The archival record is versioned on Zenodo: DOI 10.5281/zenodo.21701387 always resolves to the latest version; 10.5281/zenodo.21701388 is the permanent record for version 1.0.1 specifically.

Version 1.0.1 adds public archive metadata and mirror links; observation values are unchanged from 1.0.0.