L'électron rare
138 lines
5.5 KiB
Python
138 lines
5.5 KiB
Python
from __future__ import annotations
|
|
|
|
import unittest
|
|
|
|
from core.generation.provider import ProviderConfig
|
|
from core.runtime.config import openai_base_url_for_model, runtime_probe_profile
|
|
from core.runtime.health import probe_runtime_health
|
|
from core.runtime.profiles import runtime_probe_name, runtime_profile_name_for_model
|
|
|
|
|
|
class RuntimeProfileTests(unittest.TestCase):
|
|
def test_provider_config_exposes_runtime_constraints_for_apple(self) -> None:
|
|
profile = ProviderConfig(
|
|
provider="openai_compatible",
|
|
base_url="http://127.0.0.1:8100",
|
|
api_key="",
|
|
model="apple-coreml:qwen3.5-4b-onnx-q4f16",
|
|
timeout=30.0,
|
|
max_tokens=512,
|
|
stage_max_tokens={},
|
|
).to_runtime_profile()
|
|
|
|
self.assertTrue(profile.capabilities.requires_manual_model_switch)
|
|
self.assertEqual(profile.capabilities.response_format_mode, "best_effort")
|
|
self.assertEqual(profile.name, "apple_coreml_single_model")
|
|
self.assertTrue(any(item.code == "manual-apple-switch" for item in profile.constraints))
|
|
self.assertTrue(any(item.code == "json-best-effort" for item in profile.constraints))
|
|
|
|
def test_runtime_profile_with_model_keeps_manual_switch_flag_for_cross_apple_switch(self) -> None:
|
|
profile = ProviderConfig(
|
|
provider="openai_compatible",
|
|
base_url="http://127.0.0.1:8100",
|
|
api_key="",
|
|
model="apple-coreml:qwen2.5-0.5b-instruct-onnx",
|
|
timeout=30.0,
|
|
max_tokens=512,
|
|
stage_max_tokens={},
|
|
).to_runtime_profile()
|
|
|
|
updated = profile.with_model("apple-coreml:qwen3.5-4b-onnx-q4f16")
|
|
self.assertTrue(updated.capabilities.requires_manual_model_switch)
|
|
|
|
|
|
class RuntimeHealthTests(unittest.TestCase):
|
|
def test_probe_runtime_health_can_read_model_catalog(self) -> None:
|
|
profile = ProviderConfig(
|
|
provider="openai_compatible",
|
|
base_url="http://127.0.0.1:8100",
|
|
api_key="",
|
|
model="ollama:qwen2.5:7b",
|
|
timeout=30.0,
|
|
max_tokens=512,
|
|
stage_max_tokens={},
|
|
).to_runtime_profile()
|
|
|
|
class FakeResponse:
|
|
def __enter__(self):
|
|
return self
|
|
|
|
def __exit__(self, exc_type, exc, tb):
|
|
return False
|
|
|
|
def read(self):
|
|
return b'{\"status\": \"ok\"}'
|
|
|
|
def fetch(url: str, timeout: float):
|
|
if url.endswith("/v1/models"):
|
|
return {"data": [{"id": "ollama:qwen2.5:7b"}]}
|
|
raise AssertionError(url)
|
|
|
|
health = probe_runtime_health(profile, opener=lambda *args, **kwargs: FakeResponse(), json_fetcher=fetch)
|
|
self.assertTrue(health.ok)
|
|
self.assertEqual(health.available_models, ["ollama:qwen2.5:7b"])
|
|
|
|
def test_probe_runtime_health_reports_failure(self) -> None:
|
|
profile = ProviderConfig(
|
|
provider="openai_compatible",
|
|
base_url="http://127.0.0.1:8100",
|
|
api_key="",
|
|
model="ollama:qwen2.5:7b",
|
|
timeout=30.0,
|
|
max_tokens=512,
|
|
stage_max_tokens={},
|
|
).to_runtime_profile()
|
|
|
|
health = probe_runtime_health(profile, opener=lambda *args, **kwargs: (_ for _ in ()).throw(OSError("down")))
|
|
self.assertFalse(health.ok)
|
|
self.assertIn("down", health.detail or "")
|
|
|
|
|
|
class RuntimeConfigHelpersTests(unittest.TestCase):
|
|
def test_runtime_probe_profile_uses_runtime_probe_defaults(self) -> None:
|
|
profile = runtime_probe_profile("http://127.0.0.1:8100", timeout=7.0)
|
|
|
|
self.assertEqual(profile.base_url, "http://127.0.0.1:8100")
|
|
self.assertEqual(profile.model, "runtime-probe")
|
|
self.assertEqual(profile.timeout, 7.0)
|
|
self.assertEqual(profile.name, "openai_probe")
|
|
|
|
def test_openai_base_url_for_model_prefers_ollama_openai_runtime_when_enabled(self) -> None:
|
|
self.assertEqual(
|
|
openai_base_url_for_model(
|
|
"ollama:qwen2.5:7b",
|
|
core_base_url="http://127.0.0.1:8100",
|
|
ollama_runtime="openai_compatible",
|
|
ollama_openai_base_url="http://127.0.0.1:8091",
|
|
),
|
|
"http://127.0.0.1:8091",
|
|
)
|
|
|
|
def test_openai_base_url_for_model_defaults_to_core_runtime(self) -> None:
|
|
self.assertEqual(
|
|
openai_base_url_for_model(
|
|
"apple-coreml:qwen3.5-4b-onnx-q4f16",
|
|
core_base_url="http://127.0.0.1:8100",
|
|
ollama_runtime="openai_compatible",
|
|
ollama_openai_base_url="http://127.0.0.1:8091",
|
|
),
|
|
"http://127.0.0.1:8100",
|
|
)
|
|
|
|
def test_runtime_profile_name_for_model_maps_known_runtimes(self) -> None:
|
|
self.assertEqual(runtime_profile_name_for_model("apple-coreml:qwen3.5-4b-onnx-q4f16"), "apple_coreml_single_model")
|
|
self.assertEqual(runtime_profile_name_for_model("ollama:qwen2.5:7b"), "ollama_openai_compatible")
|
|
self.assertEqual(
|
|
runtime_profile_name_for_model("ollama:qwen2.5:7b", ollama_runtime="native"),
|
|
"ollama_native",
|
|
)
|
|
|
|
def test_runtime_probe_name_maps_named_probe_profiles(self) -> None:
|
|
self.assertEqual(runtime_probe_name("core"), "mascarade_local")
|
|
self.assertEqual(runtime_probe_name("apple"), "apple_coreml_single_model")
|
|
self.assertEqual(runtime_probe_name("ollama_openai"), "llama_cpp_local")
|
|
self.assertEqual(
|
|
runtime_probe_name("remote", base_url="http://tower.example.test:8100"),
|
|
"mascarade_remote_tower_example_test",
|
|
)
|