Proposal: Report a Public Workflow

[yangzhang75]

Proposal: Report a Public Workflow

What it does

Texera lets any user make a workflow public (workflow.is_public), after which it shows up on the public Hub where everyone can browse, view, like, and clone it. Today there is no way for a viewer to flag a public workflow that contains inappropriate content — offensive text in operator/comment fields, misleading or harmful pipelines, spam, copyrighted data, etc. The only recourse is to email a maintainer out of band, and admins have no in-product action to take a bad workflow down.

This proposal adds a lightweight "Report" flow: any logged-in user can report a public workflow with a reason, and admins get a moderation queue where they can review reports and, if warranted, unpublish it.

flowchart LR
    A[User sees a public<br/>workflow on the Hub] -->|clicks Report,<br/>picks a reason| B[(Report saved)]
    B --> C{Admin reviews<br/>the report}
    C -->|looks fine| D[Dismiss]
    C -->|inappropriate| E[Unpublish<br/>removed from Hub]

Loading

Use case: A public workflow's operator labels contain harassing language toward a named person. A viewer clicks Report → "Harassment / offensive content". An admin sees it in the queue, confirms, and unpublishes it with one click — the workflow is no longer discoverable on the Hub, without deleting the author's private copy.

How it works

Reporting (any logged-in user)

• On the Hub workflow card and the public-workflow detail view, add a Report action.
• It opens a small dialog: a reason category (dropdown) + an optional free-text detail field.
• Suggested categories: Offensive / hateful content, Harassment, Spam / advertising, Copyright / privacy violation, Other.
• Guardrails: only public workflows are reportable; a user can't report the same workflow twice while an open report exists; the workflow owner can't report their own.

Moderation (admins only)

• A new Reports tab in the existing admin dashboard.
• It lists reports grouped by workflow, sorted by report count / recency, showing: workflow title + owner, number of distinct reporters, reason breakdown, latest detail.
• From a row an admin can:
  • Dismiss → close the workflow's open reports, no action taken.
  • Unpublish → set the workflow back to private (reuses the existing public/private mechanism) so it disappears from the Hub. The author keeps their private copy.
• Every action is recorded (who acted, when, resolution) for accountability.

Implementation sketch

There is no report/moderation concept in Texera today, so this is greenfield. At a high level it needs: a new table to store reports, a user-facing endpoint to submit a report, admin-only endpoints to list and resolve them (mirroring the existing admin resources), and a bit of frontend — a report dialog on the Hub plus a Reports tab in the admin dashboard. "Unpublish" reuses the workflow's existing public/private toggle, so no new take-down machinery is required. Exact schema and API can be worked out in the PR.

Open questions

1. Take-down semantics — should "Unpublish" (keep author's private copy, my default) be the only admin action, or do we also want hard-delete for egregious cases?
2. Anonymity & abuse of reporting — reporter identity is stored for accountability, but should it be visible to admins only? Do we need a per-user report rate limit to prevent report spam?
3. Author notification — should the author be told their workflow was unpublished, and why? In scope or a follow-up?
4. Scope — start with workflows only, or design it generically so datasets (also public via the Hub) can be reported with the same mechanism later?
5. Thresholds — purely manual admin review, or auto-unpublish once N distinct users report the same workflow?

[chenlica]

1. For now I suggest we allow the admin to un-publish it, without deleting it.
2. I suggest we make the author visible to admins only. No need to set a limit as users are presumably trustable.
3. Authors should be notified.
4. Let's focus on workflows for now. It shouldn't be hard to generalize it to dataset and other resources in the future.
5. We can use a threshold, and set it to "1" initially.