fix(cost): keep cache totals aligned with resolved rates

This commit is contained in:
liming 2026-09-14 16:13:09 +08:00
parent b9db42ac1c
commit bfca7aa236
3 changed files with 40 additions and 7 deletions

View file

@ -60,9 +60,6 @@ from litellm.llms.databricks.cost_calculator import (
from litellm.llms.deepseek.cost_calculator import (
cost_per_token as deepseek_cost_per_token,
)
from litellm.llms.fireworks_ai.cost_calculator import (
cost_per_token as fireworks_ai_cost_per_token,
)
from litellm.llms.gemini.cost_calculator import cost_per_token as gemini_cost_per_token
from litellm.llms.lemonade.cost_calculator import (
cost_per_token as lemonade_cost_per_token,
@ -76,9 +73,6 @@ from litellm.llms.openai.cost_calculation import (
from litellm.llms.openai.cost_calculation import (
cost_per_token as openai_cost_per_token,
)
from litellm.llms.perplexity.cost_calculator import (
cost_per_token as perplexity_cost_per_token,
)
from litellm.llms.tencent.cost_calculator import (
cost_per_token as tencent_cost_per_token,
)
@ -647,6 +641,10 @@ def cost_per_token(
elif custom_llm_provider == "databricks":
return databricks_cost_per_token(model=model, usage=usage_block)
elif custom_llm_provider == "fireworks_ai":
from litellm.llms.fireworks_ai.cost_calculator import (
cost_per_token as fireworks_ai_cost_per_token,
)
return fireworks_ai_cost_per_token(model=model, usage=usage_block)
elif custom_llm_provider == "azure":
return azure_openai_cost_per_token(
@ -662,6 +660,10 @@ def cost_per_token(
elif custom_llm_provider == "tencent":
return tencent_cost_per_token(model=model, usage=usage_block)
elif custom_llm_provider == "perplexity":
from litellm.llms.perplexity.cost_calculator import (
cost_per_token as perplexity_cost_per_token,
)
return perplexity_cost_per_token(model=model, usage=usage_block)
elif custom_llm_provider == "xai":
return xai_cost_per_token(model=model, usage=usage_block)

View file

@ -1283,12 +1283,19 @@ def generic_cost_per_token(
current_time=billing_time,
threshold_is_inclusive=_uses_inclusive_token_thresholds(custom_llm_provider),
)
resolved_cache_read_cost: Final = _resolve_cache_read_cost_rate(
model_info=model_info,
usage=usage,
current_time=billing_time,
service_tier=service_tier,
)
effective_cache_read_cost: Final = cache_read_cost if resolved_cache_read_cost is None else resolved_cache_read_cost
prompt_cost = _calculate_input_cost(
prompt_tokens_details=prompt_tokens_details,
model_info=model_info,
prompt_base_cost=prompt_base_cost,
cache_read_cost=cache_read_cost,
cache_read_cost=effective_cache_read_cost,
cache_creation_cost=cache_creation_cost,
cache_creation_cost_above_1hr=cache_creation_cost_above_1hr,
service_tier=service_tier,

View file

@ -19,6 +19,7 @@ import litellm
from litellm.litellm_core_utils.llm_cost_calc.utils import (
_resolve_cache_read_cost_rate,
calculate_prompt_caching_savings,
generic_cost_per_token,
get_token_type_cost_breakdown,
)
from litellm.llms.dashscope.cost_calculator import (
@ -404,6 +405,29 @@ class TestDashscopeCostCalculator:
assert math.isclose(savings, 600 * (1e-06 - 2e-07), rel_tol=1e-10)
def test_openai_compatible_qwen_total_uses_implicit_cache_read_rate(self):
model_info: ModelInfo = {
"key": "qwen-generic-cache-mode-test",
"input_cost_per_token": 1e-06,
"output_cost_per_token": 4e-06,
"cache_read_input_token_cost": 1e-07,
"implicit_cache_read_input_token_cost": 2e-07,
}
usage = Usage(
prompt_tokens=1000,
completion_tokens=0,
prompt_tokens_details={"cached_tokens": 600},
)
prompt_cost, _ = generic_cost_per_token(
model="custom-qwen",
usage=usage,
custom_llm_provider="openai",
model_info=model_info,
)
assert math.isclose(prompt_cost, (400 * 1e-06) + (600 * 2e-07), rel_tol=1e-10)
@pytest.mark.parametrize(
("cache_type", "expected_cache_rate"),
[("ephemeral", 1e-07), (None, 2e-07)],