puter/paseo
1
0
Fork 0
mirror of https://github.com/getpaseo/paseo.git synced 2026-11-06 08:24:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

test(voice): clarify playback and timeline regression journeys

Browse Source
This commit is contained in:
Mohamed Boudra
2026-09-16 00:15:52 +02:00
parent 072c23fd48
commit 71a51ba6da
4 changed files with 78 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/server/src/server/session/voice/voice-session.test.ts
+12 -6
View File
@@ -125,6 +125,10 @@ function isAudioOutput(message: SessionOutboundMessage): boolean {
return message.type === "audio_output";
}
async function waitForAudioOutput(host: FakeVoiceHost): Promise<void> {
await vi.waitFor(() => expect(host.emitted.filter(isAudioOutput)).toHaveLength(1));
}
async function settle(): Promise<void> {
await Promise.resolve();
await Promise.resolve();
@@ -140,13 +144,14 @@ describe("VoiceSession streaming transcription", () => {
};
const { voiceSession, detector, host, speak } = createVoiceSession(tts);
const interrupted: string[] = [];
host.interruptAgentIfRunning = async (agentId) => {
async function recordInterruption(agentId: string) {
interrupted.push(agentId);
};
}
host.interruptAgentIfRunning = recordInterruption;
await voiceSession.handleSetVoiceMode(true, VOICE_AGENT_ID);
const playback = speak({ text: "A spoken response." });
try {
await vi.waitFor(() => expect(host.emitted.filter(isAudioOutput)).toHaveLength(1));
await waitForAudioOutput(host);
detector.emit("speech_started");
await settle();
expect(host.emitted).toContainEqual({
@@ -171,18 +176,19 @@ describe("VoiceSession streaming transcription", () => {
const external = new AbortController();
let abortRequested = false;
const emit = host.emit;
host.emit = (message) => {
function acknowledgeLaterAudio(message: SessionOutboundMessage) {
emit(message);
if (message.type === "audio_output" && abortRequested)
voiceSession.handleAudioPlayed(message.payload.id);
};
}
host.emit = acknowledgeLaterAudio;
await voiceSession.handleSetVoiceMode(true, VOICE_AGENT_ID);
const playback = speak({
text: "First sentence. Second sentence. Third sentence.",
signal: external.signal,
});
try {
await vi.waitFor(() => expect(host.emitted.filter(isAudioOutput)).toHaveLength(1));
await waitForAudioOutput(host);
abortRequested = true;
await voiceSession.handleAbort();
await playback;
packages/server/src/server/speech/providers/local/sherpa/sherpa-tts.test.ts
+21 -15
View File
@@ -46,6 +46,18 @@ function createTts(numThreads?: number, beforeGenerate?: () => Promise<void>) {
return { tts, requests, nativeConfig, released };
}
async function yieldNativeQueue(): Promise<void> {
await new Promise<void>((resolve) => setImmediate(resolve));
}
function holdNativeSynthesis() {
const pending: Array<() => void> = [];
function wait() {
return new Promise<void>((resolve) => pending.push(resolve));
}
return { pending, wait };
}
describe("SherpaOnnxTTS", () => {
it("requests Electron-compatible copied samples and returns PCM audio", async () => {
const { tts, requests } = createTts();
@@ -58,21 +70,18 @@ describe("SherpaOnnxTTS", () => {
});
it("keeps the event loop available and serializes native synthesis", async () => {
const pending: Array<() => void> = [];
const { tts, requests } = createTts(
undefined,
() => new Promise<void>((resolve) => pending.push(resolve)),
);
const { pending, wait } = holdNativeSynthesis();
const { tts, requests } = createTts(undefined, wait);
const first = tts.synthesizeSpeech("first");
const second = tts.synthesizeSpeech("second");
try {
await new Promise<void>((resolve) => setImmediate(resolve));
await yieldNativeQueue();
expect(requests).toEqual([]);
expect(pending).toHaveLength(1);
pending.shift()!();
const audio = await first;
audio.stream.destroy();
await new Promise<void>((resolve) => setImmediate(resolve));
await yieldNativeQueue();
expect(requests).toHaveLength(1);
expect(pending).toHaveLength(1);
pending.shift()!();
@@ -103,21 +112,18 @@ describe("SherpaOnnxTTS", () => {
});
it("waits for active native synthesis before freeing and drops queued synthesis", async () => {
let finish!: () => void;
const gate = new Promise<void>((resolve) => {
finish = resolve;
});
const { tts, requests, released } = createTts(undefined, () => gate);
const { pending, wait } = holdNativeSynthesis();
const { tts, requests, released } = createTts(undefined, wait);
const first = tts.synthesizeSpeech("first");
const second = tts.synthesizeSpeech("second");
const rejected = expect(second).rejects.toThrow("TTS provider closed");
await new Promise<void>((resolve) => setImmediate(resolve));
await yieldNativeQueue();
tts.free();
expect(released).toEqual([]);
finish();
pending.shift()!();
(await first).stream.destroy();
await rejected;
await new Promise<void>((resolve) => setImmediate(resolve));
await yieldNativeQueue();
expect(requests).toHaveLength(1);
expect(released).toEqual([true]);
await expect(tts.synthesizeSpeech("third")).rejects.toThrow("TTS provider closed");