pi-scoped-models-extra-info

Interactive table of your pi coding agent's scoped models — pricing, context window, thinking levels, modalities, and optional coding benchmarks.

Features

• Rich model table — shows all your enabled (scoped) models in one view
• Columns: model slug, input price, output price, context window, input modalities, thinking levels, and coding benchmarks
• Sortable — press n (name), i (input price), o (output price), c (coding index)
• Model switching — press Enter on any row to switch to that model
• Keyboard navigation — ↑↓/jk to move, Home/End to jump, q/Esc to close
• Optional coding benchmarks — per-thinking-level coding index from Artificial Analysis (if AA_API_KEY is set)

Installation

Via git (pi package manager)

pi install git:github.com/kdejaeger/pi-scoped-models-extra-info

Then reload pi (/reload) or restart.

Via local clone

Clone the repo and add to ~/.pi/agent/settings.json:

{
  "extensions": [
    "~/pi-scoped-models-extra-info/index.ts"
  ]
}

Usage

Run the command:

/scoped-models-extra-info

Or use the keyboard shortcut: Ctrl+Alt+F

Navigation

Key	Action
↑ / k	Move selection up
↓ / j	Move selection down
Home / Ctrl+A	Jump to first model
End / Ctrl+E	Jump to last model
Enter / Space	Switch to selected model
n	Sort by model name
i	Sort by input price
o	Sort by output price
c	Sort by coding index
q / Esc	Close table

Configuration

Coding benchmarks (optional)

The extension can show Artificial Analysis coding index scores per thinking level. To enable this:

1. Get an API key from artificialanalysis.ai
2. Set the environment variable before starting pi:

export AA_API_KEY="aa_your_key_here"
pi

Or add to your ~/.bashrc / ~/.zshrc:

export AA_API_KEY="aa_your_key_here"

Without this variable, the extension works perfectly — it just omits the coding index column.

What it shows

The table displays only the models you have enabled/scoped in your pi settings.json (enabledModels). If you haven't scoped any models, it shows all available models.

Pricing comes from pi's built-in model registry — it's accurate for all providers pi supports (OpenAI, Anthropic, Google, OpenRouter, Groq, etc.).

Thinking levels are resolved from each model's capabilities, with smart fallback for OpenRouter-proxied models.

Package structure

pi-scoped-models-extra-info/
├── index.ts              # Extension source
├── package.json          # Pi package manifest
├── screenshot.png         # Table screenshot
├── LICENSE               # MIT
└── README.md             # This file