Admin preview · live evidence moderation queue

Reporter Desk

Production auth/profile foundations now support audited editorial role gates and a live Convex-backed evidence moderation queue. Research automation remains disabled; no AI calls, provider calls, source-ledger mutations, or automated publishing are connected.

Admin role gate

Role-aware admin shell

Checking session

Viewer: No active session

Role: not synced

Role permission summary

Sync profile before evaluating role-based workflow affordances.
Request evidence: blocked — Sync a viewer profile before requesting evidence.
Research automation: always blocked in the manual MVP.

Founder admin bootstrap

Use this once after the founder account has authenticated and synced a profile. The server only promotes the configured founder email, and only while no admin exists.

Bootstrap is only for the configured founder account and closes after the first admin exists.

This role panel is explanatory; action buttons stay controlled by separately tested row-level gates. It keeps assignment/review/publish controls non-mutating and never starts research automation.

Assignment workflow

Admin queue locked

Sign in with a synced contributor, editor, or admin profile to load workflow rows. Manual Convex actions stay hidden while signed out, and research automation remains disabled.

Internal manual dry-runAdmin onlyDefault disabledNo public route trigger

Provider search dry-run

No live provider call is wired from this UI. It only documents the tested server-only boundary for a future manual admin Serper dry-run.

Dry-run query

Requires explicit budget-capped approval before any live Serper verification; use mocked server tests until then.

Evidence moderation queue

Live Convex submissions awaiting human editor review

Human editors review submitted links before they affect topic score, claims, dossiers, or source ledgers.

No AI extraction, crawling, search, or provider calls run from this queue

Queue state: Sign-in required

Editor next step: Sign in with an editor or admin profile, then return to this queue.

Evidence moderation actions stay disabled until sign-in, profile sync, role, status, and note gates pass.

No query, mutation, scoring, provider, source-ledger, automatic moderation, or publish action runs from this guidance.

RR-001

researching

Manual scoping first; provider-backed extraction waits for budget and source-cache guardrails.

Scope approved
Search queries drafted
Evidence intake not yet connected
Claim extraction pending human review

RR-002

researching

Manual source list and human review first; provider-backed synthesis remains gated.

Benchmark source list created
Caveats tagged
Expert commentary needed

Cost ledger

AI calls require budget + cache checks

Topic classification · DeepSeek Flash · estimate: <$0.01 · cache eligible

Source summarization · Gemini Flash 2.0 · estimate: low · must cache by URL/hash

Final trust review · DeepSeek v4/pro · estimate: medium · manual approval required

Publish Checklist

No auto-publishing

Mixed-state checklist: publication is gated even when research looks complete.

Done ✓Sources & Method completeSource ledger exists for the launch dossier.
Done ✓Important claims citedEach visible claim maps to source IDs.
Done ✓AI disclosure visibleAI support is disclosed as workflow assistance.
Done ✓Correction log initializedPublic correction trail starts at launch.
Done ✓Admin approvalApprove publish now has its own tested admin-only UI and backend gate.
Needs review ◐Sensitive-topic reviewSafe AI/tech scope now; review again before expansion.
Needs review ◐Cost ledger reviewedRequires budget confirmation before live AI calls.
Blocked until credentials ×Start research runBlocked until AI/search/provider cost guardrails and tests exist.