Aller au contenu
TakoAPI
Retour aux skills

memo-persistent-memory

par webdevtodayjason

Search & Research
—Pas encore noté
Ta note

Résumé

Persistent memory system - automatic context capture and semantic search.

Description

Ce que fait ce skill

Adds persistent memory to OpenClaw sessions using a local SQLite database with full-text search. Memories are captured automatically after each response and injected before the next prompt, keeping context available across sessions.

Unlike in-session context that resets when the conversation ends, this system persists observations to a local database and retrieves only semantically relevant memories, avoiding token bloat.

Quand l'utiliser

  • Remembering project-specific decisions across multiple work sessions
  • Recalling user preferences without re-explaining them each time
  • Tracking ongoing tasks and their current state between conversations
  • Surfacing relevant past context when starting a new related task
  • Storing and retrieving notes from previous debugging sessions

Exemple de workflow

ENTRÉE

User asks: 'What was the database schema we decided on last week?'

AGENT
  1. 1

    auto-recall triggers and searches the memory store for relevant observations

  2. 2

    memory_search queries the SQLite FTS5 index with natural language

  3. 3

    matching memories are injected into the prompt context before the agent responds

  4. 4

    agent uses the retrieved schema details to answer directly

SORTIE

Agent answers with the previously stored schema decision without the user needing to repeat it

Installation

clawhub install webdevtodayjason/memo-persistent-memory

Exécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.

Liens