teknium1
dfd4aa4a94
When a proxy (Ollama, OpenRouter) rejects the MODEL's own unparseable tool-call JSON with `400 invalid tool call arguments`, the classifier returned the generic format_error verdict (`should_fallback=True`) and the non-retryable client-error path cascaded through every fallback provider: 4-5 sequential calls, 20-60s per occurrence, ending on a model that produced the same broken JSON (#12770). - error_classifier: explicit `_MALFORMED_TOOL_ARGS_PATTERNS` checked before the request-validation and overflow heuristics, returning format_error with `retryable=False, should_fallback=False`. - turn_api_error: the client-error settlement honours `should_fallback`; the verdicts that legitimately reach that branch (policy block, TLS chain, MoA shape/preset errors) now state `should_fallback=True` explicitly, so the gate changes behaviour only for the new verdict. Local validation errors keep their historical fallback. Fixes #12770. Pattern list and gating approach from #16022 by @cuyua9 (stale base); tests trimmed to two invariants. Co-authored-by: cuyua9 <2114364329@qq.com>
76 lines
3.0 KiB
Python
76 lines
3.0 KiB
Python
"""A 400 for the model's own malformed tool-call JSON must not walk the fallback chain (#12770)."""
|
|
from types import SimpleNamespace
|
|
from unittest.mock import patch
|
|
|
|
import pytest
|
|
|
|
from agent.error_classifier import FailoverReason, classify_api_error
|
|
from agent.turn_api_error import settle_unrecovered_error
|
|
|
|
|
|
class _Err(Exception):
|
|
status_code = 400
|
|
response = None
|
|
|
|
def __init__(self, message, body=None):
|
|
super().__init__(message)
|
|
self.body = body or {"error": {"message": message, "type": "invalid_request_error"}}
|
|
|
|
|
|
@pytest.mark.parametrize("wording", ["invalid tool call arguments", "Invalid function_call arguments"])
|
|
def test_malformed_tool_args_400_is_terminal_without_fallback(wording):
|
|
"""Large session included: the bare-ish message must not read as context overflow either."""
|
|
verdict = classify_api_error(_Err(f"Error code: 400 - {wording}"), provider="ollama",
|
|
approx_tokens=90_000, context_length=128_000, num_messages=120)
|
|
assert verdict.reason is FailoverReason.format_error
|
|
assert (verdict.retryable, verdict.should_compress, verdict.should_fallback) == (False, False, False)
|
|
|
|
# Unrelated request-shape 400s keep their fallback (another provider may accept the request).
|
|
assert classify_api_error(_Err("Unsupported parameter: 'max_tokens'")).should_fallback is True
|
|
|
|
|
|
class _Agent:
|
|
"""Only the fallback seam is real; every other helper the terminal path touches is a no-op."""
|
|
log_prefix = ""
|
|
verbose = False
|
|
provider = "ollama"
|
|
_fallback_chain = [object()]
|
|
_fallback_index = 0
|
|
_credential_pool = None
|
|
|
|
def __init__(self):
|
|
self.activated = []
|
|
|
|
def _has_pending_fallback(self):
|
|
return True
|
|
|
|
def _try_activate_fallback(self, **kwargs):
|
|
self.activated.append(True)
|
|
return True
|
|
|
|
def _summarize_api_error(self, error):
|
|
return str(error)
|
|
|
|
def __getattr__(self, name):
|
|
return lambda *args, **kwargs: None
|
|
|
|
|
|
def test_client_error_settlement_skips_fallback_when_classifier_says_so():
|
|
agent = _Agent()
|
|
retry = SimpleNamespace(copilot_stale_cred_retry_attempted=False, primary_recovery_attempted=False)
|
|
err = _Err("invalid tool call arguments")
|
|
classified = classify_api_error(err, provider="ollama")
|
|
|
|
with patch("agent.conversation_loop._is_copilot_provider", lambda a: False):
|
|
verdict = settle_unrecovered_error(
|
|
agent, api_error=err, classified=classified, _retry=retry, status_code=400, error_msg=str(err),
|
|
is_context_length_error=False, is_rate_limited=False, _is_zai_coding_overload=False,
|
|
_provider="ollama", _base="http://127.0.0.1:11434/v1", _model="glm", messages=[], api_messages=[],
|
|
api_kwargs={}, active_system_prompt="", conversation_history=None, approx_tokens=10,
|
|
retry_count=0, max_retries=3, compression_attempts=0, api_call_count=1,
|
|
)
|
|
|
|
assert verdict.action == "return"
|
|
assert verdict.result["failure_reason"] == FailoverReason.format_error.value
|
|
assert agent.activated == []
|