pi-model-selector-x

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# pi-model-selector-x

ModelSelectorX enhances pi's native `/model` selector with a detail pane showing context window, cost, input modalities, API protocol, and reasoning capability.

It patches the built-in `/model` view.

## WARNING

This extension patches the `/model` internals and could break if pi updates in an incompatible way. I chose this approach rather than re-implementing the internal model selector rendering so that it would automatically update the rendering.

## Features

- Bottom detail pane for the selected model showing full metadata
- Context window size (e.g. `200k`, `1M`)
- Max output tokens
- API protocol (`resp` / `comp` / `anth`)
- Input modalities (`txt`, `txt+img`, `txt+img+aud`)
- Reasoning capability indicator
- Cost breakdown (input / output / cache read / cache write)
- Base URL for the selected model/provider
- Free model detection

## Screenshots

<table>
  <tr>
    <th width="50%">Before</th>
    <th width="50%">After</th>
  </tr>
  <tr>
    <td width="50%"><img src="screenshots/before.svg" alt="Built-in /model selector showing only model-id and provider" width="100%" /></td>
    <td width="50%"><img src="screenshots/after.svg" alt="Enhanced /model selector with detail pane showing context, cost, protocol, input, and reasoning" width="100%" /></td>
  </tr>
  <tr>
    <td>The built-in selector shows only <code>model-id [provider] ✓</code>. No context window, cost, or capability info.</td>
    <td>ModelSelectorX adds a detail pane with context window, max output, API protocol, input modalities, reasoning indicator, and cost breakdown.</td>
  </tr>
</table>

## Detail Pane

The detail pane appears below the model list and shows:

| Field | Source | Example |
|-------|--------|---------|
| Context | `model.contextWindow` | `200k` |
| Max Output | `model.maxTokens` | `64k` |
| Protocol | `model.api` | `resp` / `comp` / `anth` |
| Input | `model.input` | `txt+img` / `txt` |
| Reasoning | `model.reasoning` | `⚡ reasoning` |
| Cost | `model.cost` | `$2.50 / $15.00` / `free` |
| Cache | `model.cost.cacheRead/Write` | `$0.25` |
| BaseURL | `model.baseUrl` | `https://api.example.com/v1` |

### Protocol abbreviations

| Short | Full |
|-------|------|
| `resp` | OpenAI Responses API |
| `comp` | OpenAI Completions API |
| `anth` | Anthropic Messages API |

## Installation

### npm

```bash
pi install npm:pi-model-selector-x
```

To try it for one run without adding it to your settings:

```bash
pi -e npm:pi-model-selector-x
```

### git

```bash
pi install git:github.com/Dwsy/pi-model-selector-x
```

## Usage

After installation, open the model selector with the built-in keybinding or:

```text
/model
```

## Notes

- Tested with pi 0.74.0
- ModelSelectorX patches the native `/model` path, so the built-in slash command and model hotkey keep using pi's own navigation and selection flow.
- ModelSelectorX relies on private model-selector internals, so upstream pi changes may require ModelSelectorX updates.

pi-model-selector-x

ModelSelectorX enhances pi's native /model selector with a detail pane showing context window, cost, input modalities, API protocol, and reasoning capability.

It patches the built-in /model view.

WARNING

This extension patches the /model internals and could break if pi updates in an incompatible way. I chose this approach rather than re-implementing the internal model selector rendering so that it would automatically update the rendering.

Features

Bottom detail pane for the selected model showing full metadata
Context window size (e.g. 200k, 1M)
Max output tokens
API protocol (resp / comp / anth)
Input modalities (txt, txt+img, txt+img+aud)
Reasoning capability indicator
Cost breakdown (input / output / cache read / cache write)
Base URL for the selected model/provider
Free model detection

Screenshots

Before	After

The built-in selector shows only `model-id [provider] ✓`. No context window, cost, or capability info.	ModelSelectorX adds a detail pane with context window, max output, API protocol, input modalities, reasoning indicator, and cost breakdown.

Detail Pane

The detail pane appears below the model list and shows:

Field	Source	Example
Context	`model.contextWindow`	`200k`
Max Output	`model.maxTokens`	`64k`
Protocol	`model.api`	`resp` / `comp` / `anth`
Input	`model.input`	`txt+img` / `txt`
Reasoning	`model.reasoning`	`⚡ reasoning`
Cost	`model.cost`	`$2.50 / $15.00` / `free`
Cache	`model.cost.cacheRead/Write`	`$0.25`
BaseURL	`model.baseUrl`	`https://api.example.com/v1`

Protocol abbreviations

Short	Full
`resp`	OpenAI Responses API
`comp`	OpenAI Completions API
`anth`	Anthropic Messages API

Installation

npm

pi install npm:pi-model-selector-x

To try it for one run without adding it to your settings:

pi -e npm:pi-model-selector-x

git

pi install git:github.com/Dwsy/pi-model-selector-x

Usage

After installation, open the model selector with the built-in keybinding or:

/model

Notes

Tested with pi 0.74.0
ModelSelectorX patches the native /model path, so the built-in slash command and model hotkey keep using pi's own navigation and selection flow.
ModelSelectorX relies on private model-selector internals, so upstream pi changes may require ModelSelectorX updates.