pi-verbosity-control

Apply per-model OpenAI text.verbosity overrides and cycle the current model's setting from the keyboard.

Install

pi install npm:pi-verbosity-control

Or via git:

pi install git:github.com/ferologics/pi-verbosity-control

Restart Pi or use /reload if you are developing locally.

What it does

• Reads global config from ~/.pi/agent/verbosity.json
• Supports model-specific overrides by bare model id (gpt-5.4) or exact provider/model (openai-codex/gpt-5.4)
• Applies text.verbosity to supported OpenAI Responses-family requests right before they are sent
• Cycles the current model's verbosity with a shortcut and saves it back to the config file
• Optionally shows the active verbosity inline in Pi's footer

Exact provider/model entries win over bare model ids.

Supported APIs

• openai-responses
• openai-codex-responses
• azure-openai-responses

Anthropic / Claude

This extension is intentionally OpenAI-only.

Anthropic does not currently expose a direct equivalent to OpenAI text.verbosity. Claude's output_config.effort is a separate effort/thoroughness control, not a drop-in verbosity setting, so this extension does not map verbosity onto Anthropic models.

Shortcuts

• Cycle verbosity
  • macOS: Alt+V
  • Other platforms: Ctrl+Alt+V
• Toggle footer indicator
  • macOS: Alt+Shift+V
  • Other platforms: Ctrl+Alt+Shift+V

The verbosity cycle is:

low -> medium -> high -> low

Config

Path:

~/.pi/agent/verbosity.json

Example:

{
    "showIndicator": false,
    "models": {
        "gpt-5.4": "low",
        "openai/gpt-5.4": "medium"
    }
}

• showIndicator defaults to false
• When showIndicator is false, the extension does not patch Pi's footer at all

If you edit the file manually while Pi is already running, use /reload.

Notes

• The optional footer indicator uses a runtime monkeypatch of Pi's built-in FooterComponent, not a public footer-composition API.