Connecting models
Set API keys and default models under Settings → AI & Models.
What you will learn
- How to open Settings
- Paste keys under Provider API Keys
- Choose Default LLM / Fallback LLM
- Add custom OpenAI-compatible providers
- What to check when requests fail
1. Open Settings
Sidebar → Settings. Full page /settings stays open until you close it.
Top actions: Discard · Close · Save Changes
Real tab names
| Tab | Purpose |
|---|---|
| General | Language, skills directory, … |
| AI & Models | API keys, default/fallback model, temperature |
| Chat Interface | Chat UI / context options |
| System | System options |
| Advanced | Advanced (shell runtime PATH, …) |
| Experimental | Experimental features |
Names that do not exist: 「LLM Provider」 section, standalone 「AI Models」, 「Preferred Model」, Settings only as a top-right gear.
2. Provider API Keys
- Open AI & Models.
- Under Provider API Keys, pick a card (OpenAI, Anthropic, Google Gemini, Ollama, Groq, Fireworks AI, Cerebras, OpenRouter, …).
- Paste API Key (some cards also have Base URL).
- Click Save Changes.
Keys are stored locally. Never paste keys into chat.
Where to get keys
| Provider | Portal |
|---|---|
| Anthropic | console.anthropic.com |
| OpenAI | platform.openai.com/api-keys |
| Google Gemini | aistudio.google.com |
| Groq | console.groq.com/keys |
3. Model Preferences
Same AI & Models tab:
| Field | Meaning |
|---|---|
| Default LLM | Default provider/model for new sessions |
| Fallback LLM | Used when the default fails |
| Override temperature | When on, set Temperature yourself |
You can also change Provider / Model in Chat for the current session only.
4. Custom OpenAI providers
For OpenRouter, Ollama, etc., use Custom OpenAI Providers → Add Custom OpenAI Provider.
OpenRouter example:
| Field | Example |
|---|---|
| Display name | OpenRouter |
| Base URL | https://openrouter.ai/api/v1 |
| API Key | your OpenRouter key |
Ollama example:
| Field | Example |
|---|---|
| Base URL | http://localhost:11434/v1 |
| API Key | empty or ollama |
Local servers must already be running.
5. Troubleshooting
| Symptom | Check |
|---|---|
| Requests fail | Key / Base URL under Provider API Keys, and Save Changes |
| Empty model list | Save key, then Refresh models in the picker |
| Switch provider | Add key on that card, change Default LLM |
| Python/Node runtime | Not Settings — Chat → App Wizard + setup-wizard |
There is no Settings “connected / verifying” badge. Send a Chat message to verify.
More: Troubleshooting