feat(vela): start mocked response flow after push-to-talk commit
This commit is contained in:
@@ -67,21 +67,23 @@ function scheduleMockedTurnStep(session, turnId, delay, callback) {
|
||||
session.mockedTurnTimers.push(timer);
|
||||
}
|
||||
|
||||
function startMockedTurn(socket, session) {
|
||||
function startMockedTurn(socket, session, { transcript = MOCKED_USER_TRANSCRIPT, includeListeningState = true } = {}) {
|
||||
if (session.mockedTurnInFlight) {
|
||||
sendSocketError(socket, 'mocked_turn_in_flight', 'Only one mocked turn can run per session at a time.');
|
||||
return;
|
||||
}
|
||||
|
||||
clearMockedTurn(session);
|
||||
session.audioChunkCount = 0;
|
||||
session.mockedTurnInFlight = true;
|
||||
const turnId = crypto.randomUUID();
|
||||
session.activeMockedTurnId = turnId;
|
||||
updateSessionState(socket, session, 'listening');
|
||||
|
||||
if (includeListeningState) {
|
||||
updateSessionState(socket, session, 'listening');
|
||||
}
|
||||
|
||||
scheduleMockedTurnStep(session, turnId, 75, () => {
|
||||
sendSocketMessage(socket, 'transcript.final', { text: MOCKED_USER_TRANSCRIPT });
|
||||
sendSocketMessage(socket, 'transcript.final', { text: transcript });
|
||||
updateSessionState(socket, session, 'thinking');
|
||||
});
|
||||
|
||||
@@ -264,11 +266,12 @@ function handleClientMessage(socket, session, rawMessage) {
|
||||
break;
|
||||
}
|
||||
|
||||
sendSocketMessage(socket, 'transcript.final', {
|
||||
text: createPlaceholderFinalTranscript(session.audioChunkCount)
|
||||
});
|
||||
const finalTranscript = createPlaceholderFinalTranscript(session.audioChunkCount);
|
||||
session.audioChunkCount = 0;
|
||||
updateSessionState(socket, session, 'idle');
|
||||
startMockedTurn(socket, session, {
|
||||
transcript: finalTranscript,
|
||||
includeListeningState: false
|
||||
});
|
||||
break;
|
||||
case 'response.cancel':
|
||||
clearMockedTurn(session);
|
||||
|
||||
@@ -303,6 +303,26 @@ test('websocket handles valid and invalid client messages safely', async () => {
|
||||
type: 'transcript.final',
|
||||
payload: { text: '[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'thinking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'speaking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: '[mocked assistant] ' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: 'This is a deterministic mocked response from the gateway vertical slice.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.completed',
|
||||
payload: {}
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'idle' }
|
||||
@@ -358,6 +378,26 @@ test('websocket accepts a placeholder input cycle before a mocked turn on the sa
|
||||
type: 'transcript.final',
|
||||
payload: { text: '[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'thinking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'speaking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: '[mocked assistant] ' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: 'This is a deterministic mocked response from the gateway vertical slice.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.completed',
|
||||
payload: {}
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'idle' }
|
||||
@@ -404,6 +444,37 @@ test('websocket emits deterministic partials for repeated appends and a determin
|
||||
type: 'transcript.final',
|
||||
payload: { text: '[mocked final] Placeholder push-to-talk transcript completed from 2 appended chunks.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'thinking' }
|
||||
});
|
||||
|
||||
client.sendJson({ type: 'input_audio.commit', payload: {} });
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'error',
|
||||
payload: {
|
||||
code: 'mocked_turn_in_flight',
|
||||
message: 'Wait for the mocked turn to finish before committing input.',
|
||||
retryable: true
|
||||
}
|
||||
});
|
||||
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'speaking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: '[mocked assistant] ' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: 'This is a deterministic mocked response from the gateway vertical slice.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.completed',
|
||||
payload: {}
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'idle' }
|
||||
@@ -414,7 +485,30 @@ test('websocket emits deterministic partials for repeated appends and a determin
|
||||
type: 'transcript.final',
|
||||
payload: { text: '[mocked final] Placeholder push-to-talk transcript completed without appended audio.' }
|
||||
});
|
||||
await assert.rejects(() => client.nextMessage(150), /timed out waiting for websocket message/);
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'thinking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'speaking' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: '[mocked assistant] ' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.text.delta',
|
||||
payload: { text: 'This is a deterministic mocked response from the gateway vertical slice.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'response.completed',
|
||||
payload: {}
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'idle' }
|
||||
});
|
||||
|
||||
await client.close();
|
||||
} finally {
|
||||
@@ -542,6 +636,53 @@ test('websocket cancel stops an active mocked turn and allows a new one without
|
||||
}
|
||||
});
|
||||
|
||||
test('websocket cancel stops a push-to-talk commit response and allows another turn', async () => {
|
||||
const server = await startServer();
|
||||
|
||||
try {
|
||||
const client = await connectWebSocket(server.port);
|
||||
await client.nextMessage();
|
||||
await client.nextMessage();
|
||||
|
||||
client.sendJson({ type: 'input_audio.append', payload: { chunk: 'chunk-1' } });
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'listening' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'transcript.partial',
|
||||
payload: { text: '[mocked partial] Placeholder push-to-talk transcript in progress.' }
|
||||
});
|
||||
|
||||
client.sendJson({ type: 'input_audio.commit', payload: {} });
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'transcript.final',
|
||||
payload: { text: '[mocked final] Placeholder push-to-talk transcript completed from 1 appended chunk.' }
|
||||
});
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'thinking' }
|
||||
});
|
||||
|
||||
client.sendJson({ type: 'response.cancel', payload: {} });
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'idle' }
|
||||
});
|
||||
await assert.rejects(() => client.nextMessage(150), /timed out waiting for websocket message/);
|
||||
|
||||
client.sendJson({ type: 'mocked.turn.trigger', payload: {} });
|
||||
assert.deepEqual(await client.nextMessage(), {
|
||||
type: 'session.state',
|
||||
payload: { value: 'listening' }
|
||||
});
|
||||
|
||||
await client.close();
|
||||
} finally {
|
||||
await server.close();
|
||||
}
|
||||
});
|
||||
|
||||
test('websocket safely accepts cancel when no turn is active', async () => {
|
||||
const server = await startServer();
|
||||
|
||||
|
||||
Reference in New Issue
Block a user