Help
Troubleshooting
Every refusal the desk can show, with its cause and fix: writing off, unpaid, kill switch, held opening lines, tokens, connecting, vendor review, rate limits, motion.
Each entry is a sentence the desk or a mail actually prints. Match the words on the screen, then follow the fix. If none matches, open /status and then /contact. — Writing is off on this deployment — no model key is configured. You can still write posts yourself, approve them and schedule them. Cause: this deployment has no model key, or writing has been switched off on it, so Generate and Rebuild Brain are refused. The keyless demo also shows this. Fix: contact support (or, if you run the server yourself, see DEPLOY.md). Until then write drafts by hand in Studio, approve them on the Queue and let the scheduled time send. Where: the AI callout on Home and Studio; Generate is disabled. — Writing is not available on this desk. SuperPost will not invent a week. Cause: the same missing model key, hit from Generate directly. Fix: as above. No week is fabricated; that is deliberate. Where: Studio → Generate. — Workspace unpaid · Publishing needs a paid plan. Cause: the workspace has no active or trialing subscription (payment failed past the grace period, the paid window ended, or nothing was bought). Native publish is parked for six hours and retried; nothing is deleted. Fix: Settings → Billing → Manage billing to pay or reactivate. Copy and send by hand still works. Where: the header badge, the Queue send button, the unpaid mail. — Kill switch is on. Every adapter is paused. · Kill switch is on. Cause: someone turned pause on. Unattended sending turns off and stays off. A publish waits until the next day, at the hour it was meant to go out. A first reply waits an hour. Approved drafts stay in the queue. Fix: Settings → Kill switch → turn it off. Turning the switch off makes those two jobs due then. Where: the red banner on every desk screen, the publish job note. — Held: opening line <n> is under the <floor> auto-publish floor. (for example: Held: opening line 65 is under the 70 auto-publish floor.) Cause: an Autopilot post sends on its own only when its opening line scores at least the bar set for that account. This one scored lower, so it stays approved and waits for you. Send on the Queue does not apply this hold. Fix: send it yourself from the Queue, or sharpen the first line and re-score. The floor is per account on Connections; lowering it is a policy change, not a fix for one draft. Where: the Queue row's error line on that draft. — Token expired. Reconnect. · Reconnect X before SuperPost goes silent · <network> expires in under 7 days. Reconnect the account. Cause: the network's OAuth token expired, or refresh failed. The account shows expired and publishing to it stops. Fix: Connections → Reconnect on that account. Approved drafts for it resume at the next scheduled time. Where: Connections, the token_expiring mail, the red health tile. — This account is not set up for native publish yet. Copy and send still works. Cause: this deployment is not set up to publish to that network yet, or the network is not switched on here. Fix: contact support; until then use Copy and send from the Queue. Where: the note at the top of Connections after a failed Connect. — Connect came back to Connections with ?error=bad_state, missing_code, exchange or scopes. Cause: bad_state — the sign-in took too long or was started by another user; missing_code — you cancelled at the network; exchange — the network refused the sign-in; scopes — you granted fewer permissions than publishing needs, so the account is set to copy and send; elsewhere — that account is already connected in another SuperPost workspace. Fix: retry Connect once; for scopes, reconnect and accept every permission shown; for elsewhere, disconnect it there first. If exchange keeps happening, contact support. Where: Connections; the account's last error line. — Waiting on vendor review. Cause: the network is still reviewing SuperPost's access, so that connection shows Needs review. This is the network's gate, not a SuperPost fault. Fix: wait for the network; until the connection leaves review, nothing is sent there. Otherwise nothing is sent until you approve. YouTube uploads as private. You publish it in YouTube Studio. Where: Connections; the publish job ends as permanent with the vendor's message. — SuperPost could not publish (mail) · Still did not attach. Post not sent. Cause: the network rejected the post, or the media never attached (the X upload returned 429 or failed to finalise). The draft goes back to approved and a mail is sent. Fix: open the draft on the Queue, read the error line, fix the media or copy, and send again. A URL in an X body is refused unless this is a release post. Where: the Queue row's error line, the publish_failed mail. — Rate limit: <n> per workspace per hour. · GitHub rate limited this request. Wait and retry, or Connect GitHub. Cause: a per-workspace hourly counter for generate, a repo read or the public Pull Test was hit, or GitHub throttled an anonymous read. Fix: wait for the window to pass; for GitHub, Connect GitHub so reads use your installation's quota. Where: the toast on the action you tried; the Pull Test page. — This month's … are used, so they are paused. · This month's X allowance is used · This plan does not include <feature>. Cause: the month's allowance (plan plus top-ups) is used, or the feature (a second product, seats) is not in the plan. Fix: Settings → Billing: add a top-up on Pro, or change plan. Nothing is charged automatically. Where: the toast on Generate or Send; Billing. — Motion render is not configured on this host. Cause: this deployment has no video renderer set up, so motion renders nothing. Stills still render. Fix: pick a still or skip motion for this draft; contact support if you need motion. Where: the Queue when you request motion. — SuperPost health is red (mail and header tile) Cause: a health check failed — the database, the worker heartbeat, a publish adapter, or the still renderer. The header tile says which. Fix: open /status for the failing part. If the worker is stale, the background worker is not running; contact support. Where: the header tile, /status, the health_red mail.