After storing the new memory in the database, the returned message includes the memory storage ID

[petersing]

Problem / Motivation

Feature Request: Add Memory ID to memory_store Return Value

📋 Summary

Add Memory ID to the return text of memory_store tool to prevent AI hallucination and provide verifiable confirmation of successful memory storage.

---

🐛 Problem Description

Current Behavior

When calling the memory_store tool, the return value only shows:

Stored: "..." in scope 'agent:main'

The Memory ID exists in the details.id field but is NOT visible in the response text shown to the AI model.

Impact

This creates several critical issues:

1. AI Hallucination Risk: The AI cannot verify whether a memory was actually stored successfully. It may claim "已更新記憶 📂" without having proof of the actual Memory ID.

2. No Verifiable Confirmation: Users cannot confirm that their important data (preferences, facts, decisions) was actually persisted to the memory system.

3. Debugging Difficulty: When troubleshooting memory-related issues, there's no way to reference a specific memory entry by ID.

4. Trust Issues: Users may lose confidence in the system if the AI claims to have stored memories but cannot provide concrete evidence (ID) of the operation.

---

Proposed Solution

✅ Proposed Solution

Change

Modify the return text format in memory_store tool to include the Memory ID:

Before:

text: `Stored: "${text.slice(0, 100)}${text.length > 100 ? "..." : ""}" in scope '${targetScope}'`,

After:

text: `Stored: "${text.slice(0, 100)}${text.length > 100 ? "..." : ""}" (ID: ${entry.id}) in scope '${targetScope}'`,

Example Output

Stored: "Test Store..." (ID: 8f82f232-abc1-4def-9012-345680501234) in scope 'agent:main'

---

🎯 Benefits

1. Prevents AI Hallucination: AI can only claim memory was stored if it has the actual ID as proof.

2. User Confidence: Users can see concrete evidence that their data was persisted.

3. Audit Trail: Each memory operation has a unique identifier for tracking and debugging.

4. Compliance with Best Practices: Follows the principle of "explicit confirmation for state-changing operations."

5. Enables Advanced Features: Future features like memory editing, deletion, or referencing can use these IDs.

---

🔧 Implementation Details

File to Modify

• Path: /root/node_modules/memory-lancedb-pro/src/tools.ts
• Function: registerMemoryStoreTool
• Line: ~578 (return statement in execute function)

Code Change

  return {
    content: [
      {
        type: "text",
-       text: `Stored: "${text.slice(0, 100)}${text.length > 100 ? "..." : ""}" in scope '${targetScope}'`,
+       text: `Stored: "${text.slice(0, 100)}${text.length > 100 ? "..." : ""}" (ID: ${entry.id}) in scope '${targetScope}'`,
      },
    ],
    details: {
      action: "created",
      id: entry.id,
      scope: entry.scope,
      category: entry.category,
      importance: entry.importance,
    },
  };

---

📝 Use Case Example

Scenario: User Shares Important Preference

User: "我未來 5 年計劃挑戰 traffic around the world"

AI Action: Calls memory_store with this information

Before (Problem):

• AI receives: Stored: "我未來 5 年計劃挑戰..." in scope 'agent:main'
• AI claims: "已更新記憶 📂" (but has no ID to prove it)
• Risk: AI might claim memory was stored even if the operation failed silently

After (Solution):

• AI receives: Stored: "我未來 5 年計劃挑戰..." (ID: 8f82f232-abc1-4def-9012-345678901234) in scope 'agent:main'
• AI claims: "已更新記憶 📂: 8f82f232" (with verifiable ID)
• Benefit: User can confirm the memory was actually stored with a specific ID

---

⚠️ Considerations

Privacy

• Memory IDs are UUIDs and do not contain sensitive information
• IDs are already stored in the details field, so this change only makes them visible in the response text

Performance

• No performance impact: the ID is already generated, this just includes it in the response text
• Text length increase is minimal (~50 characters for a typical UUID)

Backward Compatibility

• This is a non-breaking change
• Existing code that parses the return value can still access details.id
• Only the content.text format changes

---

🧪 Testing Recommendations

1. Unit Test: Verify that entry.id is correctly interpolated into the return text
2. Integration Test: Call memory_store and confirm the response includes a valid UUID
3. AI Behavior Test: Ensure AI models correctly use the ID in their responses
4. Edge Case Test: Test with very long text (truncation should still work correctly)

---

📚 Related Documentation

• AGENTS.md (Fiona's operational guidelines) specifies:

  「（已更新記憶 📂: <Memory_ID>）」標誌只可以在以下兩個條件同時成立時才加入回應：

  1. 本次回應中確實呼叫了 memory_store tool；且
  2. memory_store 返回結果包含實際 Memory ID

This feature enables compliance with such guidelines.

---

Alternatives Considered

No response

Area

None

Additional Context

No response

[AliceLJY]

这个需求已经有人提 PR 了：#146。方向是把 memory ID 直接放进 memory_store / memory_update 的返回文本里，方便模型和用户确认具体写入项。静态看改动不大，主要注意点是补最小回归测试，锁住新的返回文案。