@thebinaryguy/pi-fast-mode

A Pi extension that adds a persistent, toggleable Fast mode for supported OpenAI Codex models.

When Fast mode is enabled, matching requests include:

{
  "service_tier": "priority"
}

Install

pi install npm:@thebinaryguy/pi-fast-mode

Or install directly from GitHub:

pi install git:github.com/TheBinaryGuy/pi-fast-mode

Restart Pi after installing, or run /reload in an existing session.

Tested with Pi 1.0.0. Requires Node.js 22.19.0 or newer.

Usage

/fast          Toggle Fast mode
/fast on       Enable Fast mode
/fast off      Disable Fast mode
/fast status   Show the current state

When enabled, Pi adds fast to the built-in model status:

gpt-5.6-sol • high • fast

The custom footer is active only while Fast mode is enabled. Pi supports one custom footer at a time, so another extension that installs a custom footer may override this indicator.

Persistence

The extension stores its global state in:

~/.pi/agent/openai-codex-fast-mode.json

The file is created automatically the first time Fast mode is toggled or explicitly enabled/disabled. New Pi processes load the saved state. If PI_CODING_AGENT_DIR is set, the file is stored in that directory instead.

Scope

Fast mode is applied only when all of these conditions are true:

• Provider is openai-codex
• API is openai-codex-responses
• Model is GPT-5.5 or a GPT-5.6 variant
• The outgoing payload does not already define service_tier

Priority processing may have higher usage costs. The extension changes only the outgoing request payload; depending on the service-tier value echoed by Codex, Pi's locally displayed cost may not reflect priority pricing.

Security

Pi extensions execute with your user permissions. This extension reads and writes only its state file in Pi's agent directory and modifies matching OpenAI Codex request payloads.

License

MIT