Pensieve MCP Server

Instructions

Implementation Reference

• mcp_server/server.py:93-115 (handler)

  The handler implementation for search_conversations which scans JSON files in the storage directory for the query string.

  def search_conversations(query: str, limit: int = 20) -> List[Dict[str, Any]]:
      """대화 내용 검색"""
      results = []
      query_lower = query.lower()
      
      for file_path in STORAGE_DIR.glob("*.json"):
          try:
              with open(file_path, 'r', encoding='utf-8') as f:
                  data = json.load(f)
                  
                  # 메시지 내용에서 검색
                  for message in data.get("messages", []):
                      content = message.get("content", "").lower()
                      if query_lower in content:
                          results.append({
                              "id": data["id"],
                              "metadata": data.get("metadata", {}),
                              "created_at": data.get("created_at"),
                              "matched_message": message,
                              "message_count": len(data.get("messages", []))
                          })
                          break

• mcp_server/server.py:202-214 (registration)

  MCP Tool registration for search_conversations, including its description and input schema.

  Tool(
      name="search_conversations",
      description="대화 내용을 검색합니다",
      inputSchema={
          "type": "object",
          "properties": {
              "query": {
                  "type": "string",
                  "description": "검색할 텍스트"
              },
              "limit": {
                  "type": "integer",
                  "description": "최대 결과 수 (기본값: 20)",