NNyquest docs

Common Error Messages

Before retrying a generation

A connection error does not prove the provider did no work. Use this recovery table before submitting the same operation again:

OutcomeNext step
Rejected before dispatch (invalid input or authentication)Correct the request and submit it again
Rate/capacity rejectionHonor Retry-After and back off
Answer returned with x-nyquest-settlement: pendingKeep the answer; check Usage rather than generating again
Audio/media settlement pendingCheck Usage and the result before retrying; contact support if it remains unresolved
Video submitting, running, or requires_reconciliationPoll the existing job ID; do not create another job to recover the first
Stream interrupted or provider outcome unknownCheck history, Usage, and job status; a new request may generate and charge again

Nyquest does not provide a universal client idempotency-key/response-replay guarantee. Confirmed video submission rejection or confirmed render failure can release a hold; an unknown provider outcome needs reconciliation first. Contact support with the request/job ID if it stays unresolved.

What each error message means and how to recover.

Error messages in Nyquest follow a consistent format: a short title, a one-sentence explanation, sometimes a suggested action. This article maps the most common ones to their root causes.

"Insufficient balance"

Where: chat input shows it on send, agent runs reject it on start. Means: wallet is at $0 and the request would charge it. Fix: add funds (click the balance pill in the top bar; phones: nav drawer → + Add funds) OR set up BYOK (Settings → Providers) so requests bypass the wallet.

"Rate limit exceeded"

Where: chat input rejects the send. Means: you've sent more requests than the per-minute cap allows. Limits:

  • Free tier: 20 chat requests/min, 10 auth requests/min
  • Pro tier: 120 chat requests/min, 10 auth requests/min

Fix: wait 60 seconds. If you regularly hit it, fund the wallet (becomes Pro) or talk to a real human.

"Model not found" or "Model unavailable"

Where: model picker shows strikethrough or chat fails on send. Means: the model is retired or your provider revoked access. Fix: pick a different model from the picker. See Models, Providers, and BYOK Issues.

"API key invalid"

Where: Settings → API Keys (Test button), or chat fails on a BYOK provider. Means: Nyquest can't authenticate with your provider using the saved key. Fix: verify the key at the provider's dashboard, rotate if needed, update in Nyquest.

"Provider unreachable" or "Provider error 5xx"

Where: chat fails mid-stream or right after send. Means: the upstream model provider had a transient outage. Fix: preserve the response/request ID and check Usage or job status first. Retry only once you have distinguished a rejected request from pending settlement or an unknown provider outcome. See the recovery table above.

"Conversation not found"

Where: clicking a conversation link. Means: the conversation was deleted or belongs to a different account. Fix: none — sign in to the right account, or accept it's gone if deleted.

"Project not found"

Where: clicking a project URL or refresh. Means: the project was deleted or you're on a different account. Fix: check Settings → Projects for your active project list.

"Forbidden" or 403

Where: admin endpoints, or rare auth-related calls. Means: your account doesn't have permission for that action. Fix: if you're trying to access /admin, you need admin status — only granted by request. Otherwise, this usually indicates a session issue: sign out and back in.

"Unauthorized" or 401

Where: any backend call after the JWT expires. Means: your session expired (default: 8 hours after sign-in). Fix: sign out and sign back in.

"CORS error" (in browser console)

Where: browser dev tools. Means: a frontend API call is being blocked by CORS policy. Almost always indicates a deploy mismatch. Fix: hard-refresh the page (Cmd/Ctrl+Shift+R). If persistent, talk to a real human — could be a backend deploy issue.

"Service Unavailable" or 503

Where: any endpoint. Means: the backend is briefly down (deploy in progress, restart, etc.). Fix: inspect the error type and the recovery table above. A capacity rejection can be retried after backoff; pending settlement or unknown completion needs a status check first. If unresolved, contact support.

"Network error" / "Failed to fetch"

Where: browser console or chat input. Means: your network couldn't reach api.nyquest.ai at all. Fix:

  • Check your internet connection
  • Disable VPN if you have one (some block our domain)
  • Try a different network
  • Check status.nyquest.ai (when deployed)

"Image generation failed"

Where: image-capable model returns no image. Means: the upstream image model errored — usually content policy or provider outage. Fix: rephrase the prompt (avoid trigger words), or try a different image model.

"Memory recall service unavailable"

Where: chat completes successfully but no recall happened. Means: the memory layer was briefly unavailable. The chat itself worked; recall just didn't trigger. Fix: none needed — recall will resume on next message. Won't affect chat correctness.

"Discord support unavailable"

Where: /help/live-support shows a 503 or "service disabled" message. Means: the Discord bot is offline or env vars are misconfigured. Fix: talk to support via email (the email path documented in Contacting Support). This is rare; usually means a deploy hiccup on the backend.

Where to next