refactor: 清理 litellm 冗余代码并锁定依赖版本

backend/tests/unit/services/test_llm_pricing_service.py

@@ -2,14 +2,14 @@ from __future__ import annotations
 
 import pytest
 
-from services.litellm.service import LiteLLMService
+from services.llm_pricing.service import LlmPricingService
 
 
 def test_calculate_cost_uses_first_qwen_tier() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     cost = service.calculate_cost(
-        model="dashscope/qwen3.5-flash",
+        model="qwen3.5-flash",
         prompt_tokens=100_000,
         completion_tokens=1_000,
         cached_prompt_tokens=10_000,
@@ -19,10 +19,10 @@ def test_calculate_cost_uses_first_qwen_tier() -> None:
 
 
 def test_calculate_cost_uses_second_qwen_tier() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     cost = service.calculate_cost(
-        model="dashscope/qwen3.5-flash",
+        model="qwen3.5-flash",
         prompt_tokens=200_000,
         completion_tokens=5_000,
         cached_prompt_tokens=20_000,
@@ -32,10 +32,10 @@ def test_calculate_cost_uses_second_qwen_tier() -> None:
 
 
 def test_build_usage_metadata_calculates_cost_from_usage_summary() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     metadata = service.build_usage_metadata(
-        model="dashscope/qwen3.5-flash",
+        model="qwen3.5-flash",
         usage_summary={
             "input_tokens": 2000,
             "output_tokens": 100,
@@ -44,7 +44,7 @@ def test_build_usage_metadata_calculates_cost_from_usage_summary() -> None:
         },
     )
 
-    assert metadata["model"] == "dashscope/qwen3.5-flash"
+    assert metadata["model"] == "qwen3.5-flash"
     assert metadata["inputTokens"] == 2000
     assert metadata["outputTokens"] == 100
     assert metadata["totalTokens"] == 2100
@@ -59,7 +59,7 @@ def test_build_usage_metadata_calculates_cost_from_usage_summary() -> None:
 
 
 def test_build_usage_metadata_prefers_provider_direct_cost() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     metadata = service.build_usage_metadata(
         model="deepseek-chat",
@@ -80,7 +80,7 @@ def test_build_usage_metadata_prefers_provider_direct_cost() -> None:
 
 
 def test_build_usage_metadata_falls_back_when_provider_cost_incomplete() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     metadata = service.build_usage_metadata(
         model="deepseek-chat",
@@ -100,7 +100,7 @@ def test_build_usage_metadata_falls_back_when_provider_cost_incomplete() -> None
 
 
 def test_build_usage_metadata_marks_incomplete_usage_fallback() -> None:
-    service = LiteLLMService()
+    service = LlmPricingService()
 
     metadata = service.build_usage_metadata(
         model="deepseek-chat",