2a95791992
run_agent.py: delete the `# noqa: F401` re-export block (agent.process_bootstrap
OpenAI/_SafeWriter/_get_proxy_*, model_tools get_tool_definitions/
handle_function_call/check_toolset_requirements, FailoverReason,
_qwen_portal_headers/_routermint_headers, session_persistence names,
estimate_request_tokens_rough, ContextCompressor + friends, jittered_backoff,
prompt_builder names, message_sanitization names, tool_dispatch_helpers
names) — 41 names run_agent never used itself — and the `_STREAM_DIAG_HEADERS`
back-compat class alias (no in-tree reader). run_agent now imports only what
it uses (get_toolset_for_tool, is_local_endpoint, coalesce/uniquify tool-call
ids, cleanup_vm/get_active_env from terminal_tool_lifecycle).
agent/*: `_ra().X` late-binds that only reached a re-export now import the
defining module directly (agent_runtime_helpers -> process_bootstrap.OpenAI,
model_tools.handle_function_call, session_persistence._safe_session_filename_component;
agent_init -> model_tools.get_tool_definitions/check_toolset_requirements,
_lazy_headers("agent.client_lifecycle", ...) for qwen/routermint;
system_prompt -> agent.prompt_builder / model_tools directly, dropping its
own _ra() shim and the `_r` parameter threading). `_ra()` stays for
run_agent-resident names (logger, AIAgent, _hermes_home, _set_interrupt, ...).
toolsets.py: remove resolve_multiple_toolsets (shim-only, restored by
34abf954bd); tests/test_toolsets.py pins the same union behavior via
resolve_toolset over each name.
providers/__init__.py: drop the OMIT_TEMPERATURE re-export (no callers via the
package); ProviderProfile stays because __init__ uses it for annotations —
2 tests repointed to providers.base.
agent/iteration_budget.py: drop the "run_agent re-exports the class"
docstring pointer; 4 tests import IterationBudget from its home.
model_tools.py (arg_coercion names), agent/tool_executor.py, and
hermes_cli/cli_session_mixin.py repoints landed via a sibling commit on this
shared worktree.
Callers repointed: gateway/run.py, hermes_cli/cli_chat_turn_mixin.py,
hermes_cli/cli_tui_mixin.py, tui_gateway/session_workdir.py,
agent/transports/codex.py (one-line imports) + comment pointers in
tools/file_state.py, tools/schema_sanitizer.py, scripts/tool_search_livetest.py.
Tests: patch("run_agent.X") / monkeypatch.setattr(run_agent, "X") /
`from run_agent import X` -> defining module across 99 test files.
124 lines
4.8 KiB
Python
124 lines
4.8 KiB
Python
"""Auth-failure provider failover (conversation loop).
|
|
|
|
A 401/403 that survives the per-provider credential-refresh attempt
|
|
(revoked OAuth, blocked/expired key, an account pinned to a dead/staging
|
|
endpoint) must escalate to the configured fallback chain instead of
|
|
thrashing on the same dead credential every turn.
|
|
|
|
Before the fix, the conversation loop's generic failover dispatch only
|
|
fired for ``{rate_limit, billing}`` reasons; ``auth`` / ``auth_permanent``
|
|
fell through to "switch providers manually" advice and never called
|
|
``_try_activate_fallback()``. These tests pin:
|
|
|
|
1. 401/403 classify as auth (``classified.is_auth`` True).
|
|
2. ``_try_activate_fallback`` advances the chain on an auth reason.
|
|
3. The one-shot guard flag exists on TurnRetryState.
|
|
"""
|
|
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
from run_agent import AIAgent
|
|
from agent.error_classifier import classify_api_error, FailoverReason
|
|
from agent.turn_retry_state import TurnRetryState
|
|
|
|
|
|
def _make_agent(fallback_model=None):
|
|
with (
|
|
patch("model_tools.get_tool_definitions", return_value=[]),
|
|
patch("model_tools.check_toolset_requirements", return_value={}),
|
|
patch("agent.process_bootstrap.OpenAI"),
|
|
):
|
|
agent = AIAgent(
|
|
api_key="test-key",
|
|
base_url="https://openrouter.ai/api/v1",
|
|
quiet_mode=True,
|
|
skip_context_files=True,
|
|
skip_memory=True,
|
|
fallback_model=fallback_model,
|
|
)
|
|
agent.client = MagicMock()
|
|
return agent
|
|
|
|
|
|
def _mock_client(base_url="https://openrouter.ai/api/v1", api_key="fb-key"):
|
|
mock = MagicMock()
|
|
mock.base_url = base_url
|
|
mock.api_key = api_key
|
|
return mock
|
|
|
|
|
|
def _auth_error(status=401, msg="Your API key is invalid, blocked or out of funds."):
|
|
err = Exception(f"Error code: {status} - {msg}")
|
|
err.status_code = status
|
|
return err
|
|
|
|
|
|
class TestAuthErrorClassification:
|
|
def test_401_is_auth(self):
|
|
c = classify_api_error(_auth_error(401))
|
|
assert c.reason in {FailoverReason.auth, FailoverReason.auth_permanent}
|
|
assert c.is_auth is True
|
|
|
|
|
|
def test_500_is_not_auth(self):
|
|
err = Exception("Error code: 500 - internal server error")
|
|
err.status_code = 500
|
|
c = classify_api_error(err)
|
|
assert c.is_auth is False
|
|
|
|
|
|
class TestAuthFailoverGuardFlag:
|
|
def test_flag_defaults_false(self):
|
|
assert TurnRetryState().auth_failover_attempted is False
|
|
|
|
|
|
class TestAuthFailoverActivation:
|
|
"""The decision the loop makes on a persistent auth failure: when a
|
|
fallback chain exists and the guard hasn't fired, escalate to it."""
|
|
|
|
def _should_failover(self, agent, classified, retry):
|
|
# Mirror the exact gating condition added to conversation_loop.py.
|
|
return (
|
|
classified.is_auth
|
|
and not retry.auth_failover_attempted
|
|
and agent._fallback_index < len(agent._fallback_chain)
|
|
)
|
|
|
|
def test_auth_failover_fires_when_chain_present(self):
|
|
agent = _make_agent(fallback_model=[{"provider": "openai", "model": "gpt-4o"}])
|
|
retry = TurnRetryState()
|
|
classified = classify_api_error(_auth_error(401))
|
|
assert self._should_failover(agent, classified, retry) is True
|
|
# And the activation primitive actually advances on an auth reason.
|
|
with patch(
|
|
"agent.auxiliary_client.resolve_provider_client",
|
|
return_value=(_mock_client(), "gpt-4o"),
|
|
):
|
|
advanced = agent._try_activate_fallback(reason=classified.reason)
|
|
assert advanced is True
|
|
assert agent._fallback_index == 1
|
|
|
|
def test_no_failover_without_chain(self):
|
|
"""A user with no fallback configured (the common case for the
|
|
original incident) does NOT failover — falls through to the
|
|
existing terminal handling + troubleshooting advice."""
|
|
agent = _make_agent(fallback_model=None)
|
|
retry = TurnRetryState()
|
|
classified = classify_api_error(_auth_error(401))
|
|
assert self._should_failover(agent, classified, retry) is False
|
|
|
|
def test_guard_blocks_repeat_failover(self):
|
|
agent = _make_agent(fallback_model=[{"provider": "openai", "model": "gpt-4o"}])
|
|
retry = TurnRetryState()
|
|
retry.auth_failover_attempted = True # already escalated this attempt
|
|
classified = classify_api_error(_auth_error(401))
|
|
assert self._should_failover(agent, classified, retry) is False
|
|
|
|
def test_non_auth_error_does_not_trigger_auth_failover(self):
|
|
agent = _make_agent(fallback_model=[{"provider": "openai", "model": "gpt-4o"}])
|
|
retry = TurnRetryState()
|
|
err = Exception("Error code: 500 - internal server error")
|
|
err.status_code = 500
|
|
classified = classify_api_error(err)
|
|
assert self._should_failover(agent, classified, retry) is False
|