Random Studios

vLLM Gateway

Use https://ollama.random-studios.net/v1 as your OpenAI-compatible base URL.

OpenCode

In OpenCode, type /connect and select ollama.random-studios.net.

Example provider config:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Random Studios Ollama",
      "options": {
        "baseURL": "https://ollama.random-studios.net/v1"
      },
      "models": {
        "qwen3-coder": {
          "name": "Qwen3.6 27B AWQ INT4 (vLLM default)"
        }
      }
    }
  }
}

qwen3-coder is the alias for cyankiwi/Qwen3.6-27B-AWQ-INT4, currently served on the box.

pi-coding-agent

Add this provider to ~/.pi/agent/models.json:

{
  "providers": {
    "remote-vllm": {
      "baseUrl": "https://ollama.random-studios.net/v1",
      "api": "openai-completions",
      "apiKey": "<your-token>",
      "compat": {
        "supportsDeveloperRole": false,
        "supportsReasoningEffort": false
      },
      "models": [
        {
          "id": "qwen3-coder",
          "name": "qwen3-coder",
          "reasoning": true,
          "contextWindow": 262144,
          "maxTokens": 32768
        }
      ]
    }
  }
}

Replace <your-token> with your bearer token from the gateway.

Auth

Requests to the API require a bearer token.

Authorization: Bearer <your-token>

The homepage is public. API endpoints stay protected.

Serving Notes

This endpoint is now backed by vLLM, not Ollama.
The gateway currently serves one shared model at a time.
Clients should continue using the same bearer-token auth and /v1 OpenAI-compatible API path.
Some Ollama-specific features such as model listing and per-user upstream headers are no longer part of the backend.

Available Models

qwen3-coder Qwen3.6 27B AWQ-INT4, aliased to cyankiwi/Qwen3.6-27B-AWQ-INT4