Governance

Governance

Most community backends ship content primitives and leave governance to you. Agora makes it core surface area — endpoints, schema, and admin UI in the box, enforced at the server trust boundary. There are two distinct layers: moderation judges content; stewardship tends people and relationships.

Moderation

• Report queues for entities, comments, and chat messages, with resolution.
• Server-enforced removed-content hiding — a removed row is omitted from every list, 404'd on single reads, and filtered inside the semantic-search RPC; operators bypass to review. See Security.
• Roles — space-scoped moderator roles plus a project-wide operator god-view, and DB-backed per-project owner/admin grants.
• AI Agent Moderator — the Python services/scorer subsystem assesses all new content: two RoBERTa classifiers gate a Claude Haiku adjudication, and content is either auto-hidden or flagged for human review depending on the score (configurable categories, confidence thresholds, and auto-actions, tuned per-project in admin Settings). A moderation_analyses audit trail and an operator-gated AI-flag queue back the admin AI tab.

The scorer architecture (pipeline, write-back, resource sizing) is documented in docs/SCORER.md and docs/SCORER-REQUIREMENTS.md.

Stewardship

A distinct conflict-resolution layer — restorative, not punitive — powering the admin Steward tab.

• A DB-granted steward role between member and operator (route-scoped to the caseload; stewards do not inherit the operator's global read bypass).
• A caseload that moves a dispute (complainant ↔ respondent over some content) through open → in_mediation → closed with transformative-ordered outcomes (repair → separation → protection → escalation).
• An asymmetry / "targeting" flag for power-aware, anti-false-balance handling.
• Private mediation channels built on the existing chat — a 1:1 caucus with each party, or a consensual joint room (never for a targeting case).
• Configurable participant notifications (power-aware / symmetric / resolution-only) that keep parties informed without ever leaking who raised a case.
• An append-only case timeline, and escalate-to-removal that takes the subject content down through the normal moderation path.

Full design: docs/STEWARDSHIP.md.

Both layers live behind the same /v7/:projectId/... contract and in the Postgres schema, so they're available to every client from day one.