teknium1
a6c4278084
The streaming choke point was covered only by an identity check on the imported alias (a change-detector that would pass with a no-op wiring). Replace it with a real chunk-loop run asserting STOP -> stop and MAX_TOKENS -> length on the assembled response, and collapse the duplicate transport cases into one parametrized invariant.
81 lines
3.2 KiB
Python
81 lines
3.2 KiB
Python
"""Uppercase wire finish reasons (STOP / MAX_TOKENS from Gemini-fronting
|
|
OpenAI-compatible gateways) must fold to the lowercase OpenAI contract at both
|
|
wire-intake choke points — the chat_completions transport and the streaming
|
|
chunk loop — so stop handling and length recovery see the values they compare
|
|
against.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from types import SimpleNamespace
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
import pytest
|
|
|
|
from agent.message_sanitization import normalize_finish_reason
|
|
from agent.transports.chat_completions import ChatCompletionsTransport
|
|
from hermes_constants import PARTIAL_STREAM_STUB_ID
|
|
|
|
|
|
@pytest.mark.parametrize(
|
|
"raw,expected",
|
|
[
|
|
("STOP", "stop"),
|
|
("MAX_TOKENS", "length"),
|
|
("Tool_Calls", "tool_calls"),
|
|
("end", "stop"),
|
|
("function_call", "tool_calls"),
|
|
("content_filter", "content_filter"), # contract values pass through byte-identical
|
|
],
|
|
)
|
|
def test_normalize_finish_reason_folds_to_contract(raw, expected):
|
|
assert normalize_finish_reason(raw) == expected
|
|
|
|
|
|
@pytest.mark.parametrize("raw", [None, "", 24])
|
|
def test_normalize_finish_reason_passes_falsy_and_non_string_unchanged(raw):
|
|
# Callers keep their ``or "stop"`` defaults; Poolside int reasons are untouched.
|
|
assert normalize_finish_reason(raw) is raw
|
|
|
|
|
|
def _fake_response(finish_reason):
|
|
msg = SimpleNamespace(content="hello", tool_calls=None, refusal=None)
|
|
choice = SimpleNamespace(finish_reason=finish_reason, message=msg)
|
|
return SimpleNamespace(choices=[choice], usage=None, model="gemini-3-pro")
|
|
|
|
|
|
@pytest.mark.parametrize("raw,expected", [("STOP", "stop"), ("MAX_TOKENS", "length"), (24, "24"), (None, "stop")])
|
|
def test_transport_normalize_response_folds_finish_reason(raw, expected):
|
|
assert ChatCompletionsTransport().normalize_response(_fake_response(raw)).finish_reason == expected
|
|
|
|
|
|
def _make_stream_chunk(content=None, finish_reason=None):
|
|
delta = SimpleNamespace(content=content, tool_calls=None, reasoning_content=None, reasoning=None)
|
|
return SimpleNamespace(choices=[SimpleNamespace(index=0, delta=delta, finish_reason=finish_reason)],
|
|
model=None, usage=None)
|
|
|
|
|
|
@pytest.mark.parametrize("raw,expected", [("STOP", "stop"), ("MAX_TOKENS", "length")])
|
|
@patch("run_agent.AIAgent._create_request_openai_client")
|
|
@patch("run_agent.AIAgent._close_request_openai_client")
|
|
def test_streaming_capture_folds_uppercase_finish_reason(_mock_close, mock_create, monkeypatch, raw, expected):
|
|
from run_agent import AIAgent
|
|
|
|
def _stream():
|
|
yield _make_stream_chunk(content="partial answer")
|
|
yield _make_stream_chunk(finish_reason=raw)
|
|
|
|
mock_client = MagicMock()
|
|
mock_client.chat.completions.create.side_effect = lambda *a, **kw: _stream()
|
|
mock_create.return_value = mock_client
|
|
monkeypatch.setenv("HERMES_STREAM_RETRIES", "0")
|
|
|
|
agent = AIAgent(api_key="test-key", base_url="https://example.com/v1", model="test/model",
|
|
quiet_mode=True, skip_context_files=True, skip_memory=True)
|
|
agent.api_mode = "chat_completions"
|
|
agent._interrupt_requested = False
|
|
response = agent._interruptible_streaming_api_call({})
|
|
|
|
assert response.id != PARTIAL_STREAM_STUB_ID
|
|
assert response.choices[0].finish_reason == expected
|