Files
gods-eye-view/src/voice/geminiAutomaticVad.test.mjs
T

792 lines
24 KiB
JavaScript

// Native automatic activity, reply ownership and expiry lifecycle regressions.
import test from 'node:test';
import assert from 'node:assert/strict';
import {
harness,
tick,
deferred,
} from '../testSupport/geminiNativeHarness.mjs';
const vad = (ws, type) => ws.message({ voiceActivity: { type } });
const onset = (ws) => {
vad(ws, 'ACTIVITY_START');
vad(ws, 'ACTIVITY_END');
};
const split = (ws) => {
onset(ws);
onset(ws);
};
const audio = {
serverContent: {
modelTurn: {
parts: [
{ inlineData: { data: 'AAAA', mimeType: 'audio/pcm;rate=24000' } },
],
},
},
};
const done = (ws) => ws.message({ serverContent: { turnComplete: true } });
const tools = (ws, id) =>
ws.message({
toolCall: {
functionCalls: [
{
id,
name: 'adjust_camera_zoom',
args: { direction: 'out', amount: 'little' },
},
],
},
});
const responses = (ws) =>
ws.sent.flatMap((m) => m.toolResponse?.functionResponses || []);
function close(h) {
h.c.stop();
assert.equal(h.timers.size, 0, 'owned timers released');
}
test('N1 VAD-only split admits the first real answer and completes after playback', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
split(ws);
ws.message(audio);
ws.message({ serverContent: { outputTranscription: { text: 'Answer' } } });
done(ws);
assert.equal(h.captures[0].played.length, 1);
h.captures[0].drain();
assert.equal(
h.events.findLast((e) => e.type === 'completion')?.status,
'completed',
);
assert.equal(h.c.retiredReplyPending, false);
} finally {
close(h);
}
});
test('N2 VAD-only split does not arm a nonexistent reply timeout', async () => {
const h = harness();
try {
await h.c.start();
split(h.sockets[0]);
h.advance(5001);
assert.equal(h.c.status, 'listening');
assert.equal(h.c.retiredReplyPending, false);
assert.equal(h.c.retiredBoundaryWait, null);
} finally {
close(h);
}
});
test('N3 server interruption before output still fences until ordered turnComplete', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
split(ws);
ws.message({ serverContent: { interrupted: true } });
assert.equal(h.c.retiredReplyPending, true);
ws.message(audio);
tools(ws, 'retired-before-boundary');
await h.c.toolQueue;
assert.equal(h.captures[0].played.length, 0);
assert.equal(h.calls.length, 0);
assert.equal(responses(ws).length, 0);
ws.message({
serverContent: { ...audio.serverContent, turnComplete: true },
});
assert.equal(
h.captures[0].played.length,
0,
'coalesced retired content is discarded',
);
ws.message(audio);
assert.equal(h.captures[0].played.length, 1);
done(ws);
h.captures[0].drain();
assert.equal(
h.events.findLast((e) => e.type === 'completion')?.status,
'completed',
);
} finally {
close(h);
}
});
test('N4 generationComplete is server evidence and next onset retires it', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message({ serverContent: { generationComplete: true } });
onset(ws);
assert.equal(h.c.retiredReplyPending, true);
assert.ok(h.c.retiredBoundaryWait);
ws.message(audio);
tools(ws, 'old-generation');
await h.c.toolQueue;
assert.equal(h.captures[0].played.length, 0);
assert.equal(h.calls.length, 0);
done(ws);
assert.equal(h.c.retiredReplyPending, false);
} finally {
close(h);
}
});
test('N5 goAway during VAD-only activity waits for actual reply and playback', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
split(ws);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
assert.ok(h.c.goAway);
assert.equal(ws.closed, undefined);
h.advance(5001);
assert.equal(h.c.isActive(), true);
ws.message(audio);
done(ws);
assert.equal(
h.c.isActive(),
true,
'pending actual PCM drains before expiry',
);
h.captures[0].drain();
assert.equal(h.c.status, 'error');
assert.equal(ws.closed, true);
} finally {
close(h);
}
});
test('N6 first server tool after VAD-only split executes exactly once', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
split(ws);
tools(ws, 'new-generation');
await h.c.toolQueue;
assert.equal(h.calls.length, 1);
assert.equal(
responses(ws).filter((x) => x.id === 'new-generation').length,
1,
);
assert.equal(responses(ws).at(-1).response.ok, true);
tools(ws, 'new-generation');
await h.c.toolQueue;
assert.equal(h.calls.length, 1);
assert.equal(
responses(ws).filter((x) => x.id === 'new-generation').length,
1,
);
} finally {
close(h);
}
});
test('C1 actual old audio followed by onset fences late audio and tool until boundary', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
ws.message(audio);
onset(ws);
assert.equal(h.c.retiredReplyPending, true);
ws.message(audio);
tools(ws, 'late-old');
await h.c.toolQueue;
assert.equal(h.captures[0].played.length, 1);
assert.equal(h.calls.length, 0);
ws.message({ serverContent: { interrupted: true } });
ws.message(audio);
assert.equal(h.captures[0].played.length, 1);
done(ws);
ws.message(audio);
assert.equal(h.captures[0].played.length, 2);
} finally {
close(h);
}
});
test('C2 first server tool in flight is aborted and never answered after new onset', async () => {
const gate = deferred(),
entered = [];
const h = harness({
runAction: async (_n, _a, o) => {
entered.push(o);
await gate.promise;
return { ok: true };
},
});
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
tools(ws, 'old-running');
await tick();
assert.equal(entered.length, 1);
onset(ws);
assert.equal(entered[0].signal.aborted, true);
gate.resolve();
await h.c.toolQueue;
assert.equal(responses(ws).filter((x) => x.id === 'old-running').length, 0);
assert.equal(h.c.retiredReplyPending, true);
} finally {
gate.resolve();
close(h);
}
});
test('C5 typed request before first server output still retires on next onset', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
assert.equal(h.c.sendTextCommand('Zoom in'), true);
onset(ws);
assert.equal(h.c.retiredReplyPending, true);
ws.message(audio);
tools(ws, 'late-typed');
await h.c.toolQueue;
assert.equal(h.captures[0].played.length, 0);
assert.equal(h.calls.length, 0);
} finally {
close(h);
}
});
test('N7 old retired boundary preserves the newer automatic owner for goAway', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
const oldOwner = h.c.automaticActivityOwner;
ws.message(audio);
onset(ws);
const successor = h.c.automaticActivityOwner;
assert.notEqual(successor, oldOwner);
assert.equal(h.c.retiredAutomaticActivityOwner, oldOwner);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(h.c.automaticActivityOwner, successor);
assert.equal(h.c.automaticTurnOpen, true);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
ws.message(audio);
done(ws);
h.captures[0].drain();
assert.equal(h.c.status, 'error');
} finally {
close(h);
}
});
test('N8 old automatic owner with no successor is cleared by its retired boundary', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
const owner = h.c.automaticActivityOwner;
ws.message(audio);
ws.message({ serverContent: { interrupted: true } });
assert.equal(h.c.retiredAutomaticActivityOwner, owner);
done(ws);
assert.equal(h.c.automaticTurnOpen, false);
assert.equal(h.c.automaticActivityOwner, null);
assert.equal(h.c.retiredAutomaticActivityOwner, null);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), false, 'no invented successor wait');
} finally {
close(h);
}
});
test('N9 coalesced old completion and new VAD preserves the new owner', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
vad(ws, 'ACTIVITY_END');
ws.message({
voiceActivity: { type: 'ACTIVITY_START' },
serverContent: { interrupted: true, turnComplete: true },
});
assert.equal(h.c.automaticTurnOpen, true);
assert.ok(h.c.automaticActivityOwner);
assert.equal(h.c.retiredAutomaticActivityOwner, null);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
} finally {
close(h);
}
});
test('N10 stop, restart and destroy release automatic owner identities and timers', async () => {
const h = harness();
try {
await h.c.start();
onset(h.sockets[0]);
const first = h.c.automaticActivityOwner;
h.c.stop();
assert.equal(h.c.automaticTurnOpen, false);
assert.equal(h.c.automaticActivityOwner, null);
assert.equal(h.c.retiredAutomaticActivityOwner, null);
assert.equal(h.c.automaticActivitySequence, 0);
await h.c.start();
onset(h.sockets[1]);
assert.notEqual(h.c.automaticActivityOwner, first);
assert.notEqual(h.c.automaticActivityOwner.epoch, first.epoch);
h.c.stop({ removeUi: true });
assert.equal(h.c.disposed, true);
assert.equal(h.c.automaticTurnOpen, false);
assert.equal(h.c.automaticActivityOwner, null);
assert.equal(h.c.retiredAutomaticActivityOwner, null);
assert.equal(h.timers.size, 0);
} finally {
close(h);
}
});
test('N11 deferred typed ownership counts discarded open-mic PCM without queuing or replay', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
ws.message(audio);
assert.equal(h.c.sendTextCommand('Zoom in'), true);
assert.equal(h.c.sendTextCommand('Zoom out'), true);
assert.equal(h.c.deferredInput.kind, 'user-text');
const before = ws.sent.filter((x) => x.realtimeInput?.audio).length;
const metadata = { captureId: h.c.captureOwner.captureId };
h.captures[0].callbacks.onAudio(
{ data: 'AAAA', mimeType: 'audio/pcm;rate=16000' },
metadata,
);
assert.equal(ws.sent.filter((x) => x.realtimeInput?.audio).length, before);
assert.equal(
h.c.deferredInput.messages.some((x) => x.realtimeInput?.audio),
false,
);
assert.equal(h.c.getDiagnostics().droppedAudioChunks, 1);
assert.deepEqual(h.c.getDiagnostics().deferredAudioDrops, {
nonPttOwner: 1,
closedActivity: 0,
});
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
done(ws);
assert.equal(
ws.sent.filter((x) => x.realtimeInput?.audio).length,
before,
'no wrong-owner PCM replay after boundaries',
);
} finally {
close(h);
}
});
test('N12 goAway before old completion still waits for the newer automatic owner', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
onset(ws);
const successor = h.c.automaticActivityOwner;
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(h.c.automaticActivityOwner, successor);
assert.equal(
h.c.isActive(),
true,
'old boundary during goAway cannot end successor wait',
);
ws.message(audio);
done(ws);
h.captures[0].drain();
assert.equal(h.c.status, 'error');
} finally {
close(h);
}
});
for (const interruptedBeforeExpiry of [false, true])
test(`N13 no successor clears retired ownership with goAway ${interruptedBeforeExpiry ? 'after interruption' : 'before interruption'}`, async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
if (interruptedBeforeExpiry)
ws.message({ serverContent: { interrupted: true } });
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(h.c.automaticTurnOpen, false);
assert.equal(h.c.automaticActivityOwner, null);
assert.equal(
h.c.isActive(),
false,
'no successor finishes at its actual boundary',
);
} finally {
close(h);
}
});
test('N14 native interruption during goAway aborts pending old tool and suppresses late response', async () => {
const gate = deferred(),
entered = [];
const h = harness({
runAction: async (_n, _a, o) => {
entered.push(o);
await gate.promise;
return { ok: true };
},
});
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
tools(ws, 'expiring-old-tool');
await tick();
assert.equal(entered.length, 1);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(entered[0].signal.aborted, true);
gate.resolve();
await h.c.toolQueue;
assert.equal(
responses(ws).filter((x) => x.id === 'expiring-old-tool').length,
0,
);
assert.equal(h.c.isActive(), false);
} finally {
gate.resolve();
close(h);
}
});
test('N15 goAway still refuses local PTT and typed replacement without reopening microphone', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
ws.message({ goAway: { timeLeft: '30s' } });
const turn = h.c.turn,
sent = ws.sent.length;
assert.equal(h.c.bargeIn(), false);
assert.equal(h.c.sendTextCommand('Zoom out'), false);
h.c.interrupt('push-to-talk');
assert.equal(h.c.turn, turn);
assert.equal(ws.sent.length, sent);
assert.equal(
h.tracks.every((t) => !t.enabled),
true,
);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(h.c.isActive(), false);
assert.equal(
h.tracks.every((t) => t.readyState === 'ended'),
true,
);
} finally {
close(h);
}
});
test('T-F1a PTT successor PCM survives retired boundary and a second input fences its late output', async () => {
const h = harness();
try {
await h.c.start({ pushToTalk: true });
const ws = h.sockets[0],
a = h.captures[0];
ws.message(audio);
h.c.input.pushToTalkKeyHeld = true;
assert.equal(h.c.input.bargeIn(), true);
h.c.input.setMicrophoneEnabled(true);
a.callbacks.onAudio(
{ data: 'AAAB', mimeType: 'audio/pcm;rate=16000' },
a.metadata,
);
h.c.input.setMicrophoneEnabled(false);
assert.equal(h.c.replyPending, true);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(
h.c.replyPending,
true,
'old boundary cannot clear submitted successor',
);
assert.equal(h.c.sendTextCommand('Zoom out second request'), true);
assert.ok(h.c.currentTypedInput());
const played = a.played.length;
ws.message(audio);
tools(ws, 'late-ptt-predecessor');
await h.c.toolQueue;
assert.equal(a.played.length, played);
assert.equal(
h.calls.length,
0,
'no old tool under newer local typed input',
);
done(ws);
tools(ws, 'current-typed');
await h.c.toolQueue;
assert.equal(h.calls.length, 1);
assert.equal(h.calls[0].options.userTurn, undefined);
assert.equal(h.calls[0].options.isCurrent(), true);
assert.equal(h.c.currentTypedInput().text, 'Zoom out second request');
} finally {
close(h);
}
});
for (const expiryBeforeBoundary of [false, true])
test(`T-F1b VAD then submitted text survives goAway ${expiryBeforeBoundary ? 'before' : 'after'} retired boundary`, async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
ws.message(audio);
onset(ws);
assert.equal(h.c.sendTextCommand('Zoom out successor'), true);
const typed = h.c.currentTypedInput();
assert.ok(typed);
assert.equal(h.c.replyPending, true);
if (expiryBeforeBoundary) ws.message({ goAway: { timeLeft: '30s' } });
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
assert.equal(h.c.replyPending, true);
if (!expiryBeforeBoundary) ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true, 'submitted text must finish');
tools(ws, 'valid-successor');
await h.c.toolQueue;
assert.equal(h.calls.length, 1);
assert.equal(h.calls[0].options.userTurn, undefined);
assert.equal(h.calls[0].options.isCurrent(), true);
assert.equal(h.c.currentTypedInput(), typed);
ws.message(audio);
done(ws);
assert.equal(h.c.isActive(), true);
h.captures[0].drain();
assert.equal(h.c.isActive(), false);
} finally {
close(h);
}
});
for (const repeat of [false, true])
test(`T-F2 forwarded A superseded by deferred B cannot run after goAway${repeat ? ' shortened twice' : ''}`, async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
ws.message(audio);
assert.equal(h.c.sendTextCommand('Zoom out A'), true);
assert.equal(h.c.replyPending, true);
assert.equal(h.c.sendTextCommand('Zoom out B'), true);
assert.equal(h.c.deferredInput.kind, 'user-text');
assert.ok(h.c.currentTypedInput());
onset(ws);
assert.equal(h.c.automaticTurnOpen, true);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.deferredInput, null);
if (repeat) {
ws.message({ goAway: { timeLeft: '20s' } });
ws.message({ goAway: { timeLeft: '10s' } });
}
assert.equal(h.c.goAway.supersededSuccessor, true);
assert.equal(h.c.isActive(), true);
ws.message({ serverContent: { interrupted: true, turnComplete: true } });
tools(ws, 'superseded-A');
await h.c.toolQueue;
assert.equal(h.calls.length, 0, 'A cannot run with dropped B authority');
assert.equal(
h.c.isActive(),
false,
'ordered boundary closes superseded forwarded predecessor',
);
} finally {
close(h);
}
});
test('T-F3 VAD-only no-generation expiry waits to bounded native deadline and releases resources', async () => {
const h = harness();
try {
await h.c.start();
onset(h.sockets[0]);
h.sockets[0].message({ goAway: { timeLeft: '30s' } });
h.advance(29999);
assert.equal(h.c.isActive(), true);
h.advance(1);
assert.equal(h.c.isActive(), false);
assert.equal(h.timers.size, 0);
assert.equal(
h.tracks.every((t) => t.readyState === 'ended'),
true,
);
} finally {
close(h);
}
});
test('T-C1 coalesced new START and natural old completion preserve newer automatic activity', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
ws.message({
voiceActivity: { type: 'ACTIVITY_START' },
serverContent: { turnComplete: true },
});
assert.equal(h.c.retiredReplyPending, false);
assert.equal(h.c.automaticTurnOpen, true);
ws.message({ goAway: { timeLeft: '30s' } });
assert.equal(h.c.isActive(), true);
} finally {
close(h);
}
});
test('T-C2 generationComplete alone followed by natural completion emits completion with no retirement timer', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message({ serverContent: { generationComplete: true } });
assert.equal(h.c.replyPending, true);
done(ws);
assert.equal(
h.events.findLast((e) => e.type === 'completion')?.status,
'completed',
);
assert.equal(h.c.replyPending, false);
assert.equal(h.c.automaticTurnOpen, false);
assert.equal(h.c.retiredBoundaryWait, null);
assert.equal(h.timers.size, 0);
} finally {
close(h);
}
});
test('T-C3 silent PTT terminal completion emits spoken false without fabricated audio', async () => {
const h = harness();
try {
await h.c.start({ pushToTalk: true });
const ws = h.sockets[0],
a = h.captures[0];
h.c.input.setMicrophoneEnabled(true);
a.callbacks.onAudio(
{ data: 'AAAB', mimeType: 'audio/pcm;rate=16000' },
a.metadata,
);
h.c.input.setMicrophoneEnabled(false);
assert.equal(h.c.userTurnPending, true);
ws.message({ serverContent: { generationComplete: true } });
done(ws);
const event = h.events.findLast((e) => e.type === 'completion');
assert.equal(event.status, 'completed');
assert.equal(event.spoken, false);
assert.equal(a.played.length, 0);
assert.equal(h.c.userTurnPending, false);
} finally {
close(h);
}
});
test('T-C4 silent terminal boundary waits for pending tool and native follow-up completion', async () => {
const gate = deferred(),
entered = [];
const h = harness({
runAction: async (_n, _a, o) => {
entered.push(o);
await gate.promise;
return { ok: true };
},
});
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
tools(ws, 'pending-terminal-tool');
await tick();
assert.equal(entered.length, 1);
done(ws);
assert.equal(
h.events.some((e) => e.type === 'completion'),
false,
);
assert.equal(h.c.userTurnPending, false);
gate.resolve();
await h.c.toolQueue;
h.c.finishPlayback();
assert.equal(
h.events.some((e) => e.type === 'completion'),
false,
);
assert.equal(h.c.replyPending, true, 'tool result awaits native follow-up');
done(ws);
assert.equal(
h.events.findLast((e) => e.type === 'completion')?.status,
'completed',
);
} finally {
gate.resolve();
close(h);
}
});
test('T-C5 terminal boundary with queued PCM waits for drain and preserves real spoken state', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
onset(ws);
ws.message(audio);
ws.message({ serverContent: { generationComplete: true } });
done(ws);
assert.equal(
h.events.some((e) => e.type === 'completion'),
false,
);
h.captures[0].drain();
const event = h.events.findLast((e) => e.type === 'completion');
assert.equal(event.status, 'completed');
assert.equal(event.spoken, true);
} finally {
close(h);
}
});
test('T-C6 old retired terminal boundary cannot clear successor userTurnPending', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
ws.message(audio);
assert.equal(h.c.sendTextCommand('Zoom out successor'), true);
assert.equal(h.c.userTurnPending, true);
done(ws);
assert.equal(h.c.userTurnPending, true);
assert.equal(
h.events.some((e) => e.type === 'completion'),
false,
);
} finally {
close(h);
}
});
test('T-C7 reentrant silent completion observer keeps new input pending', async () => {
const h = harness();
try {
await h.c.start();
const ws = h.sockets[0];
const emit = h.c.emit.bind(h.c);
h.c.emit = (event) => {
emit(event);
if (event.type === 'completion')
h.c.sendTextCommand('Zoom out reentrant');
};
onset(ws);
ws.message({ serverContent: { generationComplete: true } });
done(ws);
assert.equal(h.events.filter((e) => e.type === 'completion').length, 1);
assert.equal(h.c.replyPending, true);
assert.equal(h.c.userTurnPending, true);
assert.ok(h.c.currentTypedInput());
assert.equal(h.c.modelComplete, false);
} finally {
close(h);
}
});