@phoen1xcode/pi-codex-fast

A Pi Extension for OpenAI Codex Fast Mode.

Packages

Package details

extension

Install @phoen1xcode/pi-codex-fast from npm and Pi will load the resources declared by the package manifest.

$ pi install npm:@phoen1xcode/pi-codex-fast
Package
@phoen1xcode/pi-codex-fast
Version
0.1.1
Published
Aug 18, 2026
Downloads
247/mo · 13/wk
Author
phoen1xcode
License
MIT
Types
extension
Size
19.6 KB
Dependencies
1 dependency · 2 peers
Pi manifest JSON
{
  "extensions": [
    "./src/index.ts"
  ]
}

Security note

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

README

pi-codex-fast

A global Fast mode toggle for the Pi Coding Agent's ChatGPT/Codex subscription models.

Features

  • Sets serviceTier: "priority" on Pi's Codex stream only when a strict safety baseline is met
  • Never selects or modifies model, thinking level, tools, or prompts
  • Never reads or stores OAuth tokens
  • Persists state in Pi's existing ~/.pi/agent/settings.json — no separate config file
  • Fast mode is applied only when all of the following hold:
    • the global Fast mode toggle is enabled
    • the streamed model's provider is openai-codex
    • the streamed model's API is openai-codex-responses
    • the streamed model id is on the exact allowlist
    • modelRegistry.isUsingOAuth(model) is true

Install

pi install npm:@phoen1xcode/pi-codex-fast

Usage

  1. In Pi, run /login and choose OpenAI (ChatGPT Plus/Pro).
  2. Run /model and pick a supported model:
    • openai-codex/gpt-5.6-sol
    • openai-codex/gpt-5.6-terra
    • openai-codex/gpt-5.6-luna
  3. Toggle Fast mode:
/fast on
/fast off
/fast status

Configuration

State is stored in Pi's existing ~/.pi/agent/settings.json:

{
  "@phoen1xcode/pi-codex-fast": {
    "enabled": true
  }
}

The extension locates this file via Pi's getAgentDir(), so PI_CODING_AGENT_DIR is respected. Writes use a Pi-compatible file lock and atomic replacement, preserving other settings fields and the settings.json symlink.

"Global" here means persisted scope, not real-time sync across Pi processes. Each Pi instance reads the config at startup or on /reload; /fast takes effect immediately in the current instance, while other running instances need /reload to pick up the new value.

The extension passes serviceTier: "priority" through Pi's built-in Codex stream. The Codex backend may still report service_tier: "default" in responses; Pi estimates local cost based on the requested priority tier. This estimate is not equivalent to the final credits billing in the ChatGPT backend.

To add newly verified Fast-mode models, update SUPPORTED_MODEL_IDS in src/models.ts and add tests.

Development

npm install
npm run check
pi -e .