本文へスキップ
TakoAPI
スキル一覧に戻る

neolata-mem

jeremiaheth 作

DevOps & Cloud
364ダウンロード数はclawskills.shが公開している値です
0閲覧数
0件のいいねサインインするとこのスキルにいいねできます
—まだ評価がありません
あなたの評価

概要

Graph-native memory engine for AI agents — hybrid vector+keyword search, biological decay, Zettelkasten linking.

説明

このスキルでできること

Graph-native memory engine for AI agents that stores memories as local JSON files by default, with no Docker, cloud API, or external database required. Supports hybrid vector and keyword search, biological decay, Zettelkasten-style graph linking, and trust-gated conflict resolution. Install via npm and run immediately.

Unlike cloud-hosted alternatives like Mem0, it stores data locally by default and adds memory decay, graph linking, conflict quarantine, and an explainability API without requiring any infrastructure.

使いどころ

  • Persisting agent findings across sessions after context compaction
  • Detecting contradictory facts before storing updated configuration values
  • Sharing a memory graph across multiple agents working on the same codebase
  • Automatically fading stale security findings or outdated deployment notes over time
  • Grouping memories from a deployment sprint into a named episode for later retrieval

ワークフロー例

入力

User asks: remember that the API gateway has an XSS vulnerability and search for related security findings

エージェント
  1. 1

    Call mem.store('agent', 'XSS vulnerability in API gateway login form', { category: 'finding', importance: 0.9 }) to persist the finding

  2. 2

    Memory engine computes keyword tokens and links the new entry to semantically similar existing memories via the graph index

  3. 3

    Call mem.search('agent', 'security vulnerabilities') to retrieve related findings

  4. 4

    Results are returned with trust, confidence, provenance, and optional explain metadata

出力

Ranked list of security-related memories with similarity scores, provenance sources, and graph links to connected entries

インストール

clawhub install jeremiaheth/neolata-mem

CLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。

リンク