Technical issues

Before opening a support ticket, a few quick checks will resolve most issues on the spot. If something is still wrong after these, the more context you can give us in the first message, the faster we can get you back to chatting.

Quick checks

• Refresh the page — many transient issues clear with a hard reload.
• Try a different model from the picker — a single provider can have a short outage that does not affect the rest.
• Sign out and back in if anything looks stale in the sidebar.
• Check your remaining inference credit in the account menu — an empty balance shows up as a refused send.

Common errors and what they mean

• Stream interrupted: the underlying model dropped the connection mid-response. Send the message again — partial output is not billed.
• Rate limited: too many concurrent streams from the same account. Close other tabs and try again in a few seconds.
• Insufficient balance: your inference credit is exhausted. Top up from the pricing page or wait for the next renewal.
• Provider error: the model provider returned an error we cannot interpret. Switching models usually works while we follow up upstream.

What to include when contacting support

• The browser and operating system you are using.
• A rough timestamp of when the issue happened.
• The model you had selected.
• A screenshot if anything visual is wrong.

Please do not share the contents of the conversation that triggered the issue unless it is essential — we can almost always reproduce the problem from the metadata alone, and keeping chat content private is the default.