Skip to content
Guide v1

AI Problems

Check enablement, response mode and cooldown in AI Channels. Try a clear mention/reply. Verify Fork can read and send in the channel. All messages mode can increase usage and should not be enabled broadly just for testing.

Inspect tool enablement, original handler policy key, native permissions and current-message intent. A conversational response does not mean a requested action ran. Review its actual result record.

Inspect memory view and the privacy preference. Stored facts and semantic retrieval availability are separate. A missing embedding capability cannot be fixed by repeatedly asking the same question.

Read the returned wait/availability message. Ask has a per-user request window; image generation has a separate hourly limit. Provider budgets and fallback capabilities vary. See Usage Limits.

Never paste provider keys into chat. Report safe errors and times through Support.