Onboarding

Run:

odinn onboard

On a new installation, Ódinn offers:

• Quick setup for the shortest supported path.
• Guided setup for provider, model, and capability choices.
• Blank slate when you want to configure everything later.
• Detected OpenClaw or Hermes import paths when compatible state is available.

Ódinn stages onboarding changes before they touch the working configuration. Existing configuration is backed up, writes are atomic, and cancellation leaves the current setup unchanged. A connected setup is not accepted until a real model response succeeds and you approve the final review.

Choose a model path

For browser-based OpenAI/ChatGPT sign-in, choose OpenAI in guided setup. For a local Ollama model:

odinn onboard --provider ollama --model <installed-model>

For an API-key setup, place the key in its documented environment variable and select the matching provider:

export OPENAI_API_KEY="..."
odinn onboard --provider openai --auth api-key

Secrets are not stored in the ordinary configuration file. OAuth credentials live in the owner-only OAuth store.

Returning to onboarding

Running odinn onboard again does not silently rebuild an existing installation. It offers focused actions such as:

• open the current console;
• repair the connection;
• change the AI or model;
• review capabilities;
• open advanced settings;
• start over with separate confirmation.

Custom capability policies are preserved unless you explicitly replace them.

Verify later

Test the configured account and model through the same runtime path used by chat:

odinn onboard --verify --non-interactive

If verification fails, onboarding distinguishes rejected credentials, unavailable models, exhausted usage, timeouts, offline local providers, and port conflicts. See Troubleshooting for the next action.