357 lines
14 KiB
JavaScript
357 lines
14 KiB
JavaScript
import { cleanup, fireEvent, render, screen, waitFor } from '@testing-library/svelte';
|
|
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
|
|
import { createMessageEnvelope } from '@vela/protocol';
|
|
import VoiceSessionShell from '../src/lib/VoiceSessionShell.svelte';
|
|
|
|
class MockWebSocket {
|
|
static CONNECTING = 0;
|
|
static OPEN = 1;
|
|
static CLOSING = 2;
|
|
static CLOSED = 3;
|
|
static instances = [];
|
|
|
|
constructor(url) {
|
|
this.url = url;
|
|
this.readyState = MockWebSocket.CONNECTING;
|
|
this.sent = [];
|
|
this.onopen = null;
|
|
this.onmessage = null;
|
|
this.onerror = null;
|
|
this.onclose = null;
|
|
MockWebSocket.instances.push(this);
|
|
}
|
|
|
|
send(message) {
|
|
this.sent.push(message);
|
|
}
|
|
|
|
open() {
|
|
this.readyState = MockWebSocket.OPEN;
|
|
this.onopen?.();
|
|
}
|
|
|
|
message(payload) {
|
|
this.onmessage?.({ data: JSON.stringify(payload) });
|
|
}
|
|
|
|
close(code = 1000, reason = 'client disconnect', wasClean = true) {
|
|
this.readyState = MockWebSocket.CLOSED;
|
|
this.onclose?.({ code, reason, wasClean });
|
|
}
|
|
|
|
static latest() {
|
|
return MockWebSocket.instances.at(-1);
|
|
}
|
|
|
|
static reset() {
|
|
MockWebSocket.instances = [];
|
|
}
|
|
}
|
|
|
|
function getByTestId(id) {
|
|
return screen.getByTestId(id);
|
|
}
|
|
|
|
describe('voice session shell', () => {
|
|
beforeEach(() => {
|
|
MockWebSocket.reset();
|
|
vi.stubGlobal('WebSocket', MockWebSocket);
|
|
});
|
|
|
|
afterEach(() => {
|
|
cleanup();
|
|
vi.unstubAllGlobals();
|
|
});
|
|
|
|
it('keeps mocked turn unavailable while disconnected and after disconnect', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
expect(getByTestId('connection-state').textContent).toBe('not connected');
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(true);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-123' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('connection-state').textContent).toBe('connected');
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(false);
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(false);
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('disconnect-button'));
|
|
socket.close(1000, 'client disconnect', true);
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('connection-state').textContent).toBe('disconnected');
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('session-id').textContent).toBe('session-123');
|
|
});
|
|
});
|
|
|
|
it('runs a placeholder mic-control cycle, streams mocked assistant text, and keeps mocked turn usable on the same socket', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('connection-state').textContent).toBe('connected');
|
|
});
|
|
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(true);
|
|
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-mic' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(false);
|
|
});
|
|
|
|
await fireEvent.mouseDown(getByTestId('mic-control-button'));
|
|
|
|
expect(socket.sent).toHaveLength(1);
|
|
expect(JSON.parse(socket.sent[0])).toEqual({
|
|
type: 'input_audio.append',
|
|
payload: { chunk: 'placeholder-control-shell-chunk' }
|
|
});
|
|
expect(getByTestId('mic-control-status').textContent).toBe('holding');
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'listening' }));
|
|
socket.message(
|
|
createMessageEnvelope('transcript.partial', {
|
|
text: '[mocked partial] Placeholder push-to-talk transcript in progress.'
|
|
})
|
|
);
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('gateway-session-state').textContent).toBe('listening');
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('user-transcript').textContent).toBe(
|
|
'[mocked partial] Placeholder push-to-talk transcript in progress.'
|
|
);
|
|
expect(getByTestId('partial-transcript').textContent).toBe(
|
|
'[mocked partial] Placeholder push-to-talk transcript in progress.'
|
|
);
|
|
});
|
|
|
|
await fireEvent.mouseUp(getByTestId('mic-control-button'));
|
|
|
|
expect(socket.sent).toHaveLength(2);
|
|
expect(JSON.parse(socket.sent[1])).toEqual({
|
|
type: 'input_audio.commit',
|
|
payload: {}
|
|
});
|
|
expect(getByTestId('mic-control-status').textContent).toBe('idle');
|
|
|
|
socket.message(
|
|
createMessageEnvelope('transcript.final', {
|
|
text: '[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.'
|
|
})
|
|
);
|
|
socket.message(createMessageEnvelope('session.state', { value: 'thinking' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'speaking' }));
|
|
socket.message(createMessageEnvelope('response.text.delta', { text: '[mocked assistant] ' }));
|
|
socket.message(
|
|
createMessageEnvelope('response.text.delta', {
|
|
text: 'This is a deterministic mocked response from the gateway vertical slice.'
|
|
})
|
|
);
|
|
socket.message(createMessageEnvelope('response.completed', {}));
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('gateway-session-state').textContent).toBe('idle');
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(false);
|
|
expect(getByTestId('user-transcript').textContent).toBe(
|
|
'[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.'
|
|
);
|
|
expect(getByTestId('assistant-response').textContent).toBe(
|
|
'[mocked assistant] This is a deterministic mocked response from the gateway vertical slice.'
|
|
);
|
|
expect(getByTestId('partial-transcript').textContent).toBe('none');
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('mocked-turn-button'));
|
|
|
|
expect(socket.sent).toHaveLength(3);
|
|
expect(JSON.parse(socket.sent[2]).type).toBe('mocked.turn.trigger');
|
|
});
|
|
|
|
it('keeps rendered push-to-talk transcript and assistant text visible after cancel', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-ptt-cancel' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(false);
|
|
});
|
|
|
|
await fireEvent.mouseDown(getByTestId('mic-control-button'));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'listening' }));
|
|
socket.message(
|
|
createMessageEnvelope('transcript.partial', {
|
|
text: '[mocked partial] Placeholder push-to-talk transcript in progress.'
|
|
})
|
|
);
|
|
|
|
await fireEvent.mouseUp(getByTestId('mic-control-button'));
|
|
expect(JSON.parse(socket.sent[1]).type).toBe('input_audio.commit');
|
|
|
|
socket.message(
|
|
createMessageEnvelope('transcript.final', {
|
|
text: '[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.'
|
|
})
|
|
);
|
|
socket.message(createMessageEnvelope('session.state', { value: 'thinking' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'speaking' }));
|
|
socket.message(createMessageEnvelope('response.text.delta', { text: '[mocked assistant] ' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('cancel-turn-button').hasAttribute('disabled')).toBe(false);
|
|
expect(getByTestId('assistant-response').textContent).toBe('[mocked assistant] ');
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('cancel-turn-button'));
|
|
expect(JSON.parse(socket.sent[2]).type).toBe('response.cancel');
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('gateway-session-state').textContent).toBe('idle');
|
|
expect(getByTestId('mocked-turn-status').textContent).toBe('idle');
|
|
expect(getByTestId('cancel-turn-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('user-transcript').textContent).toBe(
|
|
'[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.'
|
|
);
|
|
expect(getByTestId('assistant-response').textContent).toBe('[mocked assistant] ');
|
|
expect(getByTestId('mic-control-button').hasAttribute('disabled')).toBe(false);
|
|
});
|
|
});
|
|
|
|
it('renders mocked transcript before assistant response for a connected session', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-456' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('connection-state').textContent).toBe('connected');
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('mocked-turn-button'));
|
|
|
|
expect(socket.sent).toHaveLength(1);
|
|
const sentMessage = JSON.parse(socket.sent[0]);
|
|
expect(sentMessage.type).toBe('mocked.turn.trigger');
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'listening' }));
|
|
socket.message(
|
|
createMessageEnvelope('transcript.partial', {
|
|
text: '[mocked partial] Placeholder push-to-talk transcript in progress.'
|
|
})
|
|
);
|
|
socket.message(createMessageEnvelope('transcript.final', { text: 'Turn on the office lamp.' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'thinking' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'speaking' }));
|
|
socket.message(createMessageEnvelope('response.text.delta', { text: 'Mocked ' }));
|
|
socket.message(createMessageEnvelope('response.text.delta', { text: 'assistant response.' }));
|
|
socket.message(createMessageEnvelope('response.completed', { reason: 'mocked_turn_complete' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('user-transcript').textContent).toBe('Turn on the office lamp.');
|
|
expect(getByTestId('partial-transcript').textContent).toBe('none');
|
|
expect(getByTestId('assistant-response').textContent).toBe('Mocked assistant response.');
|
|
expect(getByTestId('conversation-render-order').textContent).toBe('transcript>response');
|
|
expect(getByTestId('gateway-session-state').textContent).toBe('idle');
|
|
});
|
|
});
|
|
|
|
it('blocks mocked turn trigger before session.ready and allows it after session.ready', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('connection-state').textContent).toBe('connected');
|
|
});
|
|
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(true);
|
|
await fireEvent.click(getByTestId('mocked-turn-button'));
|
|
expect(socket.sent).toHaveLength(0);
|
|
expect(getByTestId('last-error').textContent).toBe('mocked turn requires session.ready');
|
|
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-789' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(false);
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('mocked-turn-button'));
|
|
|
|
expect(socket.sent).toHaveLength(1);
|
|
expect(JSON.parse(socket.sent[0]).type).toBe('mocked.turn.trigger');
|
|
});
|
|
|
|
it('shows cancel control during an active mocked turn and preserves rendered text after cancel', async () => {
|
|
render(VoiceSessionShell);
|
|
|
|
await fireEvent.click(getByTestId('connect-button'));
|
|
const socket = MockWebSocket.latest();
|
|
socket.open();
|
|
socket.message(createMessageEnvelope('session.ready', { sessionId: 'session-cancel' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('cancel-turn-button').hasAttribute('disabled')).toBe(true);
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('mocked-turn-button'));
|
|
|
|
expect(JSON.parse(socket.sent[0]).type).toBe('mocked.turn.trigger');
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'listening' }));
|
|
socket.message(createMessageEnvelope('transcript.final', { text: 'Keep this transcript.' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'thinking' }));
|
|
socket.message(createMessageEnvelope('session.state', { value: 'speaking' }));
|
|
socket.message(createMessageEnvelope('response.text.delta', { text: 'Partial response' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('cancel-turn-button').hasAttribute('disabled')).toBe(false);
|
|
expect(getByTestId('assistant-response').textContent).toBe('Partial response');
|
|
});
|
|
|
|
await fireEvent.click(getByTestId('cancel-turn-button'));
|
|
|
|
expect(JSON.parse(socket.sent[1]).type).toBe('response.cancel');
|
|
|
|
socket.message(createMessageEnvelope('session.state', { value: 'idle' }));
|
|
|
|
await waitFor(() => {
|
|
expect(getByTestId('mocked-turn-status').textContent).toBe('idle');
|
|
expect(getByTestId('gateway-session-state').textContent).toBe('idle');
|
|
expect(getByTestId('cancel-turn-button').hasAttribute('disabled')).toBe(true);
|
|
expect(getByTestId('mocked-turn-button').hasAttribute('disabled')).toBe(false);
|
|
expect(getByTestId('user-transcript').textContent).toBe('Keep this transcript.');
|
|
expect(getByTestId('assistant-response').textContent).toBe('Partial response');
|
|
});
|
|
});
|
|
});
|