Model Configuration
Provider Management
Settings > Models shows providers as cards.
Adding
- Click “Add” and select a template (Kimi/GLM/DeepSeek/Qwen/MiniMax/Doubao/OpenAI/Anthropic)
- Template auto-fills Base URL and model list (thinking levels, context window, multimodal)
- Paste API Key and click “Add”
Custom Providers
Manually enter name, Base URL, API type, API Key, and model list.
Fetch Model List
Pulls from /models endpoint with auto-filled metadata. All fields editable.
Model List
| Field | Description |
|---|---|
| ID | Model identifier for API calls |
| Display name | UI label |
| Thinking levels | off/minimal/low/medium/high/xhigh/max |
| Image input | Auto-verified with 1x1 PNG probe |
| Context length | Token count |
Image Recognition Fallback
When the current model doesn’t support image input, auto-switches to the configured fallback model.
Sub-Agent Tiers
Settings > Models > Sub-agent tiers: Simple/Standard/Hard, each mapping to a configured model or “Follow parent.”
Org-Synced Models
After Lansenger sign-in, org policies auto-sync:
- Managed providers and models (read-only, “Organization” group)
- Merged with user-configured models in the selector
- Cleared on sign-out
Adding a Custom Model
For services not covered by built-in provider templates (self-hosted gateways, OpenAI-compatible proxies), choose “Custom Provider”:
| Field | Notes |
|---|---|
| API type | OpenAI Chat Completions / Anthropic Messages—match the target service’s protocol |
| Base URL | Request endpoint; OpenAI-compatible services usually end at /v1 |
| API Key | Auth secret, stored locally and never uploaded |
| Model list | Add manually or fetch from the /models endpoint |
The model is selectable immediately after adding. If calls fail, first check the API key’s validity, whether the Base URL has extra path segments, and whether the model ID matches the provider’s.