@keystrokehq/keystroke 0.1.39 → 0.1.40

package/dist/agent.cjs

@@ -10,7 +10,7 @@ let node_path = require("node:path");
 let node_os = require("node:os");
 let node_fs_promises = require("node:fs/promises");
 let node_sqlite = require("node:sqlite");
-//#region ../agent/dist/schemas-C8sAmmrl.mjs
+//#region ../agent/dist/schemas-4m3ayl5o.mjs
 function capabilitiesFromGatewayTags(tags) {
 	const set = new Set(tags ?? []);
 	return {
@@ -68,7 +68,8 @@ const AgentModelIdSchema = zod.z.custom((value) => typeof value === "string" &&
 const GatewayModelPricingSchema = zod.z.object({
 	input: zod.z.string().optional(),
 	output: zod.z.string().optional(),
-	input_cache_read: zod.z.string().optional()
+	input_cache_read: zod.z.string().optional(),
+	input_cache_write: zod.z.string().optional()
 });
 const GatewayModelResponseSchema = zod.z.object({
 	id: zod.z.string().min(1),
@@ -156,7 +157,8 @@ async function fetchGatewayModelMetadata(modelId) {
 			const pricing = {
 				inputPerMillion: pricePerMillion(raw.pricing?.input),
 				outputPerMillion: pricePerMillion(raw.pricing?.output),
-				cacheReadPerMillion: pricePerMillion(raw.pricing?.input_cache_read)
+				cacheReadPerMillion: pricePerMillion(raw.pricing?.input_cache_read),
+				cacheWritePerMillion: pricePerMillion(raw.pricing?.input_cache_write)
 			};
 			gatewayModelCache.set(cacheKey, {
 				status: "found",
@@ -1138,12 +1140,14 @@ function llmUsageFromLanguageModelUsage(usage, resolved, options) {
 	const reasoning = usage.outputTokenDetails.reasoningTokens ?? 0;
 	const cacheRead = usage.inputTokenDetails.cacheReadTokens ?? 0;
 	const cacheWrite = usage.inputTokenDetails.cacheWriteTokens ?? 0;
+	const noCache = usage.inputTokenDetails.noCacheTokens ?? Math.max(0, input - cacheRead - cacheWrite);
 	const totalTokens = usage.totalTokens ?? input + output;
 	const pricing = resolved.pricing;
-	const inputCost = pricing ? input / 1e6 * pricing.inputPerMillion : 0;
+	const inputCost = pricing ? noCache / 1e6 * pricing.inputPerMillion : 0;
 	const outputCost = pricing ? output / 1e6 * pricing.outputPerMillion : 0;
 	const cacheReadCost = pricing ? cacheRead / 1e6 * pricing.cacheReadPerMillion : 0;
-	const total = inputCost + outputCost + cacheReadCost;
+	const cacheWriteCost = pricing ? cacheWrite / 1e6 * pricing.cacheWritePerMillion : 0;
+	const total = inputCost + outputCost + cacheReadCost + cacheWriteCost;
 	const { directId } = splitAgentModelId(resolved.modelId);
 	return {
 		label: resolved.modelId,
@@ -1157,15 +1161,16 @@ function llmUsageFromLanguageModelUsage(usage, resolved, options) {
 			...options?.byok !== void 0 ? { keySource: options.byok ? "byok" : "platform" } : {},
 			responseModel: resolved.modelId,
 			input,
+			noCache,
 			output,
 			reasoning,
 			cacheRead,
 			cacheWrite,
 			cost: {
-				input: inputCost,
+				noCache: inputCost,
 				output: outputCost,
 				cacheRead: cacheReadCost,
-				cacheWrite: 0,
+				cacheWrite: cacheWriteCost,
 				total
 			}
 		}