88 lines
2.9 KiB
Python
88 lines
2.9 KiB
Python
from unittest.mock import Mock, patch
|
|
|
|
import httpx
|
|
|
|
from services.chat_model_config import (
|
|
clear_chat_model_config_cache,
|
|
get_chat_model_config,
|
|
)
|
|
|
|
|
|
def setup_function():
|
|
clear_chat_model_config_cache()
|
|
|
|
|
|
def teardown_function():
|
|
clear_chat_model_config_cache()
|
|
|
|
|
|
def test_admin_runtime_config_takes_priority(monkeypatch):
|
|
monkeypatch.setenv("CHAT_MODEL_CONFIG_URL", "http://config.test/runtime")
|
|
monkeypatch.setenv("AVATAR_MODEL_CONFIG_TOKEN", "shared-secret")
|
|
response = Mock()
|
|
response.raise_for_status.return_value = None
|
|
response.json.return_value = {
|
|
"data": {
|
|
"api_base_url": "https://model.test/v1/",
|
|
"api_key": "runtime-key",
|
|
"model": "avatar-model",
|
|
"vision_model": "avatar-vision-model",
|
|
"ocr_model": "avatar-ocr-model",
|
|
"max_tokens": 2048,
|
|
"timeout_seconds": 42,
|
|
}
|
|
}
|
|
|
|
with patch("services.chat_model_config.httpx.get", return_value=response) as request:
|
|
config = get_chat_model_config()
|
|
|
|
assert config.source == "admin"
|
|
assert config.api_base_url == "https://model.test/v1"
|
|
assert config.model == "avatar-model"
|
|
assert config.vision_model == "avatar-vision-model"
|
|
assert config.ocr_model == "avatar-ocr-model"
|
|
assert config.max_tokens == 2048
|
|
request.assert_called_once_with(
|
|
"http://config.test/runtime",
|
|
headers={"X-Avatar-Config-Token": "shared-secret"},
|
|
timeout=5.0,
|
|
)
|
|
|
|
|
|
def test_runtime_failure_falls_back_to_environment(monkeypatch):
|
|
monkeypatch.setenv("CHAT_MODEL_CONFIG_URL", "http://config.test/runtime")
|
|
monkeypatch.setenv("AVATAR_MODEL_CONFIG_TOKEN", "shared-secret")
|
|
monkeypatch.setenv("CHAT_API_URL", "https://fallback.test/v1/")
|
|
monkeypatch.setenv("CHAT_API_KEY", "fallback-key")
|
|
monkeypatch.setenv("CHAT_MODEL", "fallback-model")
|
|
monkeypatch.setenv("VISION_MODEL", "fallback-vision")
|
|
monkeypatch.setenv("VISION_OCR_MODEL", "fallback-ocr")
|
|
monkeypatch.setenv("CHAT_MAX_OUTPUT_TOKENS", "1536")
|
|
|
|
request = httpx.Request("GET", "http://config.test/runtime")
|
|
with patch(
|
|
"services.chat_model_config.httpx.get",
|
|
side_effect=httpx.ConnectError("offline", request=request),
|
|
):
|
|
config = get_chat_model_config()
|
|
|
|
assert config.source == "environment"
|
|
assert config.api_base_url == "https://fallback.test/v1"
|
|
assert config.api_key == "fallback-key"
|
|
assert config.model == "fallback-model"
|
|
assert config.vision_model == "fallback-vision"
|
|
assert config.ocr_model == "fallback-ocr"
|
|
assert config.max_tokens == 1536
|
|
|
|
|
|
def test_runtime_config_is_cached(monkeypatch):
|
|
monkeypatch.setenv("CHAT_MODEL_CONFIG_URL", "")
|
|
monkeypatch.setenv("CHAT_MODEL", "first-model")
|
|
first = get_chat_model_config()
|
|
monkeypatch.setenv("CHAT_MODEL", "second-model")
|
|
|
|
second = get_chat_model_config()
|
|
|
|
assert first is second
|
|
assert second.model == "first-model"
|