Changelog

1. #Unreleased

   First public release

   Nothing has shipped publicly yet. This is what the first release will contain — see the project README for what is built, what is still being wired up, and what is not built at all.

   • AddedOpen dataset pools anyone can browse without an account, and a contribution form whose fields come from the dataset type's own contract rather than a fixed shape.
   • AddedAn automated validation pipeline on every submission: exact and near-duplicate detection, sandboxed execution, and optional LLM review. Checks that are unconfigured or unsupported say so and fail closed — they are never reported as passed.
   • AddedHuman audit windows with exclusive 24-hour validator claims, conflict-of-interest blocking so nobody reviews their own work, and automatic release back to the queue on timeout.
   • AddedKarma for accepted work — awarded, held through the dispute window, then released — shown on a public contributor profile, with an automatic badge catalog.
   • AddedSponsor dataset requests with admin review and approval before a pool goes live.
   • AddedWeekly activity charts in the contributor and validator workspaces: submission outcomes, karma earned, audit decisions, and what became of the items you flagged. Counted from your own rows over a window you choose, with a table view of the same numbers and an explicit note wherever a series is grouped by the week you submitted rather than by when each decision was made.
   • AddedAn MCP server for agent clients: JSON-RPC, OAuth 2.1 with PKCE, API-key auth, per-credential rate limiting.
   • AddedAn operator console for reviewing requests, authoring dataset types and harnesses, monitoring submissions and validators, and configuring platform settings.
   • AddedNotification delivery with a real outbox and dispatch worker. An unconfigured channel dead-letters with the actual reason instead of reporting a message as sent.
   • AddedSelf-hosting via Docker: a Dockerfile per app and a docker-compose.yml that brings up Postgres, migrations, catalog seed, all four apps, and the background worker.

Building on DataBounty? Follow this page for changes that affect datasets, karma, and the validation pipeline.