@settinghead/pi-voxlert

SHODAN, the StarCraft Adjutant, and GLaDOS narrate your pi coding sessions. LLM-generated voice notifications spoken by game characters — know which agent needs you, by ear.

Package details

← Back
extension

Install @settinghead/pi-voxlert from npm and Pi will load the resources declared by the package manifest.

npm repohome report
$ pi install npm:@settinghead/pi-voxlert
Package
@settinghead/pi-voxlert
Version
0.2.0
Published
Mar 20, 2026
Downloads
24/mo · 7/wk
Author
settinghead
License
MIT
Types
extension
Size
13.8 KB
Dependencies
0 dependencies · 2 peers
Pi manifest JSON
{
  "extensions": [
    "./extensions"
  ],
  "video": "https://youtu.be/5xFXGijwJuk"
}

Security note

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

README

pi-voxlert

SHODAN, the StarCraft Adjutant, and GLaDOS narrate your pi coding sessions.

Voice notifications for pi powered by Voxlert. When your agent finishes a task or hits an error, you hear a contextual phrase in a game character's voice — instead of silence or a generic chime.

"Awaiting further orders, Commander. Build process complete."

Install

pi install npm:@settinghead/pi-voxlert

That's it. On first session, the extension detects that the Voxlert CLI is missing and offers to install + configure it automatically:

  • Installs @settinghead/voxlert globally
  • Downloads default voice packs (SHODAN, Adjutant, Kerrigan, etc.)
  • Auto-detects your TTS backend (Qwen3-TTS on Apple Silicon, Chatterbox on CUDA)

You can also trigger setup manually anytime with /voxlert setup, or run voxlert setup in a terminal for full interactive configuration.

What it does

pi event Voxlert action
Agent finishes (agent_end) Speaks a contextual in-character phrase
Tool error (tool_result with error) Announces the error in character

Phrases are generated per-event by an LLM, so you hear things like "Pathetic authentication corrected" (SHODAN) or "Warning, Commander. Test suite failure detected" (Adjutant) — not canned sounds.

Commands

Command Description
/voxlert setup Install CLI + configure with defaults
/voxlert test Fire a test voice notification
/voxlert status Check if Voxlert CLI is available
/voxlert Show help

The LLM can also call the voxlert_speak tool to say something aloud on demand.

Configuration

All voice pack, TTS backend, and LLM settings are managed through the Voxlert CLI:

voxlert config           # interactive configuration
voxlert packs            # list available voice packs
voxlert test "Hello"     # test your setup

Supports local TTS (Qwen3-TTS on Apple Silicon, Chatterbox on CUDA) and multiple LLM backends (OpenRouter, OpenAI, Anthropic, Gemini).

Requirements

  • pi coding agent
  • Voxlert CLI installed and configured (npm install -g @settinghead/voxlert && voxlert setup)
  • A TTS backend running (or Voxlert falls back to text notifications)

Links