mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-18 09:45:23 +00:00
test(cost): cover custom pricing guards and the Fireworks cache default coercion
The Fireworks default cache read rate now trims padded numeric strings the way Python's float() does Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
parent
822f5a6d61
commit
5bf6748040
3 changed files with 75 additions and 1 deletions
|
|
@ -5,7 +5,7 @@ use crate::provider::LlmProviders;
|
|||
fn rate(value: Option<&Value>) -> Option<f64> {
|
||||
match value? {
|
||||
Value::Number(value) => value.as_f64(),
|
||||
Value::String(value) => value.parse().ok(),
|
||||
Value::String(value) => value.trim().parse().ok(),
|
||||
_ => None,
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -265,3 +265,46 @@ fn catalog_custom_pricing_falls_back_and_rejects_invalid_rates() {
|
|||
Err(CostError::InvalidRate)
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::negative_prompt_tokens(usage(-1.0, 0.0), None)]
|
||||
#[case::nan_completion_tokens(usage(0.0, f64::NAN), None)]
|
||||
#[case::negative_detail_cache_tokens(RawUsage { details_cached_tokens: Some(-1.0), ..usage(10.0, 0.0) }, None)]
|
||||
#[case::infinite_fallback_tokens(RawUsage { fallback_cache_read_tokens: Some(f64::INFINITY), ..usage(10.0, 0.0) }, None)]
|
||||
#[case::overflowing_prompt_sum(RawUsage { top_level_cache_read_tokens: Some(f64::MAX), top_level_cache_creation_tokens: Some(f64::MAX), ..usage(f64::MAX, 0.0) }, None)]
|
||||
#[case::valid_usage(usage(10.0, 2.0), Some((10.0, 2.0)))]
|
||||
fn normalize_cache_usage_rejects_negative_or_non_finite_counts(
|
||||
#[case] raw: RawUsage,
|
||||
#[case] expected: Option<(f64, f64)>,
|
||||
) {
|
||||
let normalized = normalize_cache_usage(raw);
|
||||
match expected {
|
||||
Some((prompt, completion)) => {
|
||||
let normalized = normalized.unwrap();
|
||||
assert_eq!(
|
||||
(normalized.prompt_tokens, normalized.completion_tokens),
|
||||
(prompt, completion)
|
||||
);
|
||||
}
|
||||
None => assert_eq!(normalized, Err(CostError::InvalidUsage)),
|
||||
}
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::negative_token_rate(CustomPricing { token: Some(rates(0.1, -0.1, None, None)), per_second: None }, None, CostError::InvalidRate)]
|
||||
#[case::nan_cache_read_rate(CustomPricing { token: Some(rates(0.1, 0.1, Some(f64::NAN), None)), per_second: None }, None, CostError::InvalidRate)]
|
||||
#[case::negative_second_rate(CustomPricing { token: None, per_second: Some(-0.5) }, Some(10.0), CostError::InvalidRate)]
|
||||
#[case::negative_duration(CustomPricing { token: None, per_second: Some(0.5) }, Some(-10.0), CostError::InvalidDuration)]
|
||||
#[case::infinite_duration(CustomPricing { token: None, per_second: Some(0.5) }, Some(f64::INFINITY), CostError::InvalidDuration)]
|
||||
#[case::overflowing_token_cost(CustomPricing { token: Some(rates(f64::MAX, 0.0, None, None)), per_second: None }, None, CostError::NonFiniteCost)]
|
||||
fn cost_per_token_custom_pricing_helper_rejects_unusable_rates_and_durations(
|
||||
#[case] pricing: CustomPricing,
|
||||
#[case] response_time_ms: Option<f64>,
|
||||
#[case] expected: CostError,
|
||||
) {
|
||||
let normalized = normalize_cache_usage(usage(1000.0, 0.0)).unwrap();
|
||||
assert_eq!(
|
||||
cost_per_token_custom_pricing_helper(normalized, pricing, response_time_ms),
|
||||
Err(expected)
|
||||
);
|
||||
}
|
||||
|
|
|
|||
|
|
@ -68,3 +68,34 @@ fn model_info_catalog_uses_fireworks_default_for_cached_tokens() {
|
|||
.unwrap();
|
||||
assert!((actual.0 - (60.0 * 2e-6 + 40.0 * 1e-6)).abs() < 1e-12);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::numeric_string(json!(" 2e-6 "), Some(json!(1e-6)))]
|
||||
#[case::integer(json!(2), Some(json!(1.0)))]
|
||||
#[case::boolean_is_not_a_rate(json!(true), None)]
|
||||
#[case::unparseable_string(json!("free"), None)]
|
||||
#[case::null(json!(null), None)]
|
||||
fn with_default_cache_read_rate_coerces_input_rate_like_python_as_rate(
|
||||
#[case] input_rate: serde_json::Value,
|
||||
#[case] expected: Option<serde_json::Value>,
|
||||
) {
|
||||
let info = json!({"input_cost_per_token": input_rate});
|
||||
assert_eq!(
|
||||
with_default_cache_read_rate(&info)
|
||||
.get("cache_read_input_token_cost")
|
||||
.cloned(),
|
||||
expected
|
||||
);
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
fn with_default_cache_read_rate_falls_back_to_base_rate_for_off_peak_without_input_rate() {
|
||||
let info = json!({
|
||||
"input_cost_per_token": 2e-6,
|
||||
"off_peak_pricing": {"hours_utc": "16:00-20:00"}
|
||||
});
|
||||
assert_eq!(
|
||||
with_default_cache_read_rate(&info)["off_peak_pricing"]["cache_read_input_token_cost"],
|
||||
json!(1e-6)
|
||||
);
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue