pi-test

name	pi-test
description	Test pi-voice TUI, tts tool, and auto-TTS using pilotty for PTY-based terminal automation. Use when the user wants to test, verify, or validate the /voice command UI, the tts tool, auto-TTS event handling, or any pi-voice extension behavior. Also use when the user says "test the TUI", "run e2e tests", "test voice", "run pilotty", or asks about testing pi-voice. Do not use for server integration tests (npm test), linting, type checking, or server-only testing.
compatibility	Requires pilotty (npm install -g pilotty), a running TTS server with loaded model, and a working pi installation with a model.
allowed-tools	Bash read edit write

Pi-voice E2E Testing with Pilotty

Test pi-voice's extension stack (TUI, tts tool, auto-TTS) via PTY-based terminal automation.

Prerequisites

npm run server                              # start TTS server (default: 127.0.0.1:8181)
curl -sf http://127.0.0.1:8181/health      # verify model is loaded
which pilotty                               # verify pilotty is installed

All tests require the server running with a loaded model before spawning pi.

Test Suites

Script	What it tests	Run
tests/tui.sh	/voice command: navigate, toggle, cycle, reset, close	npm run test:tui
tests/tts-tool.sh	tts tool: agent invokes tool, WAV file created	npm run test:tts-tool
tests/auto-tts.sh	Auto-TTS: agent_end fires exactly 1 POST /tts	npm run test:auto-tts
tests/run.sh	All three suites	npm run test:e2e

Test Architecture

• tests/helpers.sh — shared utilities: assertions, pilotty wrappers, session lifecycle, server checks
• Each test: sources helpers → spawns pi → interacts → asserts → cleans up
• Config is backed up and restored around each test to avoid side effects
• auto-tts.sh uses a counting HTTP proxy to verify exact request counts

Core Workflow

Every test follows this pattern:

source "$(dirname "$0")/helpers.sh"    # load utilities
require_server                         # verify server + model
# ... backup config ...
spawn_pi                               # pilotty spawn --name ... --cwd "$PACKAGE_DIR"
wait_for_pi                            # wait for [Skills] indicator

HASH=$(snapshot_content_hash)
send_type "/voice"
send_key Enter
TEXT=$(await_change_and_snapshot_text "$HASH" 500 5000)

assert_match "description" "pattern" "$TEXT"
# ... interact ...
kill_session
# ... restore config ...
print_summary

Helper Functions (tests/helpers.sh)

Assertions

• assert_match DESC PATTERN TEXT — grep for pattern in text
• assert_no_match DESC PATTERN TEXT — assert pattern absent
• assert_equals DESC EXPECTED ACTUAL — exact string equality
• assert_not_empty DESC VALUE — value is non-empty

Pilotty wrappers

• snapshot_text() — snapshot as plain text (strips header)
• snapshot_json() — full JSON snapshot
• snapshot_content_hash() — hash for change detection
• wait_for_change HASH [SETTLE] [TIMEOUT] — block until screen changes
• await_change_and_snapshot_text HASH [SETTLE] [TIMEOUT] — wait + snapshot
• send_key KEY / send_type TEXT — keyboard input

Session lifecycle

• spawn_pi() — spawn pi with extension loaded in pilotty session
• wait_for_pi [TIMEOUT] — wait for [Skills] indicator
• kill_session() — kill pilotty session
• require_server() — verify server is running with model loaded (exits if not)

Server

• require_server() — checks $SERVER_URL/health, verifies modelLoaded === true

/voice TUI Key Mapping

Action	Key	pilotty command
Navigate settings	↑/↓	send_key Up / send_key Down
Change value	←/→	send_key Left / send_key Right
Play sample	Enter	send_key Enter
Reset to defaults	r	send_key r
Close	Escape	send_key Escape

TUI Assertions

The /voice TUI renders these elements for assertion:

• → TTS / → Voice / → Speed — cursor indicator on active row
• on / off — TTS toggle state
• af_heart / af_alloy — voice names with nationality/gender hints
• Speed 1.0 — speed value
• Server running — status indicator
• navigate / ←→ change / r reset — keybinding hints
• Playing sample… — temporary playback indicator

Adding New Tests

Create tests/new-feature.sh:

#!/bin/bash
source "$(dirname "$0")/helpers.sh"

echo -e "${BOLD}  pi-voice New Feature Test  ${RESET}"

require_server
# ... backup config if needed ...

spawn_pi
wait_for_pi

# ... test logic using helpers ...

kill_session
# ... restore config ...
print_summary

Then add to tests/run.sh's default TESTS array and add a script entry in package.json:

"test:new-feature": "bash tests/new-feature.sh"

Troubleshooting

Problem	Fix
Server not running	npm run server in a separate terminal
Model not loaded	curl -X POST http://127.0.0.1:8181/models/download -d '{"dtype":"q4"}'
Pi won't start	Increase timeout: wait_for_pi 30000
TUI doesn't appear	Longer settle: await_change_and_snapshot_text "$HASH" 1000 10000
Session not found	pilotty list-sessions — sessions cleaned up on exit
WAV file missing	speak() deletes after playback — expected for tts-tool tests
Proxy port conflict	lsof -ti:18181 | xargs kill -9 (auto-tts test)