Skip to content

wizard.md

Latest commit

 

History

History
89 lines (58 loc) · 2.84 KB

wizard.md

File metadata and controls

89 lines (58 loc) · 2.84 KB

memory wizard

Use wizard to bootstrap Memory Layer configuration either globally for the machine or locally for the current repository.

Table of Contents

Global vs Repo-Local

memory wizard --global configures shared machine-level settings such as:

  • database.url
  • optional shared writer.id
  • optional LLM settings, including local Ollama via provider = "ollama"

database.url must point at a reachable PostgreSQL database before the backend can become healthy. The target database should already have pgvector enabled:

psql "$DATABASE_URL" -c "SELECT 1;"
psql "$DATABASE_URL" -c "CREATE EXTENSION IF NOT EXISTS vector;"
psql "$DATABASE_URL" -c "SELECT extversion FROM pg_extension WHERE extname = 'vector';"

For local and hosted database setup examples, see Getting Started: PostgreSQL Requirement.

The shared service API token is usually generated automatically into the adjacent memory-layer.env file. The wizard only needs an explicit token value if you want to override that generated token. If you leave writer.id unset, Memory Layer derives a stable writer identity automatically at runtime.

memory wizard inside a repository is local-first and bootstraps project files such as:

  • .mem/config.toml
  • .mem/project.toml
  • .agents/memory-layer.toml
  • .agents/skills/

Inside a repository, the wizard defaults to repo-local scope unless you explicitly choose shared/global setup.

What It Creates

At repo scope, the wizard creates:

  • .mem/ runtime and repo-local config files
  • .agents/memory-layer.toml
  • a repo-local copy of the Memory Layer skill bundle

The bundled skills are created from the packaged skill-template, or from the repo-local template during source/dev usage. That repo-local bundle now uses a shared Go helper under .agents/skills/memory-layer/scripts/, so go must be available on PATH anywhere you expect the repo-local skills to run.

Typical Usage

First machine setup:

memory wizard --global

Then inside each repo:

cd /path/to/your-project
memory wizard

Preview the resulting file, token, and service actions without writing anything:

memory wizard --dry-run
memory wizard --global --dry-run

How It Differs From init

init is the lower-level repo bootstrap command.

For normal interactive setup, prefer wizard.

Use init when you want:

  • a more scriptable bootstrap path
  • a non-interactive setup flow
  • direct control over the generated repo-local files

Both memory wizard and memory init support --dry-run for preview-only setup.

For the full onboarding flow, see Getting Started.