Full Setup Guide

A step-by-step walkthrough of the Suitcase setup wizard.

Before You Start

Decide on your LLM provider. Suitcase works with:

• Relay with llama.cpp (recommended local, free) — local OpenAI-compatible server for GGUF models
• LM Studio (local, free) — runs models on your GPU
• Ollama (local, free) — simple local model runner
• OpenAI (cloud, paid) — GPT-4o, GPT-4o-mini
• Anthropic (cloud, paid) — Claude Sonnet, Claude Haiku
• DeepSeek (cloud, paid) — DeepSeek Chat
• Custom endpoint — any OpenAI/Anthropic-compatible server

See LLM Providers for detailed configuration per provider.

Step 1: Profile

Enter your name, target role, location, and career direction. This is the foundation — Suitcase uses these to match job leads, score fit, and tailor your materials.

TIP

The target role is the single most important field. Be specific: "Staff Platform Engineer" is better than "Software Engineer."

Step 2: GitHub (Optional)

Connect your GitHub username to let Suitcase scan your repos for career evidence — languages, project descriptions, contribution patterns. A personal access token enables access to private repos and higher rate limits.

Skip this if your best work lives behind a corporate firewall. Suitcase can extract evidence from your resume and guided discovery questions instead.

Step 3: LinkedIn & Discord (Optional)

• LinkedIn: Add your profile URL as a data source
• Discord: Connect a bot to chat with Suitcase through DMs

Both are optional. Suitcase works perfectly with just the web UI.

Step 4: Resume

Upload a PDF, text, or markdown resume. Suitcase parses it into your profile — extracting skills, experience, and education. Parsed data is treated as evidence (not truth) and can be reviewed and corrected.

Step 5: GitHub Sync (Optional)

After connecting GitHub, sync your repos. Suitcase pulls public repository signals and summarizes the evidence it can use for your career profile.

Step 6: LLM Endpoint

This is where you connect your model. The setup wizard supports two modes:

• OpenAI-compatible — for Relay with llama.cpp, OpenAI, LM Studio, Ollama, DeepSeek, Groq, Together, and most self-hosted servers
• Anthropic-compatible — for Anthropic's API directly

Required fields:

• Base URL — your endpoint's address
• Model — the model name/ID
• API key — your authentication key

Optional: Custom headers — for Cloudflare Access, custom auth, etc. Enter as JSON:

{"CF-Access-Client-Id": "...", "CF-Access-Client-Secret": "..."}

When you click Save and verify, Suitcase:

1. Saves the config server-side (API key is stored in .env, never exposed to the browser)
2. Tests the endpoint — confirms the base URL is reachable and the model is listed
3. Primes the model — sends Suitcase's system prompt, world rules (WORLD.md), personality (SOUL.md), self-model (SELF.md), and your profile facts to the LLM. This means the model already knows who it is and what it knows about you before the first conversation starts.

Model Size Matters

Suitcase's system prompt, world context, and tool definitions are substantial. Very small models (<8B parameters) may struggle with instruction following. The smallest model we've verified working well is Gemma 4 26B MoE (A4B active) IQ4_XS quantized via Relay with llama.cpp. For OpenAI/Anthropic, GPT-4o-mini and Claude Haiku work well.

Step 7: Finish

Review the checklist. All required items must show Complete before the admin console unlocks. Required items are:

• Profile
• Resume
• LLM endpoint (verified)

GitHub, LinkedIn, Discord, and GitHub sync are optional. Once ready, click Enter admin console.

After Setup

The navbar shows a "Suitcase wants to chat" button. Click it to start your first conversation. After the intro, the button becomes a regular Chat link.

From the admin console, you can:

• Re-check integration status
• Sync GitHub
• Start/stop the runtime
• Export or reset your database
• Return to the setup wizard to change configuration