@ljoukov/llm 7.0.17 → 7.0.18

package/dist/index.d.cts

@@ -15,8 +15,9 @@ type LlmUsageTokens = {
     readonly totalTokens?: number;
     readonly toolUsePromptTokens?: number;
 };
-declare function estimateCallCostUsd({ modelId, tokens, responseImages, imageSize, imageQuality, }: {
+declare function estimateCallCostUsd({ modelId, pricingModelId, tokens, responseImages, imageSize, imageQuality, }: {
     modelId: string;
+    pricingModelId?: string;
     tokens: LlmUsageTokens | undefined;
     responseImages: number;
     imageSize?: string;

package/dist/index.d.ts

@@ -15,8 +15,9 @@ type LlmUsageTokens = {
     readonly totalTokens?: number;
     readonly toolUsePromptTokens?: number;
 };
-declare function estimateCallCostUsd({ modelId, tokens, responseImages, imageSize, imageQuality, }: {
+declare function estimateCallCostUsd({ modelId, pricingModelId, tokens, responseImages, imageSize, imageQuality, }: {
     modelId: string;
+    pricingModelId?: string;
     tokens: LlmUsageTokens | undefined;
     responseImages: number;
     imageSize?: string;

package/dist/index.js

@@ -344,6 +344,7 @@ function resolveChatGptServiceTier(model) {
 // src/openai/pricing.ts
 var OPENAI_GPT_55_FAST_MODEL_IDS = ["gpt-5.5-fast", "chatgpt-gpt-5.5-fast"];
 var OPENAI_GPT_55_STANDARD_MODEL_IDS = ["gpt-5.5", "chatgpt-gpt-5.5"];
+var OPENAI_GPT_55_CONCRETE_MODEL_ID_RE = /^(?:chatgpt-)?gpt-5\.5-\d{4}-\d{2}-\d{2}$/u;
 var OPENAI_GPT_54_FAST_MODEL_IDS = ["gpt-5.4-fast", "chatgpt-gpt-5.4-fast"];
 var OPENAI_GPT_54_MINI_MODEL_IDS = ["gpt-5.4-mini", "chatgpt-gpt-5.4-mini"];
 var OPENAI_GPT_54_NANO_MODEL_IDS = ["gpt-5.4-nano"];
@@ -410,7 +411,7 @@ function getOpenAiPricing(modelId) {
   if (OPENAI_GPT_55_FAST_MODEL_IDS.includes(modelId)) {
     return OPENAI_GPT_55_PRIORITY_PRICING;
   }
-  if (OPENAI_GPT_55_STANDARD_MODEL_IDS.includes(modelId)) {
+  if (OPENAI_GPT_55_STANDARD_MODEL_IDS.includes(modelId) || OPENAI_GPT_55_CONCRETE_MODEL_ID_RE.test(modelId)) {
     return OPENAI_GPT_55_PRICING;
   }
   if (OPENAI_GPT_54_FAST_MODEL_IDS.includes(modelId)) {
@@ -443,12 +444,14 @@ function resolveUsageNumber(value) {
 }
 function estimateCallCostUsd({
   modelId,
+  pricingModelId,
   tokens,
   responseImages,
   imageSize,
   imageQuality
 }) {
-  const openAiImagePricing = getOpenAiImagePricing(modelId);
+  const pricingModelIds = resolvePricingModelIds(modelId, pricingModelId);
+  const openAiImagePricing = resolvePricing(pricingModelIds, getOpenAiImagePricing);
   if (openAiImagePricing) {
     return estimateOpenAiImageCostUsd({
       pricing: openAiImagePricing,
@@ -468,7 +471,7 @@ function estimateCallCostUsd({
   const toolUsePromptTokens = resolveUsageNumber(tokens.toolUsePromptTokens);
   const promptTokenTotal = promptTokens + toolUsePromptTokens;
   const nonCachedPrompt = Math.max(0, promptTokenTotal - cachedTokens);
-  const imagePreviewPricing = getGeminiImagePricing(modelId);
+  const imagePreviewPricing = resolvePricing(pricingModelIds, getGeminiImagePricing);
   if (imagePreviewPricing) {
     const resolvedImageSize = imageSize && imagePreviewPricing.imagePrices[imageSize] ? imageSize : "2K";
     const imageRate = imagePreviewPricing.imagePrices[resolvedImageSize] ?? 0;
@@ -488,7 +491,7 @@ function estimateCallCostUsd({
     const imageOutputCost = imageTokensForPricing * imagePreviewPricing.outputImageRate;
     return inputCost + cachedCost + textOutputCost + imageOutputCost;
   }
-  const geminiPricing = getGeminiProPricing(modelId);
+  const geminiPricing = resolvePricing(pricingModelIds, getGeminiProPricing);
   if (geminiPricing) {
     const useHighTier = promptTokenTotal > geminiPricing.threshold;
     const inputRate = useHighTier ? geminiPricing.inputRateHigh : geminiPricing.inputRateLow;
@@ -500,7 +503,7 @@ function estimateCallCostUsd({
     const outputCost = outputTokens * outputRate;
     return inputCost + cachedCost + outputCost;
   }
-  const fireworksPricing = getFireworksPricing(modelId);
+  const fireworksPricing = resolvePricing(pricingModelIds, getFireworksPricing);
   if (fireworksPricing) {
     const inputCost = nonCachedPrompt * fireworksPricing.inputRate;
     const cachedCost = cachedTokens * fireworksPricing.cachedRate;
@@ -508,7 +511,7 @@ function estimateCallCostUsd({
     const outputCost = outputTokens * fireworksPricing.outputRate;
     return inputCost + cachedCost + outputCost;
   }
-  const openAiPricing = getOpenAiPricing(modelId);
+  const openAiPricing = resolvePricing(pricingModelIds, getOpenAiPricing);
   if (openAiPricing) {
     const inputCost = nonCachedPrompt * openAiPricing.inputRate;
     const cachedCost = cachedTokens * openAiPricing.cachedRate;
@@ -518,6 +521,21 @@ function estimateCallCostUsd({
   }
   return 0;
 }
+function resolvePricingModelIds(modelId, pricingModelId) {
+  if (pricingModelId && pricingModelId !== modelId) {
+    return [pricingModelId, modelId];
+  }
+  return [modelId];
+}
+function resolvePricing(modelIds, resolve) {
+  for (const modelId of modelIds) {
+    const pricing = resolve(modelId);
+    if (pricing) {
+      return pricing;
+    }
+  }
+  return void 0;
+}
 function estimateOpenAiImageCostUsd({
   pricing,
   responseImages,
@@ -8329,6 +8347,7 @@ async function runTextCall(params) {
       const outputAttachments = collectLoggedAttachmentsFromLlmParts(mergedParts, "output");
       const costUsd = estimateCallCostUsd({
         modelId: modelVersion,
+        pricingModelId: request.model,
         tokens: latestUsage,
         responseImages,
         imageSize: request.imageSize
@@ -9072,6 +9091,7 @@ async function runToolLoop(request) {
           const modelCompletedAtMs = Date.now();
           const stepCostUsd = estimateCallCostUsd({
             modelId: modelVersion,
+            pricingModelId: request.model,
             tokens: usageTokens,
             responseImages: 0
           });
@@ -9403,6 +9423,7 @@ async function runToolLoop(request) {
           usageTokens = extractChatGptUsageTokens(response.usage);
           const stepCostUsd = estimateCallCostUsd({
             modelId: modelVersion,
+            pricingModelId: request.model,
             tokens: usageTokens,
             responseImages: 0
           });
@@ -9731,6 +9752,7 @@ async function runToolLoop(request) {
           usageTokens = extractFireworksUsageTokens(response.usage);
           const stepCostUsd = estimateCallCostUsd({
             modelId: modelVersion,
+            pricingModelId: request.model,
             tokens: usageTokens,
             responseImages: 0
           });
@@ -10093,6 +10115,7 @@ async function runToolLoop(request) {
         );
         const stepCostUsd = estimateCallCostUsd({
           modelId: modelVersion,
+          pricingModelId: request.model,
           tokens: usageTokens,
           responseImages: 0
         });