recall-memory

# Recall Memory

Retrieve information stored across sessions using memU's agentic memory framework.

## Overview

This skill retrieves information from memU - a three-layer memory hierarchy that processes conversations into structured knowledge:

- **Resources**: Raw conversation data from past sessions
- **Items**: Extracted memory units (preferences, skills, opinions, habits, facts)
- **Categories**: Aggregated summaries with full traceability

**How it works**: Conversations are automatically memorized when sessions end (via SessionEnd hook). This skill retrieves that stored knowledge when needed.

## Script Path Construction

**IMPORTANT**: Always use full paths to call scripts. Do NOT use `cd` to change to the scripts directory.

The skill is located at: **Base directory for this skill** (shown when skill loads)

To call scripts, concatenate:
- **Skill base directory** + `/scripts/` + **script name**

Example:
```bash
# If skill base is: /home/user/.claude/plugins/memu/skills/recall-memory
# Then memu.py is at:
/home/user/.claude/plugins/memu/skills/recall-memory/scripts/memu.py
```

In examples below, we use `scripts/memu.py` as shorthand, but replace `scripts/` with the full path based on the skill's base directory.

## Retrieval Methods

memU supports two retrieval methods - choose based on query complexity:

### RAG (Recommended)

**When to use**: Specific factual queries, direct questions, known topics

- Fast vector-based similarity search
- Returns results in <2 seconds
- Good for: "What are my preferences?", "Do I use TypeScript or JavaScript?", "How do I handle errors?"
- Best when you know what you're looking for

**Example**:
```bash
scripts/memu.py retrieve --query "What coding style do I prefer?" --method rag
```

### LLM

**When to use**: Complex queries, ambiguous questions, need deep understanding

- Deep semantic understanding with adaptive refinement
- Slower (~5-10 seconds) but more thorough
- Good for: "What patterns do I follow when refacto