Skip to main content

Goose

Goose is your local AI agent, automating engineering tasks seamlessly. Official site: block.github.io/goose

1.1 Usage​

Method 1: Built-in OpenAI Provider​

  1. Set the OpenAI-compatible endpoint and API key in environment variables:
export OPENAI_API_KEY="your-hpc-ai-api-key"
export OPENAI_HOST="https://api.hpc-ai.com/inference/v1"
export GOOSE_PROVIDER="openai"
export GOOSE_MODEL="minimax/minimax-m2.5"

  1. Restart goose and verify the configuration with goose info -v.

  2. If you want to persist settings, edit config.yaml:

    • macOS/Linux: ~/.config/goose/config.yaml
    • Windows: %APPDATA%\Block\goose\config\config.yaml

Method 2: Custom Provider​

Create a custom provider JSON file:

  • macOS/Linux: ~/.config/goose/custom_providers/
  • Windows: %APPDATA%\Block\goose\config\custom_providers\

Example:

{
  "name": "hpcai",
  "engine": "openai",
  "display_name": "HPC-AI",
  "api_key_env": "HPC_AI_API_KEY",
  "base_url": "https://api.hpc-ai.com/inference/v1",
  "models": [
    {
      "name": "minimax/minimax-m2.5",
      "context_limit": 128000
    }
  ]
}

Then select the custom provider by name:

export HPC_AI_API_KEY="your-hpc-ai-api-key"
export GOOSE_PROVIDER="hpcai"
export GOOSE_MODEL="minimax/minimax-m2.5"

1.2 Notes​

  1. Custom providers can include custom headers in the CLI or by editing the provider file after creation.
  2. If you are using the built-in OpenAI provider with OPENAI_HOST, keep GOOSE_PROVIDER="openai". Use a value such as hpcai only after creating a custom provider with that exact name.

1.3 References​