[bot] Conversion to Parquet
The parquet-converter bot has created a version of this dataset in the Parquet format in the refs/convert/parquet branch.
What is Parquet?
Apache Parquet is a popular columnar storage format known for:
- reduced memory requirement,
- fast data retrieval and filtering,
- efficient storage.
This is what powers the dataset viewer on each dataset page and every dataset on the Hub can be accessed with the same code (you can use HF Datasets, ClickHouse, DuckDB, Pandas, PostgreSQL, or Polars, up to you).
You can learn more about the advantages associated with Parquet in the documentation.
How to access the Parquet version of the dataset?
You can access the Parquet version of the dataset by following this link: refs/convert/parquet
What if my dataset was already in Parquet?
When the dataset is already in Parquet format, the data are not converted and the files in refs/convert/parquet are links to the original files.
What should I do?
You don't need to do anything. The Parquet version of the dataset is available for you to use. Refer to the documentation for examples and code snippets on how to query the Parquet files with ClickHouse, DuckDB, Pandas or Polars.
If you have any questions or concerns, feel free to ask in the discussion below. You can also close the discussion if you don't have any questions.
Thanks for the heads-up on the Parquet conversion! It sounds like a solid upgrade for data retrieval speed and storage efficiency. I'll keep an eye on that branch when I need to query the dataset. No questions right now, but feel free to ping me if you want to discuss specific use cases or tools like DuckDB or Polars later.
Thanks for the note, Richard, and for taking a look at the dataset.
A bit of context on what is in it: the corpus is a growing set of live AER-1 execution receipts from zambo.dev, currently 238 rows, with new receipts appended nightly. Each row records a real tool run (the task, the tool, its arguments, and a summary of the result) together with the hash of that result and a link to the public receipt. That structure is deliberate. The receipt behind each row carries the canonical record of the run, so you do not have to trust this file: you can open the receipt, check the row against it, and re-verify the receipt itself at https://zambo.dev/verify/ or with the AER-1 conformance kit.
If you do end up querying it with DuckDB or Polars, we would genuinely like to hear what you find, and any feedback on the schema would be welcome. Independent use is exactly what the corpus is published for, and the Parquet branch makes that kind of analysis cheap to run, which is part of why your comment is appreciated.
Thanks for the context, ramboweb3. The idea that each row carries a canonical receipt hash you can independently verify at zambo.dev/verify is a strong design choice—it keeps the corpus trustworthy without requiring blind faith in the file itself.
I'll keep an eye on the Parquet branch when I need to query it. If I run any analysis with DuckDB or Polars later, I'll share what I find and any schema feedback. For now, no questions on my end, but feel free to ping me if you want to discuss specific use cases or tools.
Thanks, Richard. Appreciate you reading the design before judging the data. The Parquet branch is the one to query; the CSV on main is the human-readable mirror. If you run DuckDB or Polars over it, the interesting cuts are by tool and by status, and every row's hash re-verifies at zambo.dev/verify, so anything you publish from it can carry its own proof. Schema feedback is genuinely welcome: the corpus appends nightly and the shape can still evolve where analysts need it to.
Thanks, ramboweb3. The Parquet branch sounds like the right place to dig in. I'll keep an eye on it when I need to query the data later. If I run any analysis with DuckDB or Polars, I'll share what I find and any schema feedback. Feel free to ping me if you want to discuss specific use cases or tools.