pi-gpt-config

wezterm-gui_ZdkfgMfkOP.mp4

Adds a /gpt_config command to pi for Codex-parity settings on two models only:

• gpt-5.3-codex
• gpt-5.4

Outside those two models, every setting in this extension is a no-op.

Install

pi install git:github.com/edxeth/pi-gpt-config

Controls

• Fast mode: on, off
• Output style: codex, claude
• Personality: default, friendly, pragmatic
• Verbosity: inherit, low, medium, high
• Reasoning summary: inherit, none, auto, concise, detailed

Behavior

• Fast mode requests OpenAI priority service tier.
• Output style changes answer framing:
  • codex: no extra overlay
  • claude: adds a Claude-inspired overlay for more answer-first, concise, lower-overengineering, lower-check-in behavior
• Personality changes tone:
  • default: model's built-in Codex tone with no extra personality overlay
  • friendly: warmer, more collaborative
  • pragmatic: more direct, factual, compact
• Verbosity controls answer length.
• Reasoning summary controls whether a summarized reasoning trace is requested back from the API. It does not disable internal reasoning.

Defaults on parity models

• Fast mode: off
• Output style: codex
• Personality: default
• Verbosity: inherit → effective default low
• Reasoning summary: inherit → effective default none

Command

Open the panel:

/gpt_config

Subcommands:

/gpt_config status
/gpt_config reset
/gpt_config fast on
/gpt_config fast off
/gpt_config style codex
/gpt_config style claude
/gpt_config personality default
/gpt_config personality friendly
/gpt_config personality pragmatic
/gpt_config verbosity inherit
/gpt_config verbosity low
/gpt_config verbosity medium
/gpt_config verbosity high
/gpt_config summary inherit
/gpt_config summary none
/gpt_config summary auto
/gpt_config summary concise
/gpt_config summary detailed

/gpt_config personality none is accepted as a backward-compatible alias for default.

Persistence

State is stored globally at:

~/.pi/agent/cache/pi-gpt-config/state.json

If PI_CODING_AGENT_DIR is set:

$PI_CODING_AGENT_DIR/cache/pi-gpt-config/state.json

Persisted key order matches the TUI:

{
  "fastMode": false,
  "style": "codex",
  "personality": "none",
  "verbosity": "inherit",
  "summary": "inherit"
}

TUI

• The panel explains what each setting does.
• The footer only shows:
  • priority
  • style
• The footer is only shown on the two parity models.

Notes

• This extension mixes native API controls (service_tier, text.verbosity, reasoning.summary) with prompt overlays (codex personality blocks and claude output style).
• personality=default / persisted "none" now means no extra personality overlay; only explicit friendly and pragmatic inject tone prompts.
• The claude output style is a Claude-inspired prompt overlay, not a provider-native Claude mode or true Claude parity.
• The Claude overlay also counter-pressures Codex-style optional check-ins by preferring autonomous continuation until a real blocker or decision appears.
• Backend support still depends on the upstream provider honoring the request fields.