/**
* LLM Cost Calculator Utility
* Location: src/services/llm/utils/LLMCostCalculator.ts
*
* Extracted from BaseAdapter.ts to follow Single Responsibility Principle.
* Handles all cost calculation logic including caching discounts for different providers.
*
* Usage:
* - Used by BaseAdapter and all provider adapters
* - Calculates input/output costs based on token usage and model pricing
* - Applies provider-specific caching discounts (OpenAI, Anthropic, Google)
*/
import { TokenUsage, CostDetails, ModelPricing } from '../adapters/types';
export class LLMCostCalculator {
/**
* Calculate cost based on token usage and model pricing
* Supports caching discounts for providers that offer them
*/
static async calculateCost(
usage: TokenUsage,
model: string,
modelPricing: ModelPricing | null
): Promise<CostDetails | null> {
if (!modelPricing) {
return null;
}
// Determine caching discount rate based on provider and model
const cachingDiscount = this.getCachingDiscount(model);
// Calculate input cost with caching discount
let inputCost = 0;
let cachedCost = 0;
if (usage.cachedTokens && usage.cachedTokens > 0 && cachingDiscount < 1.0) {
// Split input tokens into cached and fresh
const freshTokens = usage.promptTokens - usage.cachedTokens;
const freshCost = (freshTokens / 1_000_000) * modelPricing.rateInputPerMillion;
cachedCost = (usage.cachedTokens / 1_000_000) * modelPricing.rateInputPerMillion * cachingDiscount;
inputCost = freshCost + cachedCost;
} else {
// No cached tokens, use standard pricing
inputCost = (usage.promptTokens / 1_000_000) * modelPricing.rateInputPerMillion;
}
const outputCost = (usage.completionTokens / 1_000_000) * modelPricing.rateOutputPerMillion;
const totalCost = inputCost + outputCost;
const costDetails: CostDetails = {
inputCost,
outputCost,
totalCost,
currency: modelPricing.currency || 'USD',
rateInputPerMillion: modelPricing.rateInputPerMillion,
rateOutputPerMillion: modelPricing.rateOutputPerMillion
};
// Add cached token details if applicable
if (usage.cachedTokens && usage.cachedTokens > 0) {
costDetails.cached = {
tokens: usage.cachedTokens,
cost: cachedCost
};
}
return costDetails;
}
/**
* Get caching discount multiplier for a model
* Returns the fraction of the original price (e.g., 0.1 = 90% off, 0.25 = 75% off)
*/
static getCachingDiscount(model: string): number {
// OpenAI pricing as of Oct 2025:
// GPT-5 family: 90% off cached tokens (pay 10%)
if (model.startsWith('gpt-5')) {
return 0.1;
}
// GPT-5.2 family: 75% off cached tokens (pay 25%)
if (model.startsWith('gpt-5.2')) {
return 0.25;
}
// Anthropic Claude: 90% off cached tokens
if (model.startsWith('claude')) {
return 0.1;
}
// Google Gemini: 50% off cached tokens
if (model.startsWith('gemini')) {
return 0.5;
}
// Default: no caching discount
return 1.0;
}
}
