scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-16 23:41:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 102

style(fireworks-ai): apply repository conventions to the cost component change

Browse source 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
kerry 2026-09-16 01:39:24 +00:00
parent d1aa37b0a8
commit 8352045f32
2 changed files with 57 additions and 36 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

31
litellm/llms/fireworks_ai/cost_calculator.py
View file

@ -68,24 +68,25 @@ def _resolve_model_info(model: str) -> ModelInfo:
def _with_cache_read_fallback(model_info: ModelInfo) -> ModelInfo:
"""Most fireworks_ai price-map entries publish no cache-read rate, and the provider bills
cached reads at the input rate. generic_cost_per_token prices a missing rate at $0, so the
fallback is written into a copy of the entry (the shared model-cost dict must not be
mutated), including inside off_peak_pricing so cached reads track the off-peak input rate
the way the previous calculator did."""
if model_info.get("cache_read_input_token_cost") is not None:
return model_info
"""Most fireworks_ai price-map entries publish no cache-read rate though the provider bills
cached reads at the input rate; the shared map is never mutated, so a copy carries the fallback."""
input_rate: Final = model_info.get("input_cost_per_token")
if input_rate is None:
if model_info.get("cache_read_input_token_cost") is not None or input_rate is None:
return model_info
effective: Final[dict[str, object]] = dict(model_info)
effective["cache_read_input_token_cost"] = input_rate
off_peak: Final = model_info.get("off_peak_pricing")
if off_peak is not None and "cache_read_input_token_cost" not in off_peak:
off_peak_copy: Final[dict[str, object]] = dict(off_peak)
off_peak_copy["cache_read_input_token_cost"] = off_peak_copy.get("input_cost_per_token", input_rate)
effective["off_peak_pricing"] = off_peak_copy
return cast(ModelInfo, effective)
if off_peak is None or "cache_read_input_token_cost" in off_peak:
return cast(ModelInfo, {**model_info, "cache_read_input_token_cost": input_rate})
return cast(
ModelInfo,
{
**model_info,
"cache_read_input_token_cost": input_rate,
"off_peak_pricing": {
**off_peak,
"cache_read_input_token_cost": off_peak.get("input_cost_per_token", input_rate),
},
},
)
def cost_per_token(model: str, usage: Usage, current_time: datetime | None = None) -> tuple[float, float]:

62
tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_cost_calculator.py
View file

@ -1,5 +1,6 @@
import math
from datetime import datetime, timezone
from typing import Final
import pytest
@ -51,13 +52,16 @@ STANDARD_CACHE_READ_COST = 1.5e-08
def _register_off_peak_model(
off_peak_pricing: OffPeakPricing, cache_read_cost: float | None = STANDARD_CACHE_READ_COST
) -> None:
litellm.model_cost[f"fireworks_ai/{OFF_PEAK_MODEL}"] = {
"litellm_provider": "fireworks_ai",
"mode": "chat",
"input_cost_per_token": STANDARD_INPUT_COST,
"output_cost_per_token": STANDARD_OUTPUT_COST,
"off_peak_pricing": off_peak_pricing,
**({} if cache_read_cost is None else {"cache_read_input_token_cost": cache_read_cost}),
litellm.model_cost = { # test-quality-ok: the save/restore conftest returns litellm.model_cost to the original object after each test, so replacing the map for this entry leaks nothing
**litellm.model_cost,
f"fireworks_ai/{OFF_PEAK_MODEL}": {
"litellm_provider": "fireworks_ai",
"mode": "chat",
"input_cost_per_token": STANDARD_INPUT_COST,
"output_cost_per_token": STANDARD_OUTPUT_COST,
"off_peak_pricing": off_peak_pricing,
**({} if cache_read_cost is None else {"cache_read_input_token_cost": cache_read_cost}),
},
}
@ -145,20 +149,19 @@ COMPONENT_AUDIO_OUT_COST = 6e-06
def test_cache_write_reasoning_and_audio_tokens_are_billed_at_their_component_rates():
"""Regression (LIT-7837): the hand-rolled fireworks_ai calculator billed every
cache-creation, reasoning and audio token at $0. The shared calculator treats the
prompt detail counts as subsets of prompt_tokens and the completion detail counts as
subsets of completion_tokens, billing each remainder at the text rate."""
litellm.model_cost[f"fireworks_ai/{COMPONENT_MODEL}"] = {
"litellm_provider": "fireworks_ai",
"mode": "chat",
"input_cost_per_token": COMPONENT_INPUT_COST,
"output_cost_per_token": COMPONENT_OUTPUT_COST,
"cache_read_input_token_cost": COMPONENT_CACHE_READ_COST,
"cache_creation_input_token_cost": COMPONENT_CACHE_CREATION_COST,
"output_cost_per_reasoning_token": COMPONENT_REASONING_COST,
"input_cost_per_audio_token": COMPONENT_AUDIO_IN_COST,
"output_cost_per_audio_token": COMPONENT_AUDIO_OUT_COST,
litellm.model_cost = { # test-quality-ok: the save/restore conftest returns litellm.model_cost to the original object after each test, so replacing the map for this entry leaks nothing
**litellm.model_cost,
f"fireworks_ai/{COMPONENT_MODEL}": {
"litellm_provider": "fireworks_ai",
"mode": "chat",
"input_cost_per_token": COMPONENT_INPUT_COST,
"output_cost_per_token": COMPONENT_OUTPUT_COST,
"cache_read_input_token_cost": COMPONENT_CACHE_READ_COST,
"cache_creation_input_token_cost": COMPONENT_CACHE_CREATION_COST,
"output_cost_per_reasoning_token": COMPONENT_REASONING_COST,
"input_cost_per_audio_token": COMPONENT_AUDIO_IN_COST,
"output_cost_per_audio_token": COMPONENT_AUDIO_OUT_COST,
},
}
usage = Usage(
prompt_tokens=1000,
@ -188,3 +191,20 @@ def test_cache_write_reasoning_and_audio_tokens_are_billed_at_their_component_ra
)
assert prompt_cost == pytest.approx(expected_prompt_cost)
assert completion_cost == pytest.approx(expected_completion_cost)
def test_an_entry_without_an_input_rate_gets_no_cache_read_fallback():
litellm.model_cost = { # test-quality-ok: the save/restore conftest returns litellm.model_cost to the original object after each test, so replacing the map for this entry leaks nothing
**litellm.model_cost, # pyright: ignore[reportUnknownMemberType] # the SDK types model_cost as dict[Unknown, Unknown]
"fireworks_ai/accounts/fireworks/models/no-input-rate-test": {
"litellm_provider": "fireworks_ai",
"mode": "chat",
"output_cost_per_token": 2e-06,
},
}
usage: Final = _usage(prompt_tokens=1000, cached_tokens=300, completion_tokens=200)
prompt_cost, completion_cost = cost_per_token(model="accounts/fireworks/models/no-input-rate-test", usage=usage)
assert prompt_cost == 0
assert completion_cost == 200 * 2e-06