memory-setup

name	memory-setup
description	Set up persistent agent memory. Configures identity, checks dependencies, and prepares the memory store. Use when asked to "set up memory", "configure memory", or when memory_status shows unconfigured state.

Memory Setup

Set up pi-mempalace for persistent agent memory across sessions.

Prerequisites

No external dependencies required. Everything runs in Node.js via @huggingface/transformers.

The first operation that needs embeddings (search or store) will download the all-MiniLM-L6-v2 model (~90MB, cached automatically).

Setup Steps

1. Create memory directory

mkdir -p ~/.pi/agent/memory

2. Configure identity

Ask the user about themselves and write ~/.pi/agent/memory/identity.txt. This is loaded into every session (~100 tokens). Keep it concise.

Template:

I am an AI assistant for [Name].
Projects: [list active projects with brief descriptions]
Preferences: [coding style, frameworks, communication style]
Team: [key people and their roles, if applicable]

Example:

I am an AI assistant for Alice.
Projects: Driftwood (SaaS analytics dashboard, React + PostgreSQL), Orion (CLI tool in Rust).
Preferences: TypeScript, functional style, clear error messages, test-driven.
Team: Bob (backend lead), Maya (design), Soren (DevOps).

3. Set default project

# The memory extension auto-detects from the current directory,
# but you can set a default:

Use /memory project <name> to set the current project context.

4. Verify setup

Run the memory_status tool to confirm everything is working:

• Identity should show ✅
• Total memories should be 0 (fresh install)
• Backend should show "pure TypeScript (in-process)"

5. Optional: Import existing conversations

If the user has exported conversations (from Claude, ChatGPT, etc.), help them store key exchanges:

# Read the export file, extract important exchanges, and use memory_save
# for each one. Focus on decisions, architecture choices, and key context.

How Memory Works

Once set up, memory works automatically:

• Auto-capture: Every conversation turn is embedded and stored directly (no buffering)
• Topic inference: Auto-capture and topicless manual saves use local flan-t5-small to choose a durable topic label, falling back to general
• Wake-up: Each new session starts with your identity + recent context (~600-900 tokens)
• Search: Use memory_search to find past conversations by meaning
• Save: Use memory_save to explicitly remember important decisions
• Recall: Use memory_recall to browse by project or topic

Configuration

Config is stored at ~/.pi/agent/memory/config.json:

{
  "autoCapture": true,
  "wakeUpEnabled": true,
  "wakeUpMaxTokens": 800,
  "topicInferenceEnabled": true,
  "defaultProject": null
}

Commands

• /memory status — show memory overview
• /memory project <name> — set current project
• /memory on / /memory off — enable/disable memory
• /memory topics on / /memory topics off — enable/disable local topic inference
• /memory search <query> — quick search