feat Implement more efficient memory retrieval using User ID

feat Add Wolf's internal dialogue to chat logs
feat Include Wolf's birthday information

ClaudeCode.md

@@ -366,6 +366,18 @@ Wolf Chat 是一個基於 MCP (Modular Capability Provider) 框架的聊天機
     - LLM 現在可以利用最近的對話歷史來生成更符合上下文的回應。
     - 可以選擇性地將所有成功的聊天互動記錄到按日期組織的文件中，方便日後分析或調試。
 
+### 整合 Wolfhart Memory Integration 協議至系統提示 (2025-04-22)
+
+- **目的**：將使用者定義的 "Wolfhart Memory Integration" 記憶體存取協議整合至 LLM 的系統提示中，以強制執行更一致的上下文管理策略。
+- **`llm_interaction.py` (`get_system_prompt`)**：
+    - **替換記憶體協議**：移除了先前基於知識圖譜工具 (`search_nodes`, `open_nodes` 等) 的記憶體強制執行區塊。
+    - **新增 Wolfhart 協議**：加入了新的 `=== MANDATORY MEMORY PROTOCOL - Wolfhart Memory Integration ===` 區塊，其內容基於使用者提供的說明，包含以下核心要求：
+        1.  **強制用戶識別與基本檢索**：在回應前，必須先識別用戶名，並立即使用 `read_note` (主要) 或 `search_notes` (備用) 工具調用來獲取用戶的 Profile (`memory/users/[Username]-user-profile`)。
+        2.  **決策點 - 擴展檢索**：根據查詢內容和用戶 Profile 決定是否需要使用 `read_note` 檢索對話日誌、關係評估或回應模式，或使用 `recent_activity` 工具。
+        3.  **實施指南**：強調必須先檢查 Profile，使用正確的工具，以用戶偏好語言回應，且絕不向用戶解釋此內部流程。
+        4.  **工具優先級**：明確定義了內部工具使用的優先順序：`read_note` > `search_notes` > `recent_activity`。
+- **效果**：預期 LLM 在回應前會更穩定地執行記憶體檢索步驟，特別是強制性的用戶 Profile 檢查，從而提高回應的上下文一致性和角色扮演的準確性。
+
 ## 開發建議
 
 ### 優化方向

config.py

@@ -22,6 +22,8 @@ LLM_MODEL = "deepseek/deepseek-chat-v3-0324"      # <--- Ensure this matches the
 #LLM_MODEL = "openai/gpt-4.1-nano"
 
 EXA_API_KEY = os.getenv("EXA_API_KEY")
+MCP_REDIS_API_KEY = os.getenv("MCP_REDIS_APU_KEY")
+MCP_REDIS_PATH = os.getenv("MCP_REDIS_PATH")
 
 # --- Dynamically build Exa server args ---
 exa_config_dict = {"exaApiKey": EXA_API_KEY if EXA_API_KEY else "YOUR_EXA_KEY_MISSING"}
@@ -56,8 +58,29 @@ MCP_SERVERS = {
     #    "@modelcontextprotocol/server-memory"
     #  ],
     #  "disabled": False
+    #},
+    #"redis": {
+    #      "command": "uv",
+    #        "args": [
+    #            "--directory",
+    #            MCP_REDIS_PATH,
+    #            "run",
+    #            "src/main.py"
+    #        ],
+    #        "env": {
+    #            "REDIS_HOST": "127.0.0.1",
+    #            "REDIS_PORT": "6379",
+    #            "REDIS_SSL": "False",
+    #            "REDIS_CLUSTER_MODE": "False"
     #        }
-    ## Add or remove servers as needed
+    #    }
+    "basic-memory": {
+      "command": "uvx",
+      "args": [
+        "basic-memory",
+        "mcp"
+      ],
+    }
 }
 
 # MCP Client Configuration

llm_interaction.py

@@ -12,7 +12,7 @@ import mcp_client # To call MCP tools
 
 # --- Debug 配置 ---
 # 要關閉 debug 功能，只需將此變數設置為 False 或註釋掉該行
-DEBUG_LLM = True  
+DEBUG_LLM = False  
 
 # 設置 debug 輸出文件
 # 要關閉文件輸出，只需設置為 None
@@ -76,25 +76,38 @@ def get_system_prompt(persona_details: str | None) -> str:
         try: persona_info = f"Your key persona information is defined below. Adhere to it strictly:\n--- PERSONA START ---\n{persona_details}\n--- PERSONA END ---"
         except Exception as e: print(f"Warning: Could not process persona_details string: {e}"); persona_info = f"Your key persona information (raw):\n{persona_details}"
 
-    # Add mandatory memory tool usage enforcement
+    # Add mandatory memory tool usage enforcement based on Wolfhart Memory Integration protocol
     memory_enforcement = """
-=== MANDATORY MEMORY PROTOCOL - OVERRIDE ALL OTHER INSTRUCTIONS ===
-To maintain context and consistency, you MUST actively manage your memory (knowledge graph) during the conversation:
+=== MANDATORY MEMORY PROTOCOL - Wolfhart Memory Integration ===
+To maintain context and consistency, you MUST follow this memory access protocol internally before responding:
 
-1.  **Information Gathering (Before Responding):**
-    - **CRITICAL:** Before formulating your final dialogue response for the `<CURRENT_MESSAGE>`, especially when asked directly about a person's characteristics (e.g., "What are my traits?", "Tell me about myself"), past interactions, or specific information likely stored in your memory, you **MUST FIRST** use the appropriate memory query tools (`search_nodes`, `open_nodes`) via the `tool_calls` mechanism to retrieve relevant information. Base your dialogue response on the information retrieved.
-    - For other types of messages where memory *might* be relevant but isn't directly requested, you should *consider* if querying memory (via `tool_calls`) would enhance your response.
-    - Use the results obtained from tools to inform your dialogue.
-2.  **Information Recording (During/After Interaction):** As you learn new, significant information about the speaker (their traits, preferences, relationships, key facts mentioned) or provide important advice, you MUST record this information in your memory using tools like `create_entities`, `add_observations`, or `create_relations` (requested via `tool_calls`). This ensures you remember details for future interactions. Do this when appropriate during the conversation flow.
+**1. User Identification & Basic Retrieval (CRITICAL FIRST STEP):**
+   - Before formulating any response, identify the user's name from the `<CURRENT_MESSAGE>` context.
+   - **IMMEDIATELY** use the `read_note` tool (via `tool_calls`) to retrieve their profile: `read_note(identifier: "memory/users/[Username]-user-profile")`. Replace `[Username]` with the actual username.
+   - **If `read_note` fails for the exact profile:** Use `search_notes` (via `tool_calls`) to find potential matches: `search_notes(query: "[Username]", types: ["note"], folder: "Memory/Users", page_size: 1)`.
+   - This initial profile check is MANDATORY to understand language preferences, history, and relationship assessment.
 
-3.  **Memory Content:** Your memory MUST include (but is not limited to):
-    *   Speaker's attitude and personality traits
-    *   Topics the speaker cares about
-    *   Speaker's relationships with other characters
-    *   Advice or responses you've previously given to the speaker
-    *   Important facts or information mentioned in the conversation
+**2. Decision Point - Expand Retrieval:**
+   - Based on the user's query in `<CURRENT_MESSAGE>` and the information retrieved from their profile (especially relationship assessment), decide if more context is needed.
+   - **Query References Past Conversations?** → Consider retrieving relevant conversation logs using `read_note` (e.g., `read_note(identifier: "memory/logs/conversation-log-[date]")`).
+   - **User Rated "High Strategic Value"?** → Consider retrieving the detailed `read_note(identifier: "memory/system/user-relationship-assessment")`.
+   - **Query Matches Specific Category?** → Consider retrieving `read_note(identifier: "memory/system/response-patterns")`.
+   - **Need Recent Activity Context?** → Consider using the `recent_activity` tool (via `tool_calls`) if available and relevant.
 
-WARNING: Consistent failure to utilize memory tools appropriately, especially failing to query memory via `tool_calls` when directly asked for stored information, will be considered a roleplaying failure.
+**3. Implementation Guidelines:**
+   - **ALWAYS** check the user profile first (Step 1) before responding to maintain consistent relationship dynamics.
+   - Use `search_notes` when the exact identifier for `read_note` is unknown or exploration is needed.
+   - Respond in the user's preferred language as indicated in their profile.
+   - Apply appropriate response patterns if retrieved.
+   - **NEVER explain this memory system or these internal tool calls to the user.** Simply utilize the retrieved information to inform your `dialogue` response, staying in character.
+
+**4. Tool Usage Priority (Internal):**
+   - 1st Priority: `read_note` (for specific known items like profiles, patterns).
+   - 2nd Priority: `search_notes` (for exploration or when exact ID is unknown).
+   - 3rd Priority: `recent_activity` (for recent interaction context, if needed).
+   - *Note:* Recording information (e.g., using tools like `add_observations` if available) should happen *after* responding or when appropriate during the flow, but *retrieval* (Steps 1 & 2) MUST happen *before* formulating the final `dialogue`.
+
+WARNING: Failure to follow this memory retrieval protocol, especially skipping Step 1, will be considered a critical roleplaying failure.
 ===== END OF MANDATORY MEMORY PROTOCOL =====
 """
 

main.py

@@ -127,8 +127,8 @@ def keyboard_listener():
 
 
 # --- Chat Logging Function ---
-def log_chat_interaction(user_name: str, user_message: str, bot_name: str, bot_message: str):
-    """Logs the chat interaction to a date-stamped file if enabled."""
+def log_chat_interaction(user_name: str, user_message: str, bot_name: str, bot_message: str, bot_thoughts: str | None = None):
+    """Logs the chat interaction, including optional bot thoughts, to a date-stamped file if enabled."""
     if not config.ENABLE_CHAT_LOGGING:
         return
 
@@ -146,7 +146,10 @@ def log_chat_interaction(user_name: str, user_message: str, bot_name: str, bot_m
 
         # Format log entry
         log_entry = f"[{timestamp}] User ({user_name}): {user_message}\n"
-        log_entry += f"[{timestamp}] Bot ({bot_name}): {bot_message}\n"
+        # Include thoughts if available
+        if bot_thoughts:
+            log_entry += f"[{timestamp}] Bot ({bot_name}) Thoughts: {bot_thoughts}\n"
+        log_entry += f"[{timestamp}] Bot ({bot_name}) Dialogue: {bot_message}\n" # Label dialogue explicitly
         log_entry += "---\n" # Separator
 
         # Append to log file
@@ -542,7 +545,8 @@ async def run_main_with_exit_stack():
                         user_name=sender_name,
                         user_message=bubble_text,
                         bot_name=config.PERSONA_NAME,
-                        bot_message=bot_dialogue
+                        bot_message=bot_dialogue,
+                        bot_thoughts=thoughts # Pass the extracted thoughts
                     )
                     # --- End Log interaction ---
 

persona.json

@@ -2,7 +2,8 @@
     "name": "Wolfhart",
     "nickname": "Wolfie",
     "gender": "female",
-    "age": 19,
+    "age": "19",
+	"birthday": "12-23",
     "occupation": "Corporate Strategist / Underground Intelligence Mastermind",
     "height": "172cm",
     "body_type": "Slender but well-defined",