Feature Request: Queue user messages while a reply is generating instead of cancelling the in-flight turn
Priority: High — this breaks a very common multi-question workflow.
Scenario
On the desktop app (tested on v1.18.15), when I send several messages back-to-back (for example: 1) "check BTC/ETH price", 2) "which web3 coin is the 24h dark horse", 3) "weather tomorrow in Guangzhou"), the first reply gets cancelled as soon as message #2/#3 arrive. The result: the first turn is truncated mid-response, and the queued messages are not processed in order — the whole point of sending them together is lost.
Current behaviour
- New user message during generation cancels the in-flight turn.
- The partially generated reply is discarded/truncated (no "continue" or resume).
- Messages sent together do not run sequentially.
Expected behaviour
- Messages sent while a reply is generating should be queued and processed in order after the current turn completes (UI could show "N queued").
- Keep an explicit "cancel" action for users who do want to interrupt (not the default).
- Alternative/fallback: if cancellation stays, at least preserve the truncated reply so the user can continue/resume it, and never silently drop it.
Why it matters
Queuing several independent prompts at once (quotes, weather, lookups, batch questions) is a natural way to use a chat client. Today it results in lost/truncated answers and confusing UX. This is not an edge case — any user who types faster than the model responds will hit it.
Related
- v1.18.14 note: "Waited for queued ACP session updates before ending a turn" — internal queueing already exists; extending it to user messages would be a natural step.
- No user-facing message-queue feature appears in recent releases (v1.18.6–v1.18.15).
Suggested implementation sketch
- When a user message arrives while a turn is generating, append it to a per-session FIFO instead of cancelling.
- On turn completion, start the next queued message automatically.
- Composer shows "2 messages queued" with a "Send now (interrupt)" option.
Feature Request: Queue user messages while a reply is generating instead of cancelling the in-flight turn
Priority: High — this breaks a very common multi-question workflow.
Scenario
On the desktop app (tested on v1.18.15), when I send several messages back-to-back (for example: 1) "check BTC/ETH price", 2) "which web3 coin is the 24h dark horse", 3) "weather tomorrow in Guangzhou"), the first reply gets cancelled as soon as message #2/#3 arrive. The result: the first turn is truncated mid-response, and the queued messages are not processed in order — the whole point of sending them together is lost.
Current behaviour
Expected behaviour
Why it matters
Queuing several independent prompts at once (quotes, weather, lookups, batch questions) is a natural way to use a chat client. Today it results in lost/truncated answers and confusing UX. This is not an edge case — any user who types faster than the model responds will hit it.
Related
Suggested implementation sketch