跳到正文
TakoAPI
返回技能列表

cortex-memory

由 matthewubundi 提供

Notes & PKM
515下载量由 clawskills.sh 提供
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Long-term structured memory with knowledge graph, entity tracking, temporal reasoning, and cross-session recall.

描述

该技能的功能

Cortex Memory provides structured long-term memory using a knowledge graph that extracts facts, entities, and relationships from text. It supports cross-session recall with hybrid search combining BM25, semantic, temporal, and graph traversal methods. Use it when native memory search returns noisy or incomplete results.

Unlike flat keyword search, Cortex builds a traversable knowledge graph that tracks how entities relate and how facts change over time, enabling queries that native memory cannot answer.

何时使用

  • Recalling decisions made in past coding sessions
  • Tracking how team members relate to specific projects
  • Finding when a preference or fact changed over time
  • Looking up entity relationships like service dependencies
  • Storing key context from a productive work session for later retrieval

示例工作流

输入

User asks: How does the auth service relate to the API gateway?

智能体
  1. 1

    Check if <cortex_memories> tag is already present in context

  2. 2

    If not, construct a factual retrieval query targeting the relationship between auth service and API gateway

  3. 3

    Send POST request to /v1/retrieve with query_type factual and full mode for graph traversal

  4. 4

    Parse results filtering for relationship edges between the two entities

  5. 5

    Present the retrieved relationships to the user

输出

A list of stored facts and graph edges describing how the auth service connects to the API gateway, including any temporal changes

前置要求

  • CORTEX_API_KEY environment variable
  • CORTEX_BASE_URL environment variable
  • Cortex API account (https://github.com/Ubundi/Cortex)

安装

clawhub install matthewubundi/cortex-memory

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接