Skip to content

fix(gateway): retry startup auto-resume when a failed platform reconnects#37669

Closed
Frowtek wants to merge 1 commit into
NousResearch:mainfrom
Frowtek:fix/late-reconnect-auto-resume
Closed

fix(gateway): retry startup auto-resume when a failed platform reconnects#37669
Frowtek wants to merge 1 commit into
NousResearch:mainfrom
Frowtek:fix/late-reconnect-auto-resume

Conversation

@Frowtek

@Frowtek Frowtek commented Jun 2, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

What does this PR do?

The gateway's documented startup auto-resume skips sessions whose platform
adapter isn't connected yet — and never retries them. A platform that connects
shortly after startup left its interrupted sessions behind; they only recovered
if the user sent a new message. This wires the auto-resume into the reconnect
path so a recovered platform retries the continuation for its own sessions.

Related Issue

Type of Change

  • 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

  • gateway/run.py — _schedule_resume_pending_sessions() gains an optional
    platform filter + an in-flight guard against double resumes.
  • gateway/run.py — the reconnect-watcher success path re-runs the auto-resume
    scoped to the reconnected platform.
  • tests/gateway/test_restart_resume_pending.py,
    tests/gateway/test_platform_reconnect.py — regression coverage.

How to Test

  1. Mark a session resume_pending while its adapter is offline at startup → skipped.
  2. Bring the platform online → reconnect schedules the resume for that session.
  3. scripts/run_tests.sh tests/gateway/test_restart_resume_pending.py tests/gateway/test_platform_reconnect.py -q — all pass.

Checklist

  • Read the Contributing Guide
  • Commit messages follow Conventional Commits (fix(gateway): ...)
  • Searched for duplicate PRs
  • Only changes related to this fix
  • Tests pass and I've added regression tests
  • Cross-platform impact — N/A (pure async control-flow)

@alt-glitch alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/gateway Gateway runner, session dispatch, delivery labels Jun 2, 2026
@mohamedorigami-jpg

mohamedorigami-jpg commented Jun 2, 2026

Copy link
Copy Markdown
Contributor

Clean separation of startup resume vs reconnect resume. The platform scoping is the right approach — reconnects should only retry their own sessions.

One thing worth checking: the entry.session_key in self._running_agents guard on the reconnect path assumes the original agent has already been launched and stored in _running_agents before the reconnect fires. If a platform is flapping (disconnect→reconnect in rapid succession), there's a timing window where the first reconnect triggers resume, the agent fires, the platform drops again, the second reconnect fires before the first agent is stored in _running_agents — you'd get duplicate resumes for the same session. A transition counter (disconnect_count incremented on each drop, matched on reconnect) would close that gap if flappy platforms are a real concern for your deployment.

@teknium1 teknium1 mentioned this pull request Jun 4, 2026
Merged
@teknium1

teknium1 commented Jun 4, 2026

Copy link
Copy Markdown
Contributor

Merged via PR #39018 — your commit was cherry-picked onto current main with your authorship preserved in git log (rebase-merge, commit 71a9f44). Thanks @Frowtek!

@xiaoyaner0201 xiaoyaner0201 mentioned this pull request Jun 4, 2026
Open
@xiaoyaner0201

xiaoyaner0201 commented Jun 4, 2026

Copy link
Copy Markdown

Cherry-picked this onto a v0.15.2-based fork (Discord gateway) alongside #30030. Resolves a restart-resume amnesia we were hitting — a session whose platform reconnected slightly after startup never got its interrupted turn resumed until a fresh user message arrived. The test_restart_resume_pending + test_platform_reconnect suites pass. Thanks for the fix!

@xiaoyaner0201 xiaoyaner0201 mentioned this pull request Jun 4, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

comp/gateway Gateway runner, session dispatch, delivery P2 Medium — degraded but workaround exists type/bug Something isn't working

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@Frowtek @mohamedorigami-jpg @teknium1 @xiaoyaner0201 @alt-glitch