hermes-agent/tests/run_agent/test_jsondecodeerror_retryable.py
Teknium c2b3db48f5
fix(agent): retry on json.JSONDecodeError instead of treating it as a local validation error (#15107)
json.JSONDecodeError inherits from ValueError. The agent loop's
non-retryable classifier at run_agent.py ~L10782 treated any
ValueError/TypeError as a local programming bug and short-circuited
retry. Without a carve-out, a transient JSONDecodeError from a
provider that returned a malformed response body, a truncated stream,
or a router-layer corruption would fail the turn immediately.

Add JSONDecodeError to the existing UnicodeEncodeError exclusion
tuple so the classified-retry logic (which already handles 429/529/
context-overflow/etc.) gets to run on bad-JSON errors.

Tests (tests/run_agent/test_jsondecodeerror_retryable.py):
  - JSONDecodeError: NOT local validation
  - UnicodeEncodeError: NOT local validation (existing carve-out)
  - bare ValueError: IS local validation (programming bug)
  - bare TypeError: IS local validation (programming bug)
  - source-level assertion that run_agent.py still carries the carve-out
    (guards against accidental revert)

Closes #14782
2026-04-24 05:02:58 -07:00

87 lines
3.7 KiB
Python

"""Regression guard for #14782: json.JSONDecodeError must not be classified
as a local validation error by the main agent loop.
`json.JSONDecodeError` inherits from `ValueError`. The agent loop's
non-retryable classifier at run_agent.py treats `ValueError` / `TypeError`
as local programming bugs and skips retry. Without an explicit carve-out,
a transient provider hiccup (malformed response body, truncated stream,
routing-layer corruption) that surfaces as a JSONDecodeError would bypass
the retry path and fail the turn immediately.
This test mirrors the exact predicate shape used in run_agent.py so that
any future refactor of that predicate must preserve the invariant:
JSONDecodeError → NOT local validation error (retryable)
UnicodeEncodeError → NOT local validation error (surrogate path)
bare ValueError → IS local validation error (programming bug)
bare TypeError → IS local validation error (programming bug)
"""
from __future__ import annotations
import json
def _mirror_agent_predicate(err: BaseException) -> bool:
"""Exact shape of run_agent.py's is_local_validation_error check.
Kept in lock-step with the source. If you change one, change both —
or, better, refactor the check into a shared helper and have both
sites import it.
"""
return (
isinstance(err, (ValueError, TypeError))
and not isinstance(err, (UnicodeEncodeError, json.JSONDecodeError))
)
class TestJSONDecodeErrorIsRetryable:
def test_json_decode_error_is_not_local_validation(self):
"""Provider returning malformed JSON surfaces as JSONDecodeError —
must be treated as transient so the retry path runs."""
try:
json.loads("{not valid json")
except json.JSONDecodeError as exc:
assert not _mirror_agent_predicate(exc), (
"json.JSONDecodeError must be excluded from the "
"ValueError/TypeError local-validation classification."
)
else:
raise AssertionError("json.loads should have raised")
def test_unicode_encode_error_is_not_local_validation(self):
"""Existing carve-out — surrogate sanitization handles this separately."""
try:
"\ud800".encode("utf-8")
except UnicodeEncodeError as exc:
assert not _mirror_agent_predicate(exc)
else:
raise AssertionError("encoding lone surrogate should raise")
def test_bare_value_error_is_local_validation(self):
"""Programming bugs that raise bare ValueError must still be
classified as local validation errors (non-retryable)."""
assert _mirror_agent_predicate(ValueError("bad arg"))
def test_bare_type_error_is_local_validation(self):
assert _mirror_agent_predicate(TypeError("wrong type"))
class TestAgentLoopSourceStillHasCarveOut:
"""Belt-and-suspenders: the production source must actually include
the json.JSONDecodeError carve-out. Protects against an accidental
revert that happens to leave the test file intact."""
def test_run_agent_excludes_jsondecodeerror_from_local_validation(self):
import run_agent
import inspect
src = inspect.getsource(run_agent)
# The predicate we care about must reference json.JSONDecodeError
# in its exclusion tuple. We check for the specific co-occurrence
# rather than the literal string so harmless reformatting doesn't
# break us.
assert "is_local_validation_error" in src
assert "JSONDecodeError" in src, (
"run_agent.py must carve out json.JSONDecodeError from the "
"is_local_validation_error classification — see #14782."
)