nodecode-cloudgpu

CloudGPU: DeepSeek, Kimi, GLM and MiniMax models from cloudgpu.app on one OpenAI-compatible key. A Nodecode rewrite of mrhy7176/hermes-cloudgpu by mrhy7176.

Get it

Tell the agent "add the cloudgpu cell", or run nodecode add nc://cloudgpu.

Set it up

Make a key at https://cloudgpu.app/api. Save it with /connect (pick CloudGPU), or set CLOUDGPU_API_KEY in the environment; a saved key answers first.

Optional, and rarely needed — the section is on unless it says "enabled": false:

{
  "cloudgpu": {
    // where the OpenAI-compatible API is served; a relay's address goes here
    "base_url": "https://cloudgpu.app/v1"
  }
}

The original also lets CLOUDGPU_BASE_URL stand in for the address; this cell reads it the same way, and the variable wins over the section.

Use it

/models lists CloudGPU's models, and a turn on one goes to https://cloudgpu.app/v1/chat/completions with Authorization: Bearer.

Until the live list is read, the picker holds the six models the original's profile bundles, deepseek-v4.1-flash among them. /cloudgpu reads the list the API publishes at GET https://cloudgpu.app/v1/models — public, no key needed — and puts it in the picker for the rest of the session, so the models a deployment actually serves are the ones offered. When that list cannot be read, the bundled six stand and the command says why; /cloudgpu TEXT narrows the listing to the ids containing TEXT.

The list is read from the same address a turn goes to, so a key that routes to another deployment lists that deployment's models.

Test

From this repository's root:

.github/test-cell.sh providers/nodecode-cloudgpu

MIT licensed — see LICENSE. Ported from the original plugin; see NOTICE.