Which providers should the model selector support?

Select all that you want available in the dashboard. You can always add more later.

A

NVIDIA NIM

Already configured. Nemotron, Llama, Mistral via NVIDIA's hosted API. Free tier available.

B

OpenRouter

Huge catalog (200+ models). Many free models (Llama, Gemma, Qwen, etc.). Single API key for everything. Has a /models endpoint for dynamic listing.

C

Rook Gateway

The local agent with tools, skills, and memory. Uses its own system prompt (30K+ tokens). Best for agent tasks, not basic chat. Currently has issues with NIM's think blocks.

D

Custom OpenAI-Compatible

Any endpoint that speaks OpenAI API format — Ollama, LM Studio, vLLM, text-generation-webui, etc. User enters base URL + optional API key.

Placement Question

Where should the model selector live in the UI?

1

Above Chat Input

Compact dropdown right above the message box. Always visible, easy to switch mid-conversation. Like ChatGPT's model picker.

2

In the Header Bar

Next to the "Online/Offline" indicator. Global, persistent, but takes header space.

3

Settings Page Only

Tucked away — set it and forget it. Less clutter but harder to switch on the fly.