Actor Run Dataset Reconciler

How to use Actor Run Dataset Reconciler

A succeeded run label does not establish a complete dataset handoff. Compare the run metadata with the exported item inventory and keep ambiguous run attribution visible.

1. Enter authorized run metadata and the item-ID inventory from the corresponding dataset export.
2. Inspect physical row counts, distinct IDs, shared-dataset attribution and orphan items.
3. Verify the actual actor logs and dataset before delivering results; no actor or API runs here.

Good to know

• This is run/dataset metadata reconciliation, distinct from pagination checking or marketplace payout accounting.
• A header-only item export proves only zero supplied rows. Shared datasets cannot be attributed to one run, and succeeded labels remain unverified.
• Each input field is limited to 8,000 UTF-16 characters and each exact-header CSV to 100 data rows. Quoted commas/newlines and doubled quotes are supported. Unsupported controls and unpaired surrogates are refused. Decimal magnitudes are at most 999,999,999 with a smallest nonzero magnitude of 0.000000001; exponents are refused and whole counts are checked exactly. Rational results show at most 18 fractional digits and an ellipsis when more remain.
• Editable Markdown and long-form CSV both retain all summary metrics, context, detailed findings, entered source rows and scope limitations. CSV columns are Record type,Section,Source row,Field,Value; spreadsheet-looking text is apostrophe-prefixed. These are review artifacts, not proof of source truth or a completed real-world task.
• Runs deterministically in this browser. No AI, API, vendor connection, upload, outreach, calendar change, payment, storage or background monitoring. Use authorized pseudonymous records; inspect exports before sharing.