AI tools
AI tools

AI error troubleshooter

Choose your setup, understand an error and follow a troubleshooting checklist.

0 / 8,000

Inputs are processed in this page's memory, without upload or storage. Closing the page or switching languages clears them.

Possible causes and checks

These checks are based on the error text. We have not accessed your account or confirmed the cause; an error can have several causes.

Choose your setup, understand an error and follow a troubleshooting checklist.

Find related guides →

Rules checked · 2026-11-11

Official documentation

Input or context too large
  1. Check whether the limit is model context or file-upload size.
  2. Include system instructions, history, attachments and tool definitions, leaving room for output.
  3. Save important content before reducing attachments or splitting the task; consult the tool's history-compaction instructions.
Balance or quota conditions
  1. Check whether the error concerns API balance, a key budget or a plan allowance.
  2. Check the account at the service receiving the request; a chat subscription is generally separate from API credit.
  3. Review pricing and permissions before deciding whether to add funds or adjust a budget; do not retry continuously.
Authentication mismatch
  1. Confirm the key belongs to this service and has not expired or been revoked.
  2. Check variable names, whitespace and where the tool reads credentials locally; never paste keys into websites or logs.
  3. For account-based authorization, use the tool's official sign-in flow.
Model ID, endpoint or access
  1. Copy the exact model ID from the selected service, not its display name.
  2. Check endpoint and protocol; aggregator IDs can differ from direct-provider IDs.
  3. Confirm account, region and tool support. A 404 can also mean an incorrect request path.
Rate or concurrency limit
  1. Read the error body: 429 alone does not prove your balance is exhausted.
  2. Check request, token and concurrency limits at both gateway and upstream service.
  3. Reduce concurrency and respect Retry-After; use delayed, bounded retries.
Permissions or access conditions
  1. Check account permissions, model access, region and organization policy.
  2. Read the provider's reason; content restrictions and account permissions differ.
  3. Follow official procedures or contact support; do not bypass restrictions.
Request format or parameters
  1. Confirm the protocol and message format match the endpoint.
  2. Check the named field's type and value, removing optional custom parameters one at a time.
  3. Test the official minimal example in your environment, then add features back gradually.
Internal server error
  1. Check the API service and upstream model's official status; 500 usually indicates a server-side failure.
  2. Check the error body and request outcome for possible execution or charges before retrying.
  3. Wait before making a limited number of retries. If it persists, give support a sanitized timestamp and request ID, never keys.
Network, timeout or service availability
  1. Check service status and distinguish local networking, proxy, API gateway and upstream model issues.
  2. Check DNS, certificates and proxy settings on your device; keep certificate verification enabled.
  3. A timed-out request may already have run or incurred charges. Check its outcome before retrying; give support a sanitized timestamp and request ID if it persists.

OpenRouter · DeepSeek

Rules checked · 2026-11-11