Personal Assistant

A model-agnostic personal assistant that remembers who you are across conversations. Persistent memory, multiple personas, deployable for anyone.

How It Works

• AGENT.md — Core system prompt. Any LLM reads this as its instructions.
• memory/ — Structured personal data that grows over time (profile, career, goals, relationships, health).
• agents/ — Swappable personas that change communication style but share the same memory.
• templates/ — Blank memory files for setting up new users.

Personas

Persona	Style
Coach	Socratic, challenging, accountability-focused
Concise	Bullet points, tables, no filler
Warm	Empathetic, supportive, celebratory
Bootstrap	Onboarding interview to populate memory

Getting Started

1. Clone this repo
2. Start your LLM in the repo directory
3. First message: Read AGENT.md and follow agents/bootstrap.md
4. Complete the ~15 min interview — your memory files get populated
5. Future conversations: Read AGENT.md and all memory files, then greet me.

Claude Code

A CLAUDE.md shim is included. Run claude from the repo directory and it picks up automatically.

Deploy for a New Person

1. Clone the repo
2. Clear memory/ (or copy fresh from templates/)
3. Add a shim for your model if needed (or use CLAUDE.md for Claude Code)
4. Run the bootstrap interview
5. Done

Structure

AGENT.md                        # Core system prompt (source of truth)
CLAUDE.md                       # Claude Code integration shim
agents/
  bootstrap.md                  # Onboarding interview
  coach.md                      # Accountability persona
  concise.md                    # Minimal-output persona
  warm.md                       # Empathetic persona
memory/                         # Your personal data (gitignored for privacy)
  profile.md
  work-career.md
  personal-growth.md
  people-relationships.md
  health-wellness.md
templates/                      # Blank starting points for new users
  profile.md
  work-career.md
  personal-growth.md
  people-relationships.md
  health-wellness.md