mirror of
https://github.com/GNS3/gns3-server.git
synced 2026-11-19 20:55:14 +02:00
A field log showed one request stayed silent for 604s (the OpenAI SDK default timeout), then the SDK retried on a fresh connection and got 200 OK in under a second. The abort signal did not cancel the in-flight HTTP call, so the turn could look hung for ten minutes. - set timeout (default 120s, request_timeout in the LLM config wins), max_retries=1 and stream_usage=True for the providers that accept them (openai, deepseek, anthropic, xai). ollama and google-genai reject these kwargs, so they keep the old behavior. - log the elapsed time of each LLM call. A slow first token or an SDK retry now shows up in the log line. - add request_timeout to the LLM config schemas and docs (JSON column, no migration). - tests: cover the defaults, the override, invalid values, provider filtering and the title model.
78 lines
2.6 KiB
Python
78 lines
2.6 KiB
Python
"""
|
|
Tests for the model factory request controls.
|
|
|
|
A silent provider connection used to wait for the OpenAI SDK default of
|
|
600s before the SDK retried. The factory now sets a shorter timeout and
|
|
a bounded retry count for the providers that accept them, and keeps usage
|
|
metadata in streaming mode.
|
|
"""
|
|
|
|
from typing import Any
|
|
|
|
import pytest
|
|
|
|
from gns3server.agent.gns3_copilot.agent import model_factory
|
|
|
|
|
|
def _config(**overrides: Any) -> dict[str, Any]:
|
|
config = {
|
|
"provider": "deepseek",
|
|
"model": "deepseek-v4-flash",
|
|
"api_key": "test-key",
|
|
"base_url": "https://api.deepseek.com/v1",
|
|
"temperature": 0.3,
|
|
}
|
|
config.update(overrides)
|
|
return config
|
|
|
|
|
|
@pytest.fixture
|
|
def captured(monkeypatch: pytest.MonkeyPatch) -> dict[str, Any]:
|
|
"""Capture the kwargs passed to init_chat_model."""
|
|
box: dict[str, Any] = {}
|
|
|
|
def fake_init_chat_model(**kwargs: Any) -> object:
|
|
box.update(kwargs)
|
|
return object()
|
|
|
|
monkeypatch.setattr(model_factory, "init_chat_model", fake_init_chat_model)
|
|
return box
|
|
|
|
|
|
def test_base_model_gets_timeout_and_bounded_retries(captured: dict[str, Any]) -> None:
|
|
model_factory.create_base_model(_config())
|
|
|
|
assert captured["timeout"] == model_factory.DEFAULT_REQUEST_TIMEOUT_S
|
|
assert captured["max_retries"] == model_factory.DEFAULT_MAX_RETRIES
|
|
assert captured["stream_usage"] is True
|
|
# DeepSeek thinking stays disabled
|
|
assert captured["extra_body"] == {"thinking": {"type": "disabled"}}
|
|
|
|
|
|
def test_request_timeout_from_config_wins(captured: dict[str, Any]) -> None:
|
|
model_factory.create_base_model(_config(request_timeout="60"))
|
|
assert captured["timeout"] == 60.0
|
|
|
|
|
|
@pytest.mark.parametrize("value", ["abc", 0, -5, None])
|
|
def test_invalid_request_timeout_falls_back(captured: dict[str, Any], value: Any) -> None:
|
|
model_factory.create_base_model(_config(request_timeout=value))
|
|
assert captured["timeout"] == model_factory.DEFAULT_REQUEST_TIMEOUT_S
|
|
|
|
|
|
def test_providers_without_request_controls_get_none(captured: dict[str, Any]) -> None:
|
|
"""Ollama and google-genai reject these kwargs at construction."""
|
|
model_factory.create_base_model(_config(provider="ollama", base_url="http://localhost:11434"))
|
|
|
|
assert "timeout" not in captured
|
|
assert "max_retries" not in captured
|
|
assert "stream_usage" not in captured
|
|
|
|
|
|
def test_title_model_gets_the_same_controls(captured: dict[str, Any]) -> None:
|
|
model_factory.create_title_model(_config())
|
|
|
|
assert captured["timeout"] == model_factory.DEFAULT_REQUEST_TIMEOUT_S
|
|
assert captured["max_retries"] == model_factory.DEFAULT_MAX_RETRIES
|
|
assert captured["stream_usage"] is True
|