From 93050b688949d01e55bb8a6f23b79ee86a0ea93e Mon Sep 17 00:00:00 2001
From: hailin <hailin.zhao@gdzx.xyz>
Date: Sat, 10 Jan 2026 01:42:33 -0800
Subject: [PATCH] perf(claude): enable Prompt Caching for ~90% cost savings on
 system prompt

---
 .../claude/claude-agent.service.ts            | 27 ++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)

diff --git a/packages/services/conversation-service/src/infrastructure/claude/claude-agent.service.ts b/packages/services/conversation-service/src/infrastructure/claude/claude-agent.service.ts
index 798cfd5..dfbee50 100644
--- a/packages/services/conversation-service/src/infrastructure/claude/claude-agent.service.ts
+++ b/packages/services/conversation-service/src/infrastructure/claude/claude-agent.service.ts
@@ -70,6 +70,7 @@ export class ClaudeAgentService implements OnModuleInit {
 
   /**
    * Send a message and get streaming response with tool loop support
+   * Uses Prompt Caching to reduce costs (~90% savings on cached system prompt)
    */
   async *sendMessage(
     message: string,
@@ -101,15 +102,25 @@ export class ClaudeAgentService implements OnModuleInit {
     const maxIterations = 10; // Safety limit
     let iterations = 0;
 
+    // System prompt with cache_control for Prompt Caching
+    // Cache TTL is 5 minutes, cache hits cost only 10% of normal input price
+    const systemWithCache: Anthropic.TextBlockParam[] = [
+      {
+        type: 'text',
+        text: systemPrompt,
+        cache_control: { type: 'ephemeral' },
+      },
+    ];
+
     while (iterations < maxIterations) {
       iterations++;
 
       try {
-        // Create streaming message
+        // Create streaming message with cached system prompt
         const stream = await this.client.messages.stream({
           model: 'claude-sonnet-4-20250514',
           max_tokens: 4096,
-          system: systemPrompt,
+          system: systemWithCache,
           messages,
           tools: tools as Anthropic.Tool[],
         });
@@ -243,6 +254,7 @@ export class ClaudeAgentService implements OnModuleInit {
 
   /**
    * Non-streaming message for simple queries
+   * Uses Prompt Caching for cost optimization
    */
   async sendMessageSync(
     message: string,
@@ -267,10 +279,19 @@ export class ClaudeAgentService implements OnModuleInit {
       content: message,
     });
 
+    // System prompt with cache_control for Prompt Caching
+    const systemWithCache: Anthropic.TextBlockParam[] = [
+      {
+        type: 'text',
+        text: systemPrompt,
+        cache_control: { type: 'ephemeral' },
+      },
+    ];
+
     const response = await this.client.messages.create({
       model: 'claude-sonnet-4-20250514',
       max_tokens: 4096,
-      system: systemPrompt,
+      system: systemWithCache,
       messages,
       tools: tools as Anthropic.Tool[],
     });