Streaming or chat errors

A response starts and cuts off, the stream errors mid-reply, the typing indicator never clears, or chat throws a generic error.

Likely causes

A transient network drop interrupted the streamed response.
The workspace ran out of credits, or the sender reached their monthly cap, partway through a turn.
The model provider returned an error or timed out.
The conversation tab fell out of sync after the URL changed on the first message.
A very long input or attachment exceeded what the model could process.

Stop and retry
If a stop button is visible, click it, then resend the message. If the typing indicator hangs, reload the conversation and try again. Most single-turn stream errors are transient and clear on retry.
Check credits and caps
A turn that cuts off mid-stream often means the balance hit zero or the user reached their monthly cap during generation. Confirm at Settings → Usage & Billing. See Credit / quota issues.
Confirm the model and key
A provider error or an invalid API key surfaces as a failed stream. In the agent’s AI Model tab, verify the model is active and the key is valid, then retry.
Reopen the conversation
The chat page updates its URL after the first message without remounting. If the sidebar points at a stale row, click New Task and reopen the correct conversation from history.
Reduce the input size
If the error only happens on large messages or attachments, shorten the prompt or split the attachment. See Files in chat.

Conversation — reopen the chat to read any inline error attached to the failed turn.
Usage & Billing — /settings/usage-billing for balance and caps.
API keys — /settings/api-keys to confirm the provider key is valid.