Support
Troubleshooting
Find the failing layer quickly: authentication, account, deployment, policy, routing or destination model.
Start with evidence
Record the request id, timestamp, serving name, HTTP status and deployment state. Do not include API keys or sensitive prompt content in support messages.
Isolate the layer
Test the destination directly before testing routing. Remove optional tools and firewall fields, then add them back one at a time after the base request succeeds.
Common recovery paths
Refresh an expired key, restore account balance, retry transient capacity failures and inspect the recorded LLM Firewall or routing decision before changing configuration.