hermes-admin/hermes-agent
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

fix(gateway): treat a hard-deleted session row as a dead route

Browse Source 
_is_session_ended_in_db answered False when the row was missing from a
readable owning DB, so a session hard-deleted out of the gateway process
(CLI/TUI/desktop local delete) kept its live channel->session route: the
next inbound message resolved the deleted id and run_agent's INSERT OR
IGNORE re-created the row with its old content — the deleted conversation
resurrected (#42422). A missing row is now the same verdict as an ended
one: the stale-route self-heal drops the entry, recovery finds nothing,
and the peer mints a fresh session. DB errors still answer False.
This commit is contained in:
Brooklyn Nicholson
2026-09-26 21:59:35 -05:00
committed by brooklyn!
parent a9109b07f4
commit 981ec5d150
1 changed files with 9 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
gateway/session_lifecycle.py
View File

@@ -59,15 +59,19 @@ class SessionLifecycleMixin:
"""SessionStore explicit boundaries and crash-recovery markers."""
def _is_session_ended_in_db(self, session_id: str) -> bool:
"""True iff state.db has this session with a non-null end_reason (same staleness test as
``_prune_stale_sessions_locked``; no DB/row or DB error -> False). Lets routing self-heal a
session ended while the gateway stays alive. Store resolved from the owning profile.
"""True iff state.db says the session is gone: ended (non-null end_reason) or hard-deleted
(no row in a readable owning DB). No DB or a DB error -> False (same failure mode as
``_prune_stale_sessions_locked``). Lets routing self-heal a session finalized or deleted
while the gateway stays alive. Store resolved from the owning profile.
Used by ``get_or_create_session`` to self-heal at routing time: ``_prune_stale_sessions_locked``
only runs at startup, so a session ended in the DB while the gateway stays alive (any path that
finalizes the row without clearing sessions.json) would otherwise be reused as a live routing key
and silently swallow every subsequent message until the next restart (#54878 — the live-gateway
variant of #52804/FM9). DB errors are non-fatal — never block routing on a failed lookup.
variant of #52804/FM9). A hard delete is the same shape one step further: the row is GONE, not
merely ended, and reusing the route makes run_agent's INSERT OR IGNORE resurrect the deleted
session with its old id (#42422) — so a missing row is treated exactly like an ended one. DB
errors are non-fatal — never block routing on a failed lookup.
The store is resolved from the row's owning profile rather than the ambient scope: an unscoped
background writer keeps its own copy of the same session, and comparing against that copy reports a
live session as ended (#66887).
@@ -79,7 +83,7 @@ class SessionLifecycleMixin:
row = db.get_session(session_id)
except Exception:
return False
return bool(row is not None and row.get("end_reason") is not None)
return row is None or row.get("end_reason") is not None
def _route_reset_reason(self, entry: SessionEntry) -> Optional[str]:
"""Only explicit suspension replaces a routed conversation; time never does."""