@standardagents/builder 0.24.3 → 0.24.5

package/dist/built-in-routes.js

@@ -252,14 +252,14 @@ 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;
@@ -1323,10 +1323,14 @@ function buildRequestBody(context, modelDef) {
     visionFiltering
   };
 }
-function buildProviderRequest(context, modelDef, signal) {
+function buildProviderRequest(context, modelDef, signal, user) {
   const { requestBody } = buildRequestBody(context, modelDef);
   return {
     ...requestBody,
+    // Stable per-thread identity: providers that support it (OpenRouter's
+    // provider-sticky routing) pin the thread to one upstream, which is what
+    // makes prompt caching hit across consecutive steps.
+    user,
     signal
   };
 }
@@ -1673,7 +1677,7 @@ var init_LLMRequest = __esm({
             parentSignal.addEventListener("abort", onParentAbort, { once: true });
           }
         }
-        const request = buildProviderRequest(context, modelDef, streamAbort.signal);
+        const request = buildProviderRequest(context, modelDef, streamAbort.signal, state.threadId);
         let response;
         await state.stream.waitFor(
           async () => {
@@ -2132,6 +2136,7 @@ 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,
@@ -2161,19 +2166,20 @@ 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,
@@ -2182,6 +2188,7 @@ 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,
@@ -55224,6 +55231,9 @@ function buildChatParams3(request) {
       }
     }
   }
+  if (request.user) {
+    params.user = request.user;
+  }
   const safeOptions = splitOpenRouterProviderOptions(
     request.providerOptions
   );
@@ -55745,6 +55755,9 @@ function buildCreateParams3(request) {
       }
     }
   }
+  if (request.user) {
+    params.user = request.user;
+  }
   const providerOptions = splitOpenRouterProviderOptions(
     request.providerOptions
   );