Set up Lotti with onboarding
Onboarding connects the minimum pieces needed for Lotti's AI-assisted task workflow: one provider, a recording style, at least one category, and a first real task. It is a guided setup, not a separate demo environment—providers, categories, and tasks created here remain in the normal app.
Open Settings → Onboarding whenever you want to start or replay the flow. The page also shows whether this profile has already created its first AI-structured task.
Start with the welcome promise
Select Choose your AI brain to begin. Choosing Look around first, tapping the dimmed area outside the panel, and using the system back action all leave onboarding without connecting anything. You can return from Settings later.
Lotti may also offer this welcome automatically to an eligible profile. The automatic prompt has a limited retry cadence; the Settings entry does not, so you always have a way back.
Choose where task structuring runs
The provider page identifies the available hosted and local choices:
- Melious.ai is the recommended EU-hosted default with dynamic model routing.
- Mistral, Gemini, and Qwen expose their corresponding hosted regions.
- OpenAI is available under the expanded options.
- Ollama connects to a compatible local endpoint and does not require a hosted API key.
For a hosted provider, paste the provider API key. Lotti verifies the connection before enabling Connect. Ollama verifies the configured local endpoint and reports reachable models. A successful connection saves the provider and seeds the models and inference profile used by the remaining steps.
Captured task text is sent for structuring to the provider assigned to the selected category. Choose a provider whose deployment and data-handling model fits the material you plan to capture. A reachable local Ollama endpoint keeps that structuring request on the machine running it.
Pick the recording feel and work areas
After connection succeeds, choose the recording style used by the first-task capture and later voice-capture surfaces. You can change it again under Settings → Recording Style.
Next, select one or more work areas. Presets cover common categories such as Work, Fitness, Family, and Friends; Add your own creates a category with a name that matches your workspace. Onboarding reuses an existing category with the same name instead of creating a duplicate, restores it when necessary, and connects it to the newly seeded inference profile.
The manual fixture creates two real destinations—Penguin Operations and Mission Control—so the next step demonstrates the same category choice a working account sees.
Capture the first real task
The first-task panel uses Lotti's normal capture pipeline. Choose the target category, then use one of three paths:
- Select the recording visual and speak the task in a normal sentence.
- Choose a starter suggestion to try the structuring path immediately.
- Choose Rather type? and enter the thought without using the microphone.
Lotti echoes the captured words while it structures them. The result is a real task in the chosen category, not onboarding-only sample data. When structuring produces checklist suggestions, they arrive on the task as proposals for you to confirm instead of changing the task silently.
Choose the created task card to close onboarding and open its normal task page. From there you can review checklist proposals, add cover art or evidence, set a due date and priority, and bring the task into Daily OS.
If you close the flow after connecting a provider, Lotti keeps that provider setup and treats the welcome as completed. Replaying onboarding later remains safe: it is an explicit Settings action rather than another automatic prompt.
Continue with create your first task, manage categories, and configure providers.