42f210de11
Qwen via OpenRouter / the Nous Portal relays Alibaba Model Studio's output
guardrail as an in-band 502 marked provider_unavailable ("Output data may
contain inappropriate content."), and its input guardrail as DashScope's 400
data_inspection_failed. Neither phrase was a content-policy pattern, so the
502 rode the server-error retry ladder and Hermes re-sent the same content
into the same refusal. Both now classify as content_policy_blocked.
29 lines
1.4 KiB
Python
29 lines
1.4 KiB
Python
"""Provider safety-filter refusals classify as content_policy_blocked, never as a retryable outage."""
|
|
|
|
import pytest
|
|
|
|
from agent.error_classifier import FailoverReason, classify_api_error
|
|
|
|
|
|
class _APIError(Exception):
|
|
def __init__(self, message, status_code=None):
|
|
super().__init__(message)
|
|
self.status_code = status_code
|
|
self.body = {}
|
|
|
|
|
|
@pytest.mark.parametrize("status, message", [
|
|
# Output guardrail relayed by OpenRouter / the Nous Portal as an in-band 502 marked provider_unavailable.
|
|
(502, "Upstream error from Alibaba: Output data may contain inappropriate content."),
|
|
(None, "Upstream error from Alibaba: Output data may contain inappropriate content."),
|
|
# Input guardrail: DashScope's documented 400 (code data_inspection_failed).
|
|
(400, "Error code: 400 - {'error': {'message': 'Input data may contain inappropriate content.', "
|
|
"'type': 'data_inspection_failed', 'code': 'data_inspection_failed'}}"),
|
|
])
|
|
def test_alibaba_guardrail_is_a_content_policy_block_not_an_outage(status, message):
|
|
"""Alibaba Model Studio's guardrail refuses this content; retrying the same request repeats the
|
|
refusal, so it must not ride the 5xx / unknown retry ladder."""
|
|
result = classify_api_error(_APIError(message, status), provider="nous", model="qwen/qwen3.8-flash")
|
|
assert result.reason == FailoverReason.content_policy_blocked
|
|
assert result.retryable is False
|