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

fix(state): pin the shared storage-failure action copy to the failing profile

Browse Source 
hermes_state_user_copy.py is the one table feeding the CLI banner, gateway
warning and TUI/Desktop RPC errors, and its action strings were still bare
(`hermes doctor --fix`, `hermes gateway stop`, `hermes sessions recover`).
Substitute {profile_arg} once in describe_storage_failure so those surfaces
get the same pin as the turn explainer.

Also keep final_response unbound in the turn_finalizer error fallback: it
feeds the external-memory sync and the background-review gate, which must
still see an empty response on a persistence-failed turn. One test binds the
fallback (explainer stubbed empty) and the untouched final_response.
This commit is contained in:
kshitijk4poor
2026-09-16 17:40:34 +05:30
committed by kshitij
parent 9c093c83e4
commit b6bba97f7b
5 changed files with 70 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
hermes_state_user_copy.py
View File

@@ -20,14 +20,15 @@ class StorageFailure:
action: str # what to do, one sentence naming the exact command
_DOCTOR = "Run `hermes doctor --fix` to diagnose and repair."
_DOCTOR = "Run `hermes {profile_arg}doctor --fix` to diagnose and repair."
# cause -> (code, gloss, action). "disk" is split by is_disk_full_error at lookup time.
_STORAGE_FAILURES: dict[str, tuple[str, str, str]] = {
"locked": (
"storage_locked",
"the session database is locked by another Hermes process",
"Wait a moment and try again; if it persists, stop the other Hermes process (`hermes gateway stop`).",
"Wait a moment and try again; if it persists, stop the other Hermes process "
"(`hermes {profile_arg}gateway stop`).",
),
"disk_full": (
"disk_full",
@@ -42,22 +43,22 @@ _STORAGE_FAILURES: dict[str, tuple[str, str, str]] = {
"corrupt": (
"storage_corrupt",
"the session database file is damaged",
f"{_DOCTOR} Recovery: `hermes sessions recover --source <state.db> --inspect-only`.",
f"{_DOCTOR} Recovery: `hermes {{profile_arg}}sessions recover --source <state.db> --inspect-only`.",
),
"fts_index": (
"storage_index_corrupt",
"the session search index is damaged (the messages themselves are intact)",
"Run `hermes doctor --fix` (or `hermes sessions repair`) to rebuild it.",
"Run `hermes {profile_arg}doctor --fix` (or `hermes {profile_arg}sessions repair`) to rebuild it.",
),
"replaced": (
"storage_replaced",
"the session database file was replaced while Hermes was running",
"Stop Hermes (`hermes gateway stop`), run `hermes doctor`, then start it again.",
"Stop Hermes (`hermes {profile_arg}gateway stop`), run `hermes {profile_arg}doctor`, then start it again.",
),
"deleted_wal": (
"storage_replaced",
"the session database file was changed or replaced while Hermes was running",
"Stop Hermes (`hermes gateway stop`), run `hermes doctor`, then start it again.",
"Stop Hermes (`hermes {profile_arg}gateway stop`), run `hermes {profile_arg}doctor`, then start it again.",
),
"compression": (
"storage_busy",
@@ -87,7 +88,14 @@ def describe_storage_failure(exc_or_str) -> StorageFailure:
cause = classify_persistence_error(exc_or_str)
key = "disk_full" if cause == "disk" and is_disk_full_error(exc_or_str) else cause
code, gloss, action = _STORAGE_FAILURES.get(key, _STORAGE_FAILURES["unknown"])
return StorageFailure(cause=cause, code=code, gloss=gloss, action=action)
# Pin the copy-pasteable command to the profile whose store failed: a multi-profile backend
# serves sessions whose state.db is not the process default, and a bare `hermes` follows the
# sticky active_profile (#105887).
from hermes_constants import profile_cli_selector
return StorageFailure(
cause=cause, code=code, gloss=gloss, action=action.replace("{profile_arg}", profile_cli_selector())
)
def storage_failure_details(exc_or_str, limit: int = 200) -> str: