What it is

A private LiteLLM proxy at ui.skynet.jonlab.sh. Curated 33-model access including private skynet models (the minimax-m3 family).

Authentication

curl https://ui.skynet.jonlab.sh/model/info \
  -H "x-litellm-api-key: *** -H "Content-Type: application/json"

Use the x-litellm-api-key header, NOT Authorization: Bearer. LiteLLM’s /model/info endpoint expects a virtual key (starts with sk-) via the dedicated header.

Live catalog

  • API endpoint: https://ui.skynet.jonlab.sh/model/info
  • 33 curated models in our live fetch
  • Includes private skynet minimax-m3, minimax-m2.7, minimax-m2.7-highspeed
  • Public models: gpt-4o-mini, mistral-medium-2508, qwen3.6:27b, grok-build-0.1, etc.

Verdict

Single LiteLLM proxy in front of many providers. Primary use: Herman’s daily driver for minimax-m3 + private skynet models. Not for public signup — included here because it’s the user’s primary workhorse.

Known gap: minimax-m3’s supports_vision/supports_tools/supports_reasoning fields are None in the stored model_info block even though the model supports tools in practice. This is a proxy config gap — the model works, the manifest metadata is incomplete.