General Settings
The General section groups connection, model defaults, app behavior, experimental features, voice and language settings. Changes here are applied with the Save Changes button in the footer.
Connection
API Base URL — the Kognar backend this app talks to. Below the field, a status line shows a green dot and Connected when a base URL is set, or Not connected when it is empty.
Unless you were told otherwise by your administrator, leave this field as it comes.
Default models
Set the models that are pre-selected when you start new conversations or use AI features.
| Setting | Description |
|---|---|
| Chat model | Default model for text conversations |
| Image generation model | Default model for generating images |
| Image editing model | Default model for editing existing images |
If a model you saved as a default becomes unavailable in your account's catalog, Kognar automatically falls back to another active model.
See Models & Providers for the full catalog and how the selector works.
Behavior
| Setting | Description |
|---|---|
| Save chat history locally | When enabled, conversations are stored in the local database. When disabled, conversations are not persisted after the session ends. |
| Semantic context by project | The default for new projects. Each project can override it in its own project instructions. |
| Enable UI cache | Caches the interface for faster startup. |
Experimental features
A warning card explains that experimental features may be unstable and can change or break between releases.
| Feature | Description |
|---|---|
| Local decision model | On-device router that can trim tools before each request to save tokens. Apple Silicon only. Off by default. |
Voice & transcription
| Setting | Description |
|---|---|
| Transcription engine | Automatic (recommended), Local Whisper only, or Cloud only (Gemini) |
| Voice model size | For local transcription: Tiny, Base (recommended), or Small |
| Download / load model | Prepares the local voice model ahead of first use, with a progress bar |
The hint below the engine selector changes with your choice:
- Automatic — uses the best available path: macOS native → local Whisper → cloud. No setup needed.
- Local Whisper only — always uses local Whisper, requires the model download, and falls back to cloud if local fails.
- Cloud only (Gemini) — always sends audio to Gemini through your chat connection. No model download needed.
Below the model-size selector, Kognar reports whether the model is ready, not yet loaded in memory, or a different size than the one you selected.
Voice model settings apply to the desktop app only.
See Audio Transcription for the full picture.
Language & region
| Setting | Description |
|---|---|
| Language | Interface language — English or Portuguese (Brazil). Applies to menus, alerts, modals and the native desktop menus. No restart required. |
| Timezone | Shows the timezone detected from your operating system. |
Where other settings went
The redesigned Settings window moved several things out of General:
| Looking for | Now in |
|---|---|
| Agents, MCP servers, rules | The Power Tools panel, opened from the message input |
| Backup, storage usage, reset | Data & Storage |
| Voice assistant toggle and global shortcut | Assistant |
| Global memory | Personalization |
Notification sounds still play when the assistant finishes a response, but the toggle for them is not exposed in the current Settings window.
Logout
The Logout button sits at the bottom of the Settings navigation column, below the section list. You are asked to confirm, then returned to the login screen with local session data cleared.