aicost 0.1.8

Compute LLM usage costs based on the genai::Usage type.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214

use crate::data::PROVIDERS;
use crate::{AiCost, Error, ModelPricing, Result};
use genai::ModelIden;
use genai::chat::Usage;

/// Calculates the cost for a given provider type, model name, and usage.
pub fn compute(provider_type: &str, model_name: &str, usage: &Usage) -> Result<AiCost> {
	// Since not api from genai yet, extract the eventual namespace of the modelname
	let model_name = normalize_model_name(model_name);

	let provider_type = normalize_provider_type(provider_type);

	// Find the model within the provider (longest start_with)
	let model = find_model_entry(provider_type, model_name)
		.ok_or_else(|| Error::ModelNotFound(provider_type.to_string(), model_name.to_string()))?;

	// -- Extract prompt related tokens and pricing
	let prompt_tokens = usage.prompt_tokens.unwrap_or(0) as f64;
	// Extract cached vs normal prompt tokens
	let (prompt_tokens_normal, prompt_cached_tokens, prompt_cache_creation_tokens) = match &usage.prompt_tokens_details
	{
		Some(details) => {
			let cached = details.cached_tokens.unwrap_or(0) as f64;
			let cache_creation_tokens = details.cache_creation_tokens.unwrap_or(0) as f64;
			let normal = prompt_tokens - cached - cache_creation_tokens;
			(normal, cached, cache_creation_tokens)
		}
		None => (prompt_tokens, 0.0, 0.0),
	};
	// Extract prompt prices
	let price_prompt_normal = model.input_normal;
	let price_prompt_cached = model.input_cached.unwrap_or(price_prompt_normal);
	// Note:  For now, for cache_creation_tokens assume * 1.25 for cache_creation_tokens (which is Anthropic rules, and this is only anthropic data)
	let price_prompt_cache_creation = 1.25 * price_prompt_normal;

	// -- Extract completion related tokens and pricing
	let completion_tokens = usage.completion_tokens.unwrap_or(0) as f64;
	let (completion_tokens_normal, completion_tokens_reasoning) = if let Some(reasoning_tokens) = usage
		.completion_tokens_details
		.as_ref()
		.and_then(|v| v.reasoning_tokens.map(|v| v as f64))
	{
		(completion_tokens - reasoning_tokens, reasoning_tokens)
	} else {
		(completion_tokens, 0.)
	};
	let price_completion_normal = model.output_normal;
	let price_completion_reasoning = model.output_reasoning.unwrap_or(price_completion_normal);

	// -- Compute individual cost components (raw)
	let cost_prompt_normal = prompt_tokens_normal * price_prompt_normal / 1_000_000.0;
	let cost_prompt_cached = prompt_cached_tokens * price_prompt_cached / 1_000_000.0;
	let cost_prompt_cache_creation = prompt_cache_creation_tokens * price_prompt_cache_creation / 1_000_000.0;
	let cost_completion_normal = completion_tokens_normal * price_completion_normal / 1_000_000.0;
	let cost_completion_reasoning = completion_tokens_reasoning * price_completion_reasoning / 1_000_000.0;

	// -- Compute total components (rounded)
	let input_normal = (cost_prompt_normal * 1_000_000.0).round() / 1_000_000.0;
	let input_cache_read = (cost_prompt_cached * 1_000_000.0).round() / 1_000_000.0;
	let input_cache_write = (cost_prompt_cache_creation * 1_000_000.0).round() / 1_000_000.0;
	let input_total = ((input_normal + input_cache_read + input_cache_write) * 1_000_000.0).round() / 1_000_000.0;

	let output_normal = (cost_completion_normal * 1_000_000.0).round() / 1_000_000.0;
	let output_reasoning = (cost_completion_reasoning * 1_000_000.0).round() / 1_000_000.0;
	let output_total = ((output_normal + output_reasoning) * 1_000_000.0).round() / 1_000_000.0;

	let total = ((input_total + output_total) * 1_000_000.0).round() / 1_000_000.0;

	// -- Compute cache saving
	let input_cache_saving = if prompt_cached_tokens > 0.0 {
		let would_have_cost = prompt_cached_tokens * price_prompt_normal / 1_000_000.0;
		let saving = (would_have_cost - cost_prompt_cached).max(0.0);
		(saving * 1_000_000.0).round() / 1_000_000.0
	} else {
		0.0
	};

	Ok(AiCost {
		total,
		input_total,
		input_normal,
		input_cache_read,
		input_cache_write,
		output_total,
		output_normal,
		output_reasoning,
		input_cache_saving,
	})
}

/// Calculates the cost for a given `ModelIden` and usage.
pub fn compute_iden(model_iden: &ModelIden, usage: &Usage) -> Result<AiCost> {
	let provider_type = model_iden.adapter_kind.as_lower_str();
	let model_name = &*model_iden.model_name;

	compute(provider_type, model_name, usage)
}

pub fn model_pricing(model_iden: &ModelIden) -> Option<ModelPricing> {
	let provider_type = model_iden.adapter_kind.as_lower_str();
	let model_name = &*model_iden.model_name;

	let model_name = normalize_model_name(model_name);

	let provider_type = normalize_provider_type(provider_type);

	find_model_entry(provider_type, model_name).copied()
}

// region:    --- Support

fn normalize_model_name(model_name: &str) -> &str {
	match model_name.split_once("::") {
		Some((_, after)) => after,
		None => model_name,
	}
}

fn normalize_provider_type(provider_type: &str) -> &str {
	if provider_type == "openai_resp" {
		"openai"
	} else {
		provider_type
	}
}

fn find_model_entry(provider_type: &str, model_name: &str) -> Option<&'static ModelPricing> {
	let provider = PROVIDERS.iter().find(|p| p.name == provider_type)?;

	let mut model: Option<&ModelPricing> = None;
	for m in provider.models.iter() {
		if model_name.starts_with(m.name) {
			let current_len = model.map(|m| m.name.len()).unwrap_or(0);
			if current_len < m.name.len() {
				model = Some(m)
			}
		}
	}

	model
}

// endregion: --- Support

// region:    --- Tests

#[cfg(test)]
mod tests {
	use super::*;
	use genai::chat::{PromptTokensDetails, Usage};

	type TestResult = std::result::Result<(), Box<dyn std::error::Error>>;

	#[test]
	fn test_pricing_core_cost_simple() -> TestResult {
		// -- Setup & Fixtures
		let usage = Usage {
			prompt_tokens: Some(1000),
			completion_tokens: Some(500),
			prompt_tokens_details: None,
			..Default::default()
		};

		// -- Exec
		let ai_cost = compute("openai", "gpt-4o", &usage)?;

		// -- Check
		let price = ai_cost.total;
		// gpt-4o pricing: input_normal: 2.5, output_normal: 10.0
		// Expected: (1000 * 2.5 / 1_000_000) + (500 * 10.0 / 1_000_000) = 0.0025 + 0.005 = 0.0075
		let expected = 0.0075;
		assert!((price - expected).abs() < f64::EPSILON);
		assert!(ai_cost.input_cache_write == 0.0);
		assert!(ai_cost.input_cache_saving == 0.0);

		Ok(())
	}

	#[test]
	fn test_pricing_core_cost_with_cached() -> TestResult {
		// -- Setup & Fixtures
		let fx_prompt_normal_tokens = 1000;
		let fx_completion_tokens = 500;
		let fx_cached_tokens = 400;
		let usage = Usage {
			prompt_tokens: Some(fx_prompt_normal_tokens + fx_cached_tokens),
			completion_tokens: Some(fx_completion_tokens),
			prompt_tokens_details: Some(PromptTokensDetails {
				cached_tokens: Some(fx_cached_tokens),
				..Default::default()
			}),
			..Default::default()
		};

		// -- Exec
		let ai_cost = compute("openai", "gpt-4o-mini", &usage)?;

		// -- Check
		let price = ai_cost.total;
		// gpt-4o-mini pricing: input_normal: 0.15, input_cached: 0.075, output_normal: 0.6
		let cached = fx_cached_tokens as f64 * 0.075 / 1_000_000.0;
		let prompt = fx_prompt_normal_tokens as f64 * 0.150 / 1_000_000.0;
		let completion = fx_completion_tokens as f64 * 0.6 / 1_000_000.0;
		let expected = cached + prompt + completion;
		let expected = (expected * 1_000_000.0).round() / 1_000_000.0;
		assert!((price - expected).abs() < f64::EPSILON);
		assert!(ai_cost.input_cache_saving > 0.0);
		assert!(ai_cost.input_cache_write == 0.0);

		Ok(())
	}
}

// endregion: --- Tests