@standardagents/builder 0.24.4 → 0.24.6

package/dist/built-in-routes.js

@@ -224,6 +224,18 @@ function getFallbackPricing(providerName, modelId) {
   const pricing = getProviderModelPricing(providerName, modelId);
   return pricing ? toResolvedModelPricing(pricing) : null;
 }
+function resolveUsageReportedPricing(usage) {
+  const pricing = usage.pricing;
+  if (!pricing || typeof pricing.inputPerMillion !== "number" || typeof pricing.outputPerMillion !== "number") {
+    return null;
+  }
+  return {
+    inputPrice: pricing.inputPerMillion,
+    outputPrice: pricing.outputPerMillion,
+    cachedPrice: pricing.cachedInputPerMillion,
+    source: "provider"
+  };
+}
 function resolveModelPricing(modelDef, providerName) {
   if (typeof modelDef.inputPrice === "number" && typeof modelDef.outputPrice === "number") {
     return {
@@ -252,14 +264,14 @@ function calculateUsageCost(usage, pricing) {
   const inputPrice = useLongContext ? pricing.longInputPrice ?? pricing.inputPrice : pricing.inputPrice;
   const outputPrice = useLongContext ? pricing.longOutputPrice ?? pricing.outputPrice : pricing.outputPrice;
   const cachedTokenPrice = useLongContext ? pricing.longCachedPrice ?? pricing.cachedPrice ?? pricing.inputPrice : pricing.cachedPrice ?? pricing.inputPrice;
-  const costInput = roundCost(
-    (uncachedPromptTokens * inputPrice + cachedTokens * cachedTokenPrice) / TOKENS_PER_MILLION
-  );
+  const costInput = roundCost(uncachedPromptTokens * inputPrice / TOKENS_PER_MILLION);
+  const costCached = roundCost(cachedTokens * cachedTokenPrice / TOKENS_PER_MILLION);
   const costOutput = roundCost(billableOutputTokens * outputPrice / TOKENS_PER_MILLION);
   return {
     costInput,
+    costCached,
     costOutput,
-    costTotal: roundCost(costInput + costOutput)
+    costTotal: roundCost(costInput + costCached + costOutput)
   };
 }
 var TOKENS_PER_MILLION;
@@ -1374,7 +1386,8 @@ function convertUsage(usage) {
     prompt_tokens_details: usage.cachedTokens ? { cached_tokens: usage.cachedTokens } : void 0,
     completion_tokens_details: usage.reasoningTokens ? { reasoning_tokens: usage.reasoningTokens } : void 0,
     cost: usage.cost,
-    provider: usage.provider
+    provider: usage.provider,
+    pricing: usage.pricing
   };
 }
 function inferProviderFromModelId(model) {
@@ -2069,7 +2082,7 @@ var init_LLMRequest = __esm({
           const toolsCalled = response.tool_calls ? JSON.stringify(response.tool_calls.map((tc) => tc.function.name)) : null;
           const providerName = response._provider?.name;
           const standardAgentsRouterUsed = response._viaPlatform === true || providerName === "platform";
-          const resolvedPricing = resolveModelPricing(modelDef, providerName);
+          const resolvedPricing = resolveModelPricing(modelDef) ?? resolveUsageReportedPricing(response.usage) ?? resolveModelPricing(modelDef, providerName);
           const calculatedCost = calculateUsageCost(response.usage, resolvedPricing);
           const providerReportedCost = typeof response.usage.cost === "number" ? response.usage.cost : typeof response.usage.cost === "string" ? parseFloat(response.usage.cost) : null;
           const cost_total = providerReportedCost != null && !Number.isNaN(providerReportedCost) ? providerReportedCost : calculatedCost?.costTotal ?? null;
@@ -2136,6 +2149,7 @@ var init_LLMRequest = __esm({
             reasoning_tokens: response.usage.completion_tokens_details?.reasoning_tokens || 0,
             total_tokens: response.usage.total_tokens,
             cost_input: calculatedCost?.costInput,
+            cost_cached: calculatedCost?.costCached,
             cost_output: calculatedCost?.costOutput,
             latency_ms: Date.now() - startTime,
             time_to_first_token_ms: timing?.latencyMs ?? void 0,
@@ -2165,19 +2179,20 @@ var init_LLMRequest = __esm({
               reasoning_tokens = ?5,
               total_tokens = ?6,
               cost_input = ?7,
-              cost_output = ?8,
-              latency_ms = ?9,
-              finish_reason = ?10,
-              tools_called = ?11,
-              cost_total = ?12,
-              is_complete = ?13,
-              reasoning_content = ?14,
-              provider_tools = ?15,
-              actual_provider = ?16,
-              standard_agents_router_used = ?17,
-              time_to_first_token_ms = ?18,
-              generation_ms = ?19
-            WHERE id = ?20
+              cost_cached = ?8,
+              cost_output = ?9,
+              latency_ms = ?10,
+              finish_reason = ?11,
+              tools_called = ?12,
+              cost_total = ?13,
+              is_complete = ?14,
+              reasoning_content = ?15,
+              provider_tools = ?16,
+              actual_provider = ?17,
+              standard_agents_router_used = ?18,
+              time_to_first_token_ms = ?19,
+              generation_ms = ?20
+            WHERE id = ?21
           `,
                 logData.response_body,
                 logData.input_tokens,
@@ -2186,6 +2201,7 @@ var init_LLMRequest = __esm({
                 logData.reasoning_tokens,
                 logData.total_tokens,
                 logData.cost_input ?? null,
+                logData.cost_cached ?? null,
                 logData.cost_output ?? null,
                 logData.latency_ms,
                 logData.finish_reason,
@@ -56244,6 +56260,25 @@ function getModelIconDataUri(modelId) {
   }
   return void 0;
 }
+function usagePricingFromCatalogPricing(pricing) {
+  if (!pricing) return void 0;
+  const perMillion = (value) => {
+    if (value === void 0) return void 0;
+    const parsed = Number.parseFloat(value);
+    if (!Number.isFinite(parsed)) return void 0;
+    return Number((parsed * 1e6).toFixed(12));
+  };
+  const inputPerMillion = perMillion(pricing.prompt);
+  const outputPerMillion = perMillion(pricing.completion);
+  if (inputPerMillion === void 0 || outputPerMillion === void 0) {
+    return void 0;
+  }
+  return {
+    inputPerMillion,
+    cachedInputPerMillion: perMillion(pricing.input_cache_read),
+    outputPerMillion
+  };
+}
 var OPENROUTER_SERVER_TOOL_NAMES, google_default3, meta_default4, mistralai_default3, moonshotai_default3, nvidia_default2, openai_default5, qwen_default4, zai_org_default4, ANTHROPIC_ICON2, GOOGLE_ICON3, META_ICON3, NVIDIA_ICON, ALIBABA_ICON2, DEEPSEEK_ICON2, QWEN_ICON2, MINIMAX_ICON, MOONSHOTAI_ICON, XAI_ICON, ZAI_ICON3, EXAONE_ICON, OPENROUTER_ICON, OPENAI_ICON4, MISTRAL_ICON2, COHERE_ICON, PERPLEXITY_ICON, LAB_ICONS, OPENROUTER_PROVIDER_ICON_PREFIX, OPENROUTER_PROVIDER_URL_ENTRIES, OPENROUTER_PROVIDER_URLS, OpenRouterProvider, percentileSchema, providerRoutingSchema, serverToolParametersSchema, serverToolsSchema, openrouterProviderOptions, openrouter;
 var init_dist6 = __esm({
   "../openrouter/dist/index.js"() {
@@ -56545,6 +56580,11 @@ var init_dist6 = __esm({
       static modelsCacheTime = 0;
       static CACHE_TTL = 5 * 60 * 1e3;
       // 5 minutes
+      /** In-flight catalog download shared by concurrent callers. */
+      static modelsFetchPromise = null;
+      /** Timestamp of the last failed catalog fetch, for negative caching. */
+      static modelsFetchFailedAt = 0;
+      static FAILURE_RETRY_MS = 60 * 1e3;
       /**
        * Get capabilities for a specific model.
        * Fetches from OpenRouter's models API which provides rich metadata.
@@ -56562,19 +56602,73 @@ var init_dist6 = __esm({
           return null;
         }
       }
+      /**
+       * How long a streaming finish chunk may wait on the catalog pricing lookup.
+       * The lookup is started when the stream starts, so by finish time it is
+       * almost always resolved; the bound guarantees a slow catalog download can
+       * never stall the finish chunk into the runtime's stream watchdog.
+       */
+      static USAGE_PRICING_FINISH_WAIT_MS = 2500;
+      static async settleUsagePricing(pricingPromise) {
+        let timer;
+        const timeout = new Promise((resolve2) => {
+          timer = setTimeout(() => resolve2(void 0), _OpenRouterProvider.USAGE_PRICING_FINISH_WAIT_MS);
+        });
+        try {
+          return await Promise.race([pricingPromise, timeout]);
+        } finally {
+          if (timer !== void 0) clearTimeout(timer);
+        }
+      }
+      /** Bounded-wait pricing attachment shared by the non-streaming paths. */
+      async attachUsagePricing(result, pricingPromise) {
+        const pricing = await _OpenRouterProvider.settleUsagePricing(pricingPromise);
+        if (pricing) result.usage.pricing = pricing;
+        return result;
+      }
+      /**
+       * Exact unit prices for a model from OpenRouter's live catalog (cached),
+       * attached to response usage so request logs can record specific prices.
+       * Best-effort: any failure just leaves pricing off the usage payload.
+       */
+      async getUsagePricing(modelId) {
+        if (!modelId) return void 0;
+        try {
+          const models = await this.fetchModelsWithCache();
+          const model = models.find((m) => m.id === modelId);
+          return usagePricingFromCatalogPricing(model?.pricing);
+        } catch {
+          return void 0;
+        }
+      }
       async fetchModelsWithCache() {
         const now = Date.now();
         if (_OpenRouterProvider.modelsCache && now - _OpenRouterProvider.modelsCacheTime < _OpenRouterProvider.CACHE_TTL) {
           return _OpenRouterProvider.modelsCache;
         }
-        const response = await fetch("https://openrouter.ai/api/v1/models");
-        if (!response.ok) {
-          throw new Error(`OpenRouter API error: ${response.status}`);
+        if (now - _OpenRouterProvider.modelsFetchFailedAt < _OpenRouterProvider.FAILURE_RETRY_MS) {
+          throw new Error("OpenRouter models fetch recently failed");
         }
-        const data = await response.json();
-        _OpenRouterProvider.modelsCache = data.data || [];
-        _OpenRouterProvider.modelsCacheTime = now;
-        return _OpenRouterProvider.modelsCache;
+        if (!_OpenRouterProvider.modelsFetchPromise) {
+          _OpenRouterProvider.modelsFetchPromise = (async () => {
+            try {
+              const response = await fetch("https://openrouter.ai/api/v1/models");
+              if (!response.ok) {
+                throw new Error(`OpenRouter API error: ${response.status}`);
+              }
+              const data = await response.json();
+              _OpenRouterProvider.modelsCache = data.data || [];
+              _OpenRouterProvider.modelsCacheTime = Date.now();
+              return _OpenRouterProvider.modelsCache;
+            } catch (error) {
+              _OpenRouterProvider.modelsFetchFailedAt = Date.now();
+              throw error;
+            } finally {
+              _OpenRouterProvider.modelsFetchPromise = null;
+            }
+          })();
+        }
+        return _OpenRouterProvider.modelsFetchPromise;
       }
       mapOpenRouterCapabilities(model) {
         const capabilities = {
@@ -56659,6 +56753,7 @@ var init_dist6 = __esm({
           if (this.config.providers && this.config.providers.length > 0) {
             params.provider = { only: this.config.providers };
           }
+          const pricingPromise = this.getUsagePricing(params.model);
           const response = await fetch(`${baseUrl}/chat/completions`, {
             method: "POST",
             headers: {
@@ -56680,7 +56775,7 @@ var init_dist6 = __esm({
             throw new ProviderError(errorMessage, "invalid_request", response.status);
           }
           const data = await response.json();
-          return transformChatResponse3(data);
+          return this.attachUsagePricing(transformChatResponse3(data), pricingPromise);
         } catch (error) {
           throw this.toProviderError(error);
         }
@@ -56694,6 +56789,7 @@ var init_dist6 = __esm({
         try {
           const params = buildCreateParams3(request);
           this.applyConfiguredProvidersToResponses(params);
+          const pricingPromise = this.getUsagePricing(params.model);
           if (this.hasOpenRouterServerTools(params)) {
             const response2 = await fetch(`${baseUrl}/responses`, {
               method: "POST",
@@ -56711,14 +56807,14 @@ var init_dist6 = __esm({
               await this.throwOpenRouterFetchError(response2);
             }
             const data = await response2.json();
-            return transformResponse3(data);
+            return this.attachUsagePricing(transformResponse3(data), pricingPromise);
           }
           const client = await this.getClient();
           const response = await client.beta.responses.send({
             ...this.toSdkResponsesRequest(params),
             stream: false
           });
-          return transformResponse3(response);
+          return this.attachUsagePricing(transformResponse3(response), pricingPromise);
         } catch (error) {
           throw this.toProviderError(error);
         }
@@ -56742,6 +56838,7 @@ var init_dist6 = __esm({
           if (this.config.providers && this.config.providers.length > 0) {
             params.provider = { only: this.config.providers };
           }
+          const pricingPromise = this.getUsagePricing(params.model);
           const response = await fetch(`${baseUrl}/chat/completions`, {
             method: "POST",
             headers: {
@@ -56781,6 +56878,10 @@ var init_dist6 = __esm({
                   }
                 }
                 if (finishChunk) {
+                  if (finishChunk.type === "finish") {
+                    const pricing = await _OpenRouterProvider.settleUsagePricing(pricingPromise);
+                    if (pricing) finishChunk.usage.pricing = pricing;
+                  }
                   const finishWithMeta = {
                     ...finishChunk,
                     _asyncMetadata: responseId ? {
@@ -56811,6 +56912,7 @@ var init_dist6 = __esm({
         try {
           const params = buildCreateParams3(request);
           this.applyConfiguredProvidersToResponses(params);
+          const pricingPromise = this.getUsagePricing(params.model);
           const response = await fetch(`${baseUrl}/responses`, {
             method: "POST",
             headers: {
@@ -56849,6 +56951,10 @@ var init_dist6 = __esm({
                   }
                 }
                 if (finishChunk) {
+                  if (finishChunk.type === "finish") {
+                    const pricing = await _OpenRouterProvider.settleUsagePricing(pricingPromise);
+                    if (pricing) finishChunk.usage.pricing = pricing;
+                  }
                   const finishWithMeta = {
                     ...finishChunk,
                     // Include what's needed to fetch metadata later
@@ -69109,14 +69215,31 @@ var continue_post_default = defineController2(async ({ req, params, env: env2 })
 
 // src/agents/cost-summary.ts
 function emptyCostSummary() {
-  return { totalCost: 0, requestCount: 0, messageCount: 0, byModel: {} };
+  return {
+    totalCost: 0,
+    inputCost: 0,
+    cachedCost: 0,
+    outputCost: 0,
+    requestCount: 0,
+    messageCount: 0,
+    byModel: {}
+  };
 }
 function mergeByModel(summaries) {
   const merged = {};
   for (const summary of summaries) {
     for (const [modelName, acc] of Object.entries(summary.byModel)) {
-      const entry = merged[modelName] ?? { totalCost: 0, count: 0 };
+      const entry = merged[modelName] ?? {
+        totalCost: 0,
+        inputCost: 0,
+        cachedCost: 0,
+        outputCost: 0,
+        count: 0
+      };
       entry.totalCost += acc.totalCost;
+      entry.inputCost = (entry.inputCost ?? 0) + (acc.inputCost ?? 0);
+      entry.cachedCost = (entry.cachedCost ?? 0) + (acc.cachedCost ?? 0);
+      entry.outputCost = (entry.outputCost ?? 0) + (acc.outputCost ?? 0);
       entry.count += acc.count;
       merged[modelName] = entry;
     }
@@ -69125,19 +69248,28 @@ function mergeByModel(summaries) {
 }
 function sumSummaries(summaries) {
   let totalCost = 0;
+  let inputCost = 0;
+  let cachedCost = 0;
+  let outputCost = 0;
   let requestCount = 0;
   let messageCount = 0;
   for (const summary of summaries) {
     totalCost += summary.totalCost;
+    inputCost += summary.inputCost ?? 0;
+    cachedCost += summary.cachedCost ?? 0;
+    outputCost += summary.outputCost ?? 0;
     requestCount += summary.requestCount;
     messageCount += summary.messageCount;
   }
-  return { totalCost, requestCount, messageCount };
+  return { totalCost, inputCost, cachedCost, outputCost, requestCount, messageCount };
 }
 function buildModelBreakdown(byModel) {
   return Object.entries(byModel).map(([modelName, acc]) => ({
     modelName,
     totalCost: acc.totalCost,
+    inputCost: acc.inputCost ?? 0,
+    cachedCost: acc.cachedCost ?? 0,
+    outputCost: acc.outputCost ?? 0,
     requestCount: acc.count,
     avgCostPerRequest: acc.count > 0 ? acc.totalCost / acc.count : 0
   })).sort((a, b) => b.totalCost - a.totalCost);
@@ -69212,6 +69344,9 @@ var cost_get_default = defineController2(async ({ req, params, env: env2 }) => {
         parent: null,
         isRoot: true,
         totalCost: ownSummary.totalCost,
+        inputCost: ownSummary.inputCost ?? 0,
+        cachedCost: ownSummary.cachedCost ?? 0,
+        outputCost: ownSummary.outputCost ?? 0,
         requestCount: ownSummary.requestCount
       },
       ...descendants.map((child, index) => ({
@@ -69221,11 +69356,17 @@ var cost_get_default = defineController2(async ({ req, params, env: env2 }) => {
         parent: child.parent,
         isRoot: false,
         totalCost: descendantSummaries[index]?.totalCost ?? 0,
+        inputCost: descendantSummaries[index]?.inputCost ?? 0,
+        cachedCost: descendantSummaries[index]?.cachedCost ?? 0,
+        outputCost: descendantSummaries[index]?.outputCost ?? 0,
         requestCount: descendantSummaries[index]?.requestCount ?? 0
       })).filter((row) => row.requestCount > 0)
     ];
     return Response.json({
       totalCost: totals.totalCost,
+      inputCost: totals.inputCost,
+      cachedCost: totals.cachedCost,
+      outputCost: totals.outputCost,
       ownCost: ownSummary.totalCost,
       requestCount: totals.requestCount,
       messageCount: totals.messageCount,