@siddr/pi-openai-params

Combined OpenAI fast mode and verbosity settings for pi

Package details

extension

Install @siddr/pi-openai-params from npm and Pi will load the resources declared by the package manifest.

$ pi install npm:@siddr/pi-openai-params

Pi manifest JSON

{
  "extensions": [
    "./index.ts"
  ]
}

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

Combined OpenAI fast-mode and verbosity settings for pi.

This extension adds /openai-params, which opens a small settings screen for:

Use the list to:

When enabled, it patches provider requests right before send:

Fast mode is only applied to configured supported models.
Default supported models are:
- openai/gpt-5.4
- openai-codex/gpt-5.4
Verbosity is applied only to OpenAI Responses-family APIs:
- openai-responses
- openai-codex-responses
- azure-openai-responses
The default verbosity setting is unset, so the extension does not send any text.verbosity field unless you choose one.

Config uses the same project-over-global pattern as the fast-mode package:

If neither file exists, the extension creates the global file on first run.

Default config:

{
  "fast": false,
  "verbosity": null,
  "supportedModels": [
    "openai/gpt-5.4",
    "openai-codex/gpt-5.4"
  ]
}

This extension emits its current state on pi's extension event bus over pi:openai-params with:

That lets other extensions, including status, show the active non-default fast/verbosity settings for the current workspace.

This extension combines the behavior of:

based on OpenAI GPT-5 / Responses API parameter docs.