Files
hermes-webui/tests/test_issue3929_credential_pool_classification.py
T
nesquena-hermes 93f447c7af fix(#3929 partial): classify credential-pool exhaustion with an actionable hint
'All 0 credential(s) exhausted for <provider>' did not match _is_quota_error_text
so it fell through to a generic 'Error' with no hint (the exception path's else
branch discarded the hint entirely). Adds a distinct credential_pool_empty
classification (ordered before the quota check) + an explicit exception-path
branch + an actionable pool-specific hint, so the user sees why the turn failed
and what to do. Does not over-match (doesn't swallow real quota errors) and
doesn't block partial-harvest; downstream tolerates the new type.

NOTE: this is the CLASSIFICATION half of #3929 only — the data-loss half
(reasoning/partial-work restore on reload-reconcile) is a scoped follow-up that
needs the reporter's server logs.

Self-built (nesquena-hermes); nesquena independent review APPROVED.
2026-07-05 00:35:26 +00:00

53 lines
2.5 KiB
Python

"""Regression test for #3929 — credential-pool exhaustion classification.
Opus-advise on #3929 flagged that the error text a zero-credential pool produces,
`"All 0 credential(s) exhausted for opencode_go"`, does NOT match
_is_quota_error_text ('credential(s) exhausted' != 'credits exhausted'), so it
fell through to the generic 'Error' label with an empty hint — and in the
exception path the else-branch discarded the classification hint entirely.
This does not by itself fix the partial-work loss (that needs the reload-reconcile
reasoning-restore work + reporter logs for the incident autopsy — tracked
separately on the issue), but it gives the user an accurate, actionable message
for the exact error they hit instead of a bare 'Error'.
"""
from api import streaming
CREDENTIAL_POOL_ERRORS = [
"All 0 credential(s) exhausted for opencode_go",
"All 3 credential(s) exhausted for openrouter",
"credentials exhausted",
]
def test_credential_pool_exhaustion_classified_distinctly():
"""The 'N credential(s) exhausted' pool shape gets its own classification —
not generic 'error', not mislabeled as account quota."""
for err in CREDENTIAL_POOL_ERRORS:
classified = streaming._classify_provider_error(err, Exception(err))
assert classified["type"] == "credential_pool_empty", err
# Must carry an actionable, pool-specific hint (not empty, not top-up-account).
assert classified["hint"], f"empty hint for {err!r}"
assert "credential pool" in classified["hint"].lower(), err
assert classified["label"] == "No usable credentials", err
def test_credential_pool_is_not_misclassified_as_account_quota():
"""A zero-credential POOL is a config problem, not account credit exhaustion:
it must NOT get the 'Out of credits / top up your account' quota hint."""
classified = streaming._classify_provider_error(
"All 0 credential(s) exhausted for opencode_go",
Exception("All 0 credential(s) exhausted for opencode_go"),
)
assert classified["type"] != "quota_exhausted"
assert "top up" not in classified["hint"].lower()
def test_real_account_quota_still_classifies_as_quota():
"""Guard against over-matching: genuine account-credit exhaustion phrases
must still classify as quota_exhausted, not credential_pool_empty."""
for err in ("insufficient credits", "credits exhausted", "you have exceeded your current quota"):
classified = streaming._classify_provider_error(err, Exception(err))
assert classified["type"] == "quota_exhausted", err