Docs › Help & Privacy › AI slow or stuck
AI slow or stuck
Last updated May 14, 2026 · 3 min read
AI chat hanging without a reply / streaming cut off mid-response — run through this checklist in order.
Three common scenarios
1 · AI is still working, just slow
| Signal | Interpretation / fix |
|---|---|
| The chat panel shows "Coach is reading your work.." (or "…analyzing…" in script / audiobook mode) + a spinning icon | Still processing. Long contexts + deeper models can take a while |
| Several minutes with zero movement | Actually stuck — skip to the next scenario |
Why it's slow: longer conversation context, more chapters being analyzed, or a deeper model — all add time. It's a cost, not a bug.
2 · Streaming cut off mid-response
If AI starts a reply and then stops halfway, usually:
- A brief network glitch
- Rate limit on the AI provider's side
- Output hit the token limit
Fix: click the "Retry" button in the error notice — Slima re-sends that whole message and runs it again from the start (it does not resume the partial stream).
3 · The model itself is slow
Some models (deeper-thinking ones) are inherently slower. If speed matters:
- Slima has two models: the default "Writing Model" is higher-quality, while the "General Model" is faster and lower-cost. If speed matters more than nuance, switch to the General Model
- Resend the message
Cancelling a stream manually
To cancel a streaming response: while a reply is streaming, the Send button in the bottom-right of the input box turns into a red stop button (■) — click it to cancel. The input is disabled during streaming, so you can't cancel by typing a new message.
See: Advanced chat behaviors (stream cancel, long-chat compression, cross-chat memory)
Still stuck
Most of the time, opening a fresh conversation fixes it — the old chat's context might be in a weird state.
If a new chat also fails → Report a bug.