First Run Wizard

When you launch Memo for the first time, the setup wizard guides you through configuring your assistant. The wizard runs once — all settings can be changed later via the Settings panel.

Wizard Flow

Step 1: Welcome

A clean splash screen with the Memo logo. Click "Get Started" to begin.

Step 2: Your Name

Enter your name. This is used for:

• Personalizing the assistant's responses
• Calendar event attribution
• WhatsApp message sender identification
• Session and backup metadata

Step 3: Assistant Name

Give your assistant a name. This is the persona label used in chat, system prompts, and the engine strip. Examples: "Memo", "Athena", "Dave".

Step 4: Personality & System Prompt

Choose a preset persona or write your own system prompt:

Preset	Style	Best For
Casual	Conversational, friendly	Daily chat, brainstorming
Formal	Professional, precise	Work documents, technical writing
Technical	Concise, code-focused	Programming, system administration
Creative	Expressive, imaginative	Writing, art, ideas
Fun	Playful, humorous	Entertainment, casual banter
Buddy	Warm, supportive	Personal companion, reflection
Custom	You write the prompt	Full control over personality

The system prompt defines how Memo behaves — tone, verbosity, role, and constraints. It is stored in data/identity/system-prompt.yaml and can be edited anytime at Settings → System Prompt.

You can use the incognitoPrompt field for a separate personality during incognito sessions. Both prompts support the identity style system (identity.go, styles.go).

Step 5: llama.cpp Setup

Memo checks if the bundled llama-server binary is present and functional:

• If found: Shows version and proceeds
• If missing: Auto-downloads the appropriate binary for your platform (Windows, Linux, macOS)
• GPU detection: Scans for NVIDIA CUDA, AMD ROCm, and Apple Metal capabilities

The download runs in the background. You'll see a progress indicator and can continue with configuration.

Step 6: Model Discovery

Memo scans the HuggingFace GGUF catalog and presents recommended models:

• Recommended tab: Official models from verified publishers (Google, Microsoft, Qwen/Alibaba, nomic)
• Search tab: Filter by size (1–8B, 8–14B, 14B+) or capability (Tools, Vision, Code)
• Hardware fit badge: Each model shows whether it fits your machine — "✓ Fits", "~ CPU OK", or "× Too large"

Choose an embedding model for RAG memory (default: nomic-embed-text-v1.5) and optionally a chat model. Downloads start immediately.

You can skip model download and return to the Model Store later. Memo works with cloud providers (OpenAI, Claude, Gemini) even without local models — they just need API keys.

After the Wizard

Once setup completes, first-time users see the Launchpad — a welcome screen with five feature cards:

• Chat — Write, get answers, let it remember you
• Agent — Manages files, runs commands, does tasks for you
• Orchestra — Run multiple models working as a team
• WhatsApp — Manage your chats with AI assistance
• Calendar — Auto-captures your plans from conversations

A 4-step Spotlight Tour then highlights navigation icons one at a time. Users with existing chat history skip the launchpad and go directly to chat.

Re-running the Wizard

The wizard runs only once. To reconfigure:

• Personality: Settings → System Prompt
• Name: Settings → General
• Models: Model Store (left sidebar)
• llama.cpp: Settings → GPU Config
• Spotlight Tour: Settings → General → "Re-run Setup Tour"

Configuration File

The wizard writes to config/config.yaml:

user_name: "Buğra"
assistant_name: "Memo"
language: "en"
system_prompt: "You are Memo, a helpful and concise AI assistant..."
incognito_prompt: ""
theme: "pewter"

All values can be edited directly in the file or through the Settings UI. Changes take effect immediately — no restart required.