naisys/dist/llm/cacheSavings.js

NAISYS - Autonomous AI agent runner with built-in context management and cost tracking

export function calculateModelCacheSavings(modelData, model) {
    const cacheWriteTokens = modelData.cacheWriteTokens || 0;
    const cacheReadTokens = modelData.cacheReadTokens || 0;
    const totalCacheTokens = cacheWriteTokens + cacheReadTokens;
    if (totalCacheTokens === 0 || !model.inputCost) {
        return null;
    }
    // Calculate what these cache tokens would have cost at regular input rate
    const cacheSavingsAmount = (cacheWriteTokens * (model.inputCost - (model.cacheWriteCost || 0))) /
        1_000_000 +
        (cacheReadTokens * (model.inputCost - (model.cacheReadCost || 0))) /
            1_000_000;
    // Calculate actual cache cost from tokens
    const actualCacheSpend = (cacheWriteTokens * (model.cacheWriteCost || 0)) / 1_000_000 +
        (cacheReadTokens * (model.cacheReadCost || 0)) / 1_000_000;
    // Calculate total cost for this model from tokens
    const inputTokens = modelData.inputTokens || 0;
    const outputTokens = modelData.outputTokens || 0;
    const inputCost = (inputTokens * model.inputCost) / 1_000_000;
    const outputCost = (outputTokens * model.outputCost) / 1_000_000;
    const modelTotalCost = inputCost + outputCost + actualCacheSpend;
    const costWithoutCaching = modelTotalCost + cacheSavingsAmount;
    const savingsPercent = cacheSavingsAmount > 0
        ? (cacheSavingsAmount / costWithoutCaching) * 100
        : 0;
    return {
        savingsAmount: cacheSavingsAmount,
        costWithoutCaching,
        savingsPercent,
        totalCacheTokens,
        totalCost: modelTotalCost,
        inputCost,
        outputCost,
        actualCacheSpend,
    };
}
//# sourceMappingURL=cacheSavings.js.map