fix(gateway): trigger memory provider shutdown on /new and /reset

[fancydirty]

Fixes #7759

Problem

The /new and /reset commands were not calling shutdown_memory_provider() on the cached agent before eviction. This caused OpenViking (and any memory provider that relies on session-end shutdown) to skip commit, leaving memories un-indexed until idle timeout or gateway shutdown.

Root Cause

In _handle_reset_command(), the code flushed memories and closed tool resources, but it never shut down the memory provider. The session expiry watcher already does this correctly.

Fix

Add the missing shutdown_memory_provider() call in _handle_reset_command(), matching the behavior already present in the session expiry watcher (lines ~1600–1625).

Verification

• tests/gateway/test_session_reset_notify.py — 13 passed ✅
• tests/gateway/test_gateway_inactivity_timeout.py — 8 passed ✅
• tests/gateway/test_voice_command.py — 168 passed ✅

[Copilot]

Pull request overview

This PR fixes a session-lifecycle gap in the gateway: /new and /reset now trigger the cached agent’s shutdown_memory_provider() before eviction, matching the existing session-expiry behavior so memory providers (e.g., OpenViking) can commit/index on session end.

Changes:

• Invoke shutdown_memory_provider() on the cached agent during _handle_reset_command() before closing and evicting it.
• Keep the reset flow non-blocking by leaving the transcript-based flush as a background task while still triggering provider shutdown immediately.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[fancydirty]

Hi maintainers — this is a small, targeted fix for a session-lifecycle gap. It adds the missing shutdown_memory_provider() call during /new and /reset so that memory providers like OpenViking can commit/index on session end.

• Branch is up-to-date with main (no conflicts)
• Verified locally:
  • tests/gateway/test_session_reset_notify.py — 13 passed
  • tests/gateway/test_gateway_inactivity_timeout.py — 8 passed
  • tests/gateway/test_voice_command.py — 168 passed

Would love a review when you have a moment. Thanks!

[ZaynJarvis]

agree, these issues should be fixed.

[ZaynJarvis]

Review: Gateway /new Memory Commit Fix

Verdict: LGTM — ready to merge

Surgical 5-line fix.

In the gateway, /new /reset destroys the old agent entirely and creates a fresh one for the next session, so destroy should be save.

Checklist Pass

• shutdown_memory_provider() called before close() in _handle_reset_command()
• Wrapped in its own try/except Exception block — isolated from the close() error path
• hasattr guard — non-OV agents are completely unaffected
• Confirmed MERGEABLE by GitHub, 189 gateway tests pass

This is the fastest path to fixing commit timing for gateway /new and /reset. Recommend merging before the larger PRs (#9167, #8362) to reduce diff scope.

[teknium1]

Merged via PR #9796. Your commit was cherry-picked onto current main with your authorship preserved in git log. Thanks for catching this gap, @fancydirty!