fix(gateway): emit error when turn cancels before agent ready

[deniqlab]

Summary

When a turn is cancelled during lazy agent startup, run_after_agent_ready() used to clear the inflight turn and return silently. That made the Desktop look like the first send vanished: prompt.submit returned streaming, but the user never received an error event or a turn start.

This patch emits an explicit error event before returning in the cancelled / not-running path so the client can surface the failure instead of hanging silently.

Changes

• tui_gateway/server.py: emit an error event when the deferred turn is cancelled before agent readiness or the session is no longer running.
• tests/test_tui_gateway_server.py: add regression coverage for both the user-cancelled and not-running cases.

Verification

• /Users/debuggerson/.hermes/hermes-agent/venv/bin/python -m pytest tests/test_tui_gateway_server.py -k 'interrupt_before_agent_ready_prevents_late_turn_start or prompt_submit_emits_error_when_stopped_before_agent_ready'
• /Users/debuggerson/.hermes/hermes-agent/venv/bin/python -m pytest tests/test_tui_gateway_server.py
• git diff --check

Closes #63078

[alt-glitch]

This was generated by AI during triage.

Duplicate of open #65567: both live diffs add the same deferred-turn error event for cancellation or stopped sessions before agent readiness. #65567 is the earlier focused PR.

[teknium1]

Thanks for isolating the deferred-start cancellation path. Current main still has the silent return at tui_gateway/server.py:9398-9401, while session.interrupt sets the observed flag at tui_gateway/server.py:9005. The proposed error event matches existing Ink handling at ui-tui/src/app/createGatewayEventHandler.ts:973-991 and Desktop handling at apps/desktop/src/app/session/hooks/use-message-stream/gateway-event.ts:695-744.

This is an automated hermes-sweeper review.

[GottZ]

This was generated by AI during triage.

Summary

Three PRs address or reference #63078: #65567 and #66545 add error emission for the pre-ready cancellation/not-running branch, while merged #71140 records a broader fix for the reported first-message loss by extending the agent-build wait and surfacing initialization failures. The two open PRs cover the silent-return symptom but do not resolve the underlying long-build timeout described by the issue.

Related pull requests

• fix(gateway): emit error event when a turn is cancelled before agent ready #65567 best fix — (+159/-0) — n/a: The diff emits an error and adds two regression tests for the deferred-start cancellation/not-running branch, but a blocking contributor review notes that its shared predicate incorrectly reports an intentional Stop as a generic failure; it remains relevant because merged fix(desktop): never drop the first message in a new session (patient agent-build wait + error surfacing) #71140 explicitly records this server-side path as incorporated from fix(gateway): emit error event when a turn is cancelled before agent ready #65567.
• fix(gateway): emit error when turn cancels before agent ready #66545 partial — (+86/-0) — n/a: The diff implements the same pre-ready error emission as fix(gateway): emit error event when a turn is cancelled before agent ready #65567 with a parameterized regression test, including the same generic-error behavior for an intentional Stop; it is the later duplicate identified by contributor triage.
• fix(desktop): never drop the first message in a new session (patient agent-build wait + error surfacing) #71140 [merged] best fix — (+1179/-3) — n/a: This merged reference implementation records the broader backend fix: a patient, bounded wait for live agent builds, visible build-failure handling, and incorporation of fix(gateway): emit error event when a turn is cancelled before agent ready #65567's pre-ready error-surfacing path, directly addressing the issue's timeout-driven first-message loss.

Duplicates

#66545 is substantively a duplicate of #65567; #65567's server-side change was subsequently incorporated into the broader merged #71140, yielding the explicit chain #66545 → #65567 → #71140.

Suggested consolidation

Close #66545 as duplicate of #65567, then close #65567 as duplicate of merged #71140. This departs from the keep_open reviews on #66545 and #65567 because the later merged #71140 records incorporation of #65567's server-side path and adds the broader build-wait fix; additionally, the blocking contributor review on #65567 requires intentional Stop not to emit a generic error, so neither open diff should be merged unchanged.

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
    I63078(["issue #63078 (closed)"])
    subgraph Dup65567 ["PRs duplicating each other"]
        P65567["PR #65567 (open)"]
        P66545["PR #66545 (open)"]
    end
    P66545 -.->|partial| I63078
    class I63078 closed
    class P65567 open
    class P66545 open
    class P65567 best
    class P66545 target
    click I63078 "https://github.com/NousResearch/hermes-agent/issues/63078"
    click P65567 "https://github.com/NousResearch/hermes-agent/pull/65567"
    click P66545 "https://github.com/NousResearch/hermes-agent/pull/66545"

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 3 pull requests and 1 issue in this complex. Diffs were read for 2 of 3 PRs (rest unavailable); Assessment working set: 11 kB of PR diffs, 13 kB of issue/PR text, 13 kB of discussion (15 comments), 5 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.