enable prompt caching (#10)

*Issue #, if available:*
n/a

*Description of changes:*

Support prompt cache for Sonnet3.7 for token efficiency.

https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-caching.html

By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice.

Author: tmokmss

README.md

@@ -10,7 +10,7 @@ This is an example implementation of a fully autonomous software development AI
 * Powered by AWS serverless services with minimal maintenance costs
 * No upfront or fixed costs while you don't use the system
 * MCP integration (tool servers)
-* Unlimited context window (middle-out)
+* Efficient token usage with prompt cache and middle-out strategy
 * Reads knowledge from your preferred formats (.clinerules, CLAUDE.md, etc.)
 * Can work on OSS forked repositories!
 

slack-bolt-app/src/app.ts

@@ -103,27 +103,48 @@ app.event('app_mention', async ({ event, client, logger }) => {
 
         const tokenSummary = tokenUsage
           .map((item) => {
-            const cost = calculateCost(item.SK, item.inputToken, item.outputToken);
+            const cost = calculateCost(
+              item.SK,
+              item.inputToken,
+              item.outputToken,
+              item.cacheReadInputTokens,
+              item.cacheWriteInputTokens
+            );
             return (
               `Model: ${item.SK}\n` +
               `Input tokens: ${item.inputToken}\n` +
               `Output tokens: ${item.outputToken}\n` +
+              `Cache Read tokens: ${item.cacheReadInputTokens}\n` +
+              `Cache Write tokens: ${item.cacheWriteInputTokens}\n` +
               `Cost: ${cost.toFixed(4)} USD`
             );
           })
           .join('\n\n');
 
         const totalCost = tokenUsage.reduce((acc, item) => {
-          return acc + calculateCost(item.SK, item.inputToken, item.outputToken);
+          return (
+            acc +
+            calculateCost(
+              item.SK,
+              item.inputToken,
+              item.outputToken,
+              item.cacheReadInputTokens,
+              item.cacheWriteInputTokens
+            )
+          );
         }, 0);
 
         const totalInputTokens = tokenUsage.reduce((acc, item) => acc + item.inputToken, 0);
         const totalOutputTokens = tokenUsage.reduce((acc, item) => acc + item.outputToken, 0);
+        const totalCacheReadTokens = tokenUsage.reduce((acc, item) => acc + item.cacheReadInputTokens, 0);
+        const totalCacheWriteTokens = tokenUsage.reduce((acc, item) => acc + item.cacheWriteInputTokens, 0);
 
         const historyText =
           `=== Token Usage Summary ===\n` +
           `Total Input Tokens: ${totalInputTokens}\n` +
           `Total Output Tokens: ${totalOutputTokens}\n` +
+          `Cache Read tokens: ${totalCacheReadTokens}\n` +
+          `Cache Write tokens: ${totalCacheWriteTokens}\n` +
           `Total Cost: ${totalCost.toFixed(4)} USD\n\n` +
           `=== Per Model Breakdown ===\n` +
           `${tokenSummary}\n\n` +

slack-bolt-app/src/util/cost.ts

@@ -1,11 +1,23 @@
-const modelPricing: Record<string, { input: number; output: number }> = {
-  '3-7-sonnet': { input: 0.003, output: 0.015 },
-  '3-5-sonnet': { input: 0.003, output: 0.015 },
-  '3-5-haiku': { input: 0.0008, output: 0.004 },
+const modelPricing = {
+  '3-7-sonnet': { input: 0.003, output: 0.015, cacheRead: 0.0003, cacheWrite: 0.00375 },
+  '3-5-sonnet': { input: 0.003, output: 0.015, cacheRead: 0.0003, cacheWrite: 0.00375 },
+  '3-5-haiku': { input: 0.0008, output: 0.004, cacheRead: 0.00008, cacheWrite: 0.001 },
 };
 
-export const calculateCost = (modelId: string, inputTokens: number, outputTokens: number) => {
+export const calculateCost = (
+  modelId: string,
+  inputTokens: number,
+  outputTokens: number,
+  cacheReadTokens: number,
+  cacheWriteTokens: number
+) => {
   const pricing = Object.entries(modelPricing).find(([key]) => modelId.includes(key))?.[1];
   if (pricing == null) return 0;
-  return (inputTokens * pricing.input + outputTokens * pricing.output) / 1000;
+  return (
+    (inputTokens * pricing.input +
+      outputTokens * pricing.output +
+      cacheReadTokens * pricing.cacheRead +
+      cacheWriteTokens * pricing.cacheWrite) /
+    1000
+  );
 };