fix(desktop): Esc interrupts running turn even when trigger popover is open (#74374)

[webtecnica]

Description

Fixes #74374 — pressing Esc during an active assistant turn now correctly interrupts generation, matching Stop-button behavior.

Root Cause

In handleEditorKeyDown, the trigger popover's Escape handler (line ~624-630) catches the Escape key when any @ or / completion popover is open. It closes the popover via closeTrigger() and returns — but never reaches the general Esc handler at line ~788-805 which calls haltRun() to stop the running turn.

This means: when a trigger popover was active during generation, pressing Esc would dismiss the popover but the turn kept running, requiring a second Esc press to stop. When focus was in the composer input without a trigger popover, Esc worked correctly (line 788 caught it).

Fix

After closing the trigger popover, the handler now also checks busy && !awaitingInput and calls haltRun() to stop the running turn. One Esc press now both dismisses the popover AND interrupts generation.

Changed file

• apps/desktop/src/app/chat/composer/index.tsx: Add haltRun() call after closeTrigger() when turn is busy.

[alt-glitch]

This was generated by AI during triage.

Related to #74406 for #74374, but this uses a different Esc mechanism. The branch also carries the unrelated cold-start restore patch from #74426/#74388; please separate or consolidate that carried work before selecting the Esc repair.

[teknium1]

Thanks for isolating the trigger-popover path. The reported bug is still present on current main: apps/desktop/src/app/chat/composer/index.tsx:648-653 closes the populated trigger popover and returns before the busy-turn Escape halt at :821-828. Calling the existing haltRun() is consistent with the current Stop/Esc queue-parking semantics at :272-284.

Problems

• This branch also carries the separate cold-start restore change across use-desktop-integrations.ts, use-gateway-boot.ts, and store/profile.ts (commit 6c81772c3828). Please separate or explicitly consolidate that work, as the existing maintainer comment requests.
• The Esc behavior has no regression test. Add coverage for a populated trigger popover + busy && !awaitingInput, asserting one Escape both closes the popover and invokes cancellation; retain coverage that awaitingInput does not halt.

Suggested changes

• Salvage the focused Esc hunk from 62f5e622080d after resolving its current-main context, then add the behavior test.

This is an automated hermes-sweeper review.

[teknium1]

Please add a regression test for this branch: with a populated trigger popover and busy && !awaitingInput, one Escape should close the popover and invoke cancellation. Cover the paired awaitingInput case to preserve the existing non-interrupt contract.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs reference #74374, but they address different Escape-routing paths: #74406 adds stale active-composer fallback machinery, while #74428 fixes the trigger-popover branch that returns before the existing busy-turn cancellation handler.

Related pull requests

• fix(desktop): Esc cancels a running turn when the focus bus is stale #74406 [closed] best fix — (+218/-3) — close as already implemented on main: this closed PR remains relevant as the stale-focus proposal, but commit f12e652 already heals stale or buried composer claims in apps/desktop/src/app/chat/composer/focus.ts:127-136, with coverage at apps/desktop/src/app/chat/composer/focus.test.ts:150-182, as documented by the maintainer-bot verdict.
• fix(desktop): Esc interrupts running turn even when trigger popover is open (#74374) #74428 best fix — (+46/-6) — keep open with a salvage path: unlike fix(desktop): Esc cancels a running turn when the focus bus is stale #74406, its Esc hunk fixes the populated trigger-popover path, where closeTrigger() returns before haltRun(); consistent with the keep_open review on fix(desktop): Esc interrupts running turn even when trigger popover is open (#74374) #74428, salvage the focused hunk from 62f5e62, separate the unrelated cold-start restore changes, and test both busy && !awaitingInput cancellation and the awaitingInput non-interrupt case.

Suggested consolidation

Keep #74428 open with the reviewed salvage path: retain only the trigger-popover Escape fix, split out the cold-start restore work identified by the contributor reviews, and add the requested regression coverage. Retain #74406's closure as already implemented on main; the PRs are not duplicates because #74406 concerns stale composer targeting while #74428 covers an independent early return in the trigger-popover handler.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I74374(["issue #74374 (closed)"])
    P74428["PR #74428 (open)"]
    P74428 -->|best fix| I74374
    class I74374 closed
    class P74428 open
    class P74428 best
    class P74428 target
    click I74374 "https://github.com/NousResearch/hermes-agent/issues/74374"
    click P74428 "https://github.com/NousResearch/hermes-agent/pull/74428"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 1 issue in this complex. Each diff was read against this issue; Assessment working set: 17 kB of PR diffs, 7 kB of issue/PR text, 2 kB of discussion (3 comments), 4 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.