test(39-01): add failing tests for sentence streaming and multi-lang synthesis

server/src/__tests__/39-sentence-streaming.test.ts

@@ -0,0 +1,130 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+
+// Mock ffmpeg-static BEFORE any imports
+vi.mock("ffmpeg-static", () => ({ default: "/mock/ffmpeg" }));
+
+// Mock child_process
+vi.mock("node:child_process", () => ({
+  execFile: vi.fn(),
+  spawn: vi.fn(),
+}));
+
+// Mock fs/promises for temp file operations
+vi.mock("node:fs/promises", () => ({
+  writeFile: vi.fn().mockResolvedValue(undefined),
+  unlink: vi.fn().mockResolvedValue(undefined),
+}));
+
+import { splitSentences, voicePipelineService } from "../services/voice-pipeline.js";
+import { execFile as execFileCb } from "node:child_process";
+
+const execFileMock = vi.mocked(execFileCb);
+
+describe("sentence streaming", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+  });
+
+  describe("splitSentences", () => {
+    it("splits basic sentences on period, question mark, exclamation", () => {
+      const result = splitSentences("Hello world. How are you? I am fine.");
+      expect(result).toEqual(["Hello world.", "How are you?", "I am fine."]);
+    });
+
+    it("keeps abbreviation-style dots intact (Dr., D.C.)", () => {
+      const result = splitSentences("Dr. Smith went to D.C. He liked it.");
+      // Should split only at sentence boundaries, not abbreviations
+      // The sentence "Dr. Smith went to D.C." should be one unit
+      expect(result).toHaveLength(2);
+      expect(result[0]).toContain("Dr. Smith");
+      expect(result[1]).toBe("He liked it.");
+    });
+
+    it("returns single-sentence array when text has no sentence boundaries", () => {
+      const result = splitSentences("Hello world");
+      expect(result).toEqual(["Hello world"]);
+    });
+
+    it("filters empty strings", () => {
+      const result = splitSentences("Hello.   World.");
+      expect(result.every((s) => s.length > 0)).toBe(true);
+    });
+  });
+
+  describe("synthesizeSentenceStream", () => {
+    it("yields one chunk per sentence", async () => {
+      execFileMock.mockImplementation((_cmd: any, _args: any, _opts: any, callback: any) => {
+        callback(null, Buffer.from("wav-chunk"), "");
+        return {} as any;
+      });
+
+      const svc = voicePipelineService();
+      const text = "Hello world. How are you? I am fine.";
+
+      const chunks: Array<{ index: number; total: number; audio: Buffer }> = [];
+      for await (const chunk of svc.synthesizeSentenceStream(text)) {
+        chunks.push(chunk);
+      }
+
+      expect(chunks).toHaveLength(3);
+      expect(chunks[0]).toMatchObject({ index: 0, total: 3 });
+      expect(chunks[1]).toMatchObject({ index: 1, total: 3 });
+      expect(chunks[2]).toMatchObject({ index: 2, total: 3 });
+      chunks.forEach((c) => expect(Buffer.isBuffer(c.audio)).toBe(true));
+    });
+
+    it("yields single chunk for single-sentence text", async () => {
+      execFileMock.mockImplementation((_cmd: any, _args: any, _opts: any, callback: any) => {
+        callback(null, Buffer.from("wav-data"), "");
+        return {} as any;
+      });
+
+      const svc = voicePipelineService();
+      const chunks: Array<{ index: number; total: number; audio: Buffer }> = [];
+      for await (const chunk of svc.synthesizeSentenceStream("Hello world.")) {
+        chunks.push(chunk);
+      }
+
+      expect(chunks).toHaveLength(1);
+      expect(chunks[0]).toMatchObject({ index: 0, total: 1 });
+    });
+  });
+
+  describe("synthesizeMultiLang", () => {
+    it("returns a Map with one entry per voiceId", async () => {
+      execFileMock.mockImplementation((_cmd: any, _args: any, _opts: any, callback: any) => {
+        callback(null, Buffer.from("audio-data"), "");
+        return {} as any;
+      });
+
+      const svc = voicePipelineService();
+      const text = "Hello.";
+      const voiceIds = ["en_US-lessac-medium", "da_DK-talesyntese-medium"];
+
+      const result = await svc.synthesizeMultiLang(text, voiceIds);
+
+      expect(result).toBeInstanceOf(Map);
+      expect(result.size).toBe(2);
+      expect(result.has("en_US-lessac-medium")).toBe(true);
+      expect(result.has("da_DK-talesyntese-medium")).toBe(true);
+      voiceIds.forEach((v) => expect(Buffer.isBuffer(result.get(v))).toBe(true));
+    });
+
+    it("calls piper in parallel (Promise.all semantics)", async () => {
+      const callOrder: string[] = [];
+      execFileMock.mockImplementation((_cmd: any, args: any, _opts: any, callback: any) => {
+        const voiceId = args[1]; // second arg after '--model'
+        callOrder.push(voiceId);
+        callback(null, Buffer.from("audio"), "");
+        return {} as any;
+      });
+
+      const svc = voicePipelineService();
+      await svc.synthesizeMultiLang("Test.", ["voice-a", "voice-b"]);
+
+      // Both voices should have been requested
+      expect(callOrder).toContain("voice-a");
+      expect(callOrder).toContain("voice-b");
+    });
+  });
+});