@kuralle-syrinx/realtime 4.2.0 → 4.4.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,82 +0,0 @@
1
- // SPDX-License-Identifier: MIT
2
-
3
- import { afterEach, describe, expect, it, vi } from "vitest";
4
-
5
- import { createGeminiTranslateSession } from "./gemini-translate.js";
6
-
7
- const sendRealtimeInput = vi.fn();
8
- const closeSession = vi.fn();
9
-
10
- const liveConnect = vi.fn().mockImplementation(async () => ({
11
- sendRealtimeInput,
12
- close: closeSession,
13
- }));
14
-
15
- vi.mock("@google/genai", () => ({
16
- GoogleGenAI: vi.fn().mockImplementation(() => ({
17
- live: { connect: liveConnect },
18
- })),
19
- Modality: { AUDIO: "AUDIO" },
20
- }));
21
-
22
- const FRAME_BYTES = 640;
23
-
24
- function frame(byte: number): Uint8Array {
25
- return new Uint8Array(FRAME_BYTES).fill(byte);
26
- }
27
-
28
- afterEach(() => {
29
- sendRealtimeInput.mockClear();
30
- closeSession.mockClear();
31
- liveConnect.mockClear();
32
- });
33
-
34
- describe("createGeminiTranslateSession", () => {
35
- it("coalesces 20ms frames into 100ms chunks before sendRealtimeInput", async () => {
36
- const session = await createGeminiTranslateSession({
37
- apiKey: "test-key",
38
- targetLanguageCode: "es",
39
- onAudio: () => {},
40
- });
41
-
42
- for (let i = 0; i < 5; i += 1) {
43
- session.sendAudio(frame(i));
44
- }
45
-
46
- expect(sendRealtimeInput).toHaveBeenCalledTimes(1);
47
- const arg = sendRealtimeInput.mock.calls[0]![0] as { audio: { data: string } };
48
- const chunk = Buffer.from(arg.audio.data, "base64");
49
- expect(chunk.byteLength).toBe(3200);
50
- expect(chunk[0]).toBe(0);
51
- expect(chunk[639]).toBe(0);
52
- expect(chunk[640]).toBe(1);
53
- expect(chunk[3199]).toBe(4);
54
- });
55
-
56
- it("flushes remainder on signalAudioStreamEnd and close", async () => {
57
- const session = await createGeminiTranslateSession({
58
- apiKey: "test-key",
59
- targetLanguageCode: "es",
60
- onAudio: () => {},
61
- });
62
-
63
- session.sendAudio(frame(9));
64
- session.sendAudio(frame(9));
65
- session.sendAudio(frame(9));
66
- expect(sendRealtimeInput).toHaveBeenCalledTimes(0);
67
-
68
- session.signalAudioStreamEnd();
69
- expect(sendRealtimeInput).toHaveBeenCalledTimes(2);
70
- const streamEndArg = sendRealtimeInput.mock.calls[0]![0] as { audio: { data: string } };
71
- expect(Buffer.from(streamEndArg.audio.data, "base64").byteLength).toBe(1920);
72
- expect(sendRealtimeInput.mock.calls[1]![0]).toEqual({ audioStreamEnd: true });
73
-
74
- sendRealtimeInput.mockClear();
75
- session.sendAudio(frame(7));
76
- await session.close();
77
- expect(sendRealtimeInput).toHaveBeenCalledTimes(1);
78
- const closeArg = sendRealtimeInput.mock.calls[0]![0] as { audio: { data: string } };
79
- expect(Buffer.from(closeArg.audio.data, "base64").byteLength).toBe(640);
80
- expect(closeSession).toHaveBeenCalledTimes(1);
81
- });
82
- });
@@ -1,188 +0,0 @@
1
- // SPDX-License-Identifier: MIT
2
-
3
- import { describe, expect, it } from "vitest";
4
-
5
- import type { ManagedSocket, SocketData, SocketFactory } from "@kuralle-syrinx/ws";
6
-
7
- import { createOpenAiCompatibleRealtimeAdapter } from "./openai-compatible-realtime.js";
8
- import type { RealtimeEvent } from "./realtime-adapter.js";
9
-
10
- interface MockSocketHarness {
11
- readonly factory: SocketFactory;
12
- readonly sent: string[];
13
- inject(msg: Record<string, unknown>): void;
14
- }
15
-
16
- function createMockSocketHarness(): MockSocketHarness {
17
- const sent: string[] = [];
18
- let messageHandler: ((data: SocketData, isBinary: boolean) => void) | null = null;
19
-
20
- const socket: ManagedSocket = {
21
- get isOpen() {
22
- return true;
23
- },
24
- send: (data: SocketData) => {
25
- sent.push(typeof data === "string" ? data : "");
26
- },
27
- keepAlivePing: () => {},
28
- verify: async () => true,
29
- dispose: () => {},
30
- onOpen: (handler) => {
31
- queueMicrotask(() => handler());
32
- },
33
- onMessage: (handler) => {
34
- messageHandler = handler;
35
- },
36
- onClose: () => {},
37
- onError: () => {},
38
- };
39
-
40
- return {
41
- factory: () => socket,
42
- sent,
43
- inject: (msg) => messageHandler?.(JSON.stringify(msg), false),
44
- };
45
- }
46
-
47
- async function collectEvents(
48
- events: AsyncIterable<RealtimeEvent>,
49
- max = 8,
50
- ): Promise<RealtimeEvent[]> {
51
- const out: RealtimeEvent[] = [];
52
- for await (const event of events) {
53
- out.push(event);
54
- if (out.length >= max) break;
55
- }
56
- return out;
57
- }
58
-
59
- async function waitFor(predicate: () => boolean, timeoutMs = 2000): Promise<void> {
60
- const startedAt = Date.now();
61
- while (Date.now() - startedAt < timeoutMs) {
62
- if (predicate()) return;
63
- await new Promise((resolve) => setTimeout(resolve, 5));
64
- }
65
- throw new Error("Timed out waiting for condition");
66
- }
67
-
68
- function createAdapter(mock: MockSocketHarness) {
69
- return createOpenAiCompatibleRealtimeAdapter({
70
- apiKey: "test-key",
71
- socketFactory: mock.factory,
72
- defaultModel: "gpt-realtime-2",
73
- url: () => "wss://example.test/realtime?model=gpt-realtime-2",
74
- caps: {
75
- inputSampleRateHz: 24_000,
76
- outputSampleRateHz: 24_000,
77
- supportsConcurrentToolAudio: true,
78
- supportsTruncate: true,
79
- emitsServerSpeechStarted: true,
80
- supportsTextOnlyModality: true,
81
- },
82
- buildSessionUpdate: () => ({
83
- type: "realtime",
84
- model: "gpt-realtime-2",
85
- output_modalities: ["text"],
86
- }),
87
- supportsTruncate: true,
88
- defaultErrorMessage: "OpenAI Realtime error",
89
- });
90
- }
91
-
92
- describe("createOpenAiCompatibleRealtimeAdapter text-only output", () => {
93
- it("surfaces response.output_text.delta as assistant transcript (final=false)", async () => {
94
- const mock = createMockSocketHarness();
95
- const adapter = createAdapter(mock);
96
-
97
- const eventsTask = collectEvents(adapter.events, 2);
98
- const openTask = adapter.open(new AbortController().signal);
99
- await waitFor(() => mock.sent.length > 0);
100
- mock.inject({ type: "session.updated" });
101
- await openTask;
102
-
103
- mock.inject({ type: "response.created" });
104
- mock.inject({
105
- type: "response.output_text.delta",
106
- delta: "Let me ",
107
- });
108
-
109
- const events = await eventsTask;
110
- expect(events).toEqual([
111
- { type: "response_started" },
112
- { type: "transcript", role: "assistant", text: "Let me ", final: false },
113
- ]);
114
- });
115
-
116
- it("surfaces response.output_text.done as final assistant transcript with accumulated text", async () => {
117
- const mock = createMockSocketHarness();
118
- const adapter = createAdapter(mock);
119
-
120
- const eventsTask = collectEvents(adapter.events, 3);
121
- const openTask = adapter.open(new AbortController().signal);
122
- await waitFor(() => mock.sent.length > 0);
123
- mock.inject({ type: "session.updated" });
124
- await openTask;
125
-
126
- mock.inject({ type: "response.created" });
127
- mock.inject({
128
- type: "response.output_text.delta",
129
- delta: "Let me ",
130
- });
131
- mock.inject({
132
- type: "response.output_text.done",
133
- text: "Let me check that.",
134
- });
135
-
136
- const events = await eventsTask;
137
- expect(events).toEqual([
138
- { type: "response_started" },
139
- { type: "transcript", role: "assistant", text: "Let me ", final: false },
140
- { type: "transcript", role: "assistant", text: "Let me check that.", final: true },
141
- ]);
142
- });
143
-
144
- it("uses assistantTranscript when response.output_text.done omits text", async () => {
145
- const mock = createMockSocketHarness();
146
- const adapter = createAdapter(mock);
147
-
148
- const eventsTask = collectEvents(adapter.events, 3);
149
- const openTask = adapter.open(new AbortController().signal);
150
- await waitFor(() => mock.sent.length > 0);
151
- mock.inject({ type: "session.updated" });
152
- await openTask;
153
-
154
- mock.inject({ type: "response.created" });
155
- mock.inject({
156
- type: "response.output_text.delta",
157
- delta: "Hello",
158
- });
159
- mock.inject({
160
- type: "response.output_text.done",
161
- });
162
-
163
- const events = await eventsTask;
164
- expect(events).toEqual([
165
- { type: "response_started" },
166
- { type: "transcript", role: "assistant", text: "Hello", final: false },
167
- { type: "transcript", role: "assistant", text: "Hello", final: true },
168
- ]);
169
- });
170
- });
171
-
172
- describe("createOpenAiCompatibleRealtimeAdapter requestResponse (Syrinx-owned turns)", () => {
173
- it("sends input_audio_buffer.commit then response.create", async () => {
174
- const mock = createMockSocketHarness();
175
- const adapter = createAdapter(mock);
176
-
177
- const openTask = adapter.open(new AbortController().signal);
178
- await waitFor(() => mock.sent.length > 0);
179
- mock.inject({ type: "session.updated" });
180
- await openTask;
181
-
182
- const baseline = mock.sent.length;
183
- adapter.requestResponse!();
184
-
185
- expect(JSON.parse(mock.sent[baseline]!)).toEqual({ type: "input_audio_buffer.commit" });
186
- expect(JSON.parse(mock.sent[baseline + 1]!)).toEqual({ type: "response.create" });
187
- });
188
- });