From e3bbbae3bd9660bac9f0eba2fc8af6eb85fffd58 Mon Sep 17 00:00:00 2001
From: Mohamed Boudra <boudra.moha@gmail.com>
Date: Mon, 8 Jun 2026 14:29:52 +0700
Subject: [PATCH] Clean up OpenAI TTS test seam

---
 .../speech/providers/openai/tts.test.ts       | 67 ++++++++++---------
 .../src/server/speech/providers/openai/tts.ts | 49 +++++++++-----
 2 files changed, 69 insertions(+), 47 deletions(-)

diff --git a/packages/server/src/server/speech/providers/openai/tts.test.ts b/packages/server/src/server/speech/providers/openai/tts.test.ts
index 7f9a335b0a..0c607bd7cb 100644
--- a/packages/server/src/server/speech/providers/openai/tts.test.ts
+++ b/packages/server/src/server/speech/providers/openai/tts.test.ts
@@ -1,45 +1,40 @@
 import type pino from "pino";
 import { ReadableStream } from "node:stream/web";
-import { beforeEach, describe, expect, test, vi } from "vitest";
+import pinoLogger from "pino";
+import { describe, expect, test } from "vitest";
 
-const openAiMocks = vi.hoisted(() => ({
-  createSpeech: vi.fn(),
-}));
+import {
+  OpenAITTS,
+  type OpenAITTSAudioClient,
+  type OpenAITTSSpeechRequest,
+  type OpenAITTSSpeechResponse,
+} from "./tts.js";
 
-vi.mock("openai", () => ({
-  OpenAI: vi.fn().mockImplementation(
-    class {
-      audio = {
-        speech: {
-          create: openAiMocks.createSpeech,
+class FakeOpenAITTSClient implements OpenAITTSAudioClient {
+  public readonly speechRequests: OpenAITTSSpeechRequest[] = [];
+  public readonly audio: OpenAITTSAudioClient["audio"];
+  private readonly speechResponse: OpenAITTSSpeechResponse;
+
+  constructor(speechResponse: OpenAITTSSpeechResponse) {
+    this.speechResponse = speechResponse;
+    this.audio = {
+      speech: {
+        create: async (request: OpenAITTSSpeechRequest) => {
+          this.speechRequests.push(request);
+          return this.speechResponse;
         },
-      };
-    },
-  ),
-}));
-
-import { OpenAITTS } from "./tts.js";
+      },
+    };
+  }
+}
 
 function createLogger(): pino.Logger {
-  const logger = {
-    child: vi.fn(),
-    debug: vi.fn(),
-    error: vi.fn(),
-    info: vi.fn(),
-  };
-
-  logger.child.mockReturnValue(logger);
-
-  return logger as unknown as pino.Logger;
+  return pinoLogger({ enabled: false });
 }
 
 describe("OpenAITTS", () => {
-  beforeEach(() => {
-    openAiMocks.createSpeech.mockReset();
-  });
-
   test("returns a Node stream from the OpenAI web response body", async () => {
-    openAiMocks.createSpeech.mockResolvedValue({
+    const openaiClient = new FakeOpenAITTSClient({
       body: new ReadableStream<Uint8Array>({
         start(controller) {
           controller.enqueue(Buffer.from("audio"));
@@ -48,7 +43,7 @@ describe("OpenAITTS", () => {
       }),
     });
 
-    const tts = new OpenAITTS({ apiKey: "sk-test" }, createLogger());
+    const tts = new OpenAITTS({ apiKey: "sk-test" }, createLogger(), { openaiClient });
     const result = await tts.synthesizeSpeech("hello");
 
     const chunks: Buffer[] = [];
@@ -58,5 +53,13 @@ describe("OpenAITTS", () => {
 
     expect(Buffer.concat(chunks).toString()).toBe("audio");
     expect(result.format).toBe("pcm");
+    expect(openaiClient.speechRequests).toEqual([
+      {
+        model: "tts-1",
+        voice: "alloy",
+        input: "hello",
+        response_format: "pcm",
+      },
+    ]);
   });
 });
diff --git a/packages/server/src/server/speech/providers/openai/tts.ts b/packages/server/src/server/speech/providers/openai/tts.ts
index 9813b4760a..e71129fe34 100644
--- a/packages/server/src/server/speech/providers/openai/tts.ts
+++ b/packages/server/src/server/speech/providers/openai/tts.ts
@@ -13,12 +13,35 @@ export interface TTSConfig {
   responseFormat?: "mp3" | "opus" | "aac" | "flac" | "wav" | "pcm";
 }
 
+export interface OpenAITTSSpeechRequest {
+  model: NonNullable<TTSConfig["model"]>;
+  voice: NonNullable<TTSConfig["voice"]>;
+  input: string;
+  response_format: NonNullable<TTSConfig["responseFormat"]>;
+}
+
+export interface OpenAITTSSpeechResponse {
+  body: unknown;
+}
+
+export interface OpenAITTSAudioClient {
+  audio: {
+    speech: {
+      create(request: OpenAITTSSpeechRequest): Promise<OpenAITTSSpeechResponse>;
+    };
+  };
+}
+
+export interface OpenAITTSOptions {
+  openaiClient?: OpenAITTSAudioClient;
+}
+
 export class OpenAITTS implements TextToSpeechProvider {
-  private readonly openaiClient: OpenAI;
-  private readonly config: TTSConfig;
+  private readonly openaiClient: OpenAITTSAudioClient;
+  private readonly config: Required<TTSConfig>;
   private readonly logger: pino.Logger;
 
-  constructor(ttsConfig: TTSConfig, parentLogger: pino.Logger) {
+  constructor(ttsConfig: TTSConfig, parentLogger: pino.Logger, options: OpenAITTSOptions = {}) {
     this.config = {
       model: "tts-1",
       voice: "alloy",
@@ -26,9 +49,11 @@ export class OpenAITTS implements TextToSpeechProvider {
       ...ttsConfig,
     };
     this.logger = parentLogger.child({ module: "agent", provider: "openai", component: "tts" });
-    this.openaiClient = new OpenAI({
-      apiKey: ttsConfig.apiKey,
-    });
+    this.openaiClient =
+      options.openaiClient ??
+      new OpenAI({
+        apiKey: ttsConfig.apiKey,
+      });
 
     this.logger.info(
       { voice: this.config.voice, model: this.config.model, format: this.config.responseFormat },
@@ -54,16 +79,10 @@ export class OpenAITTS implements TextToSpeechProvider {
       );
 
       const response = await this.openaiClient.audio.speech.create({
-        model: this.config.model!,
-        voice: this.config.voice!,
+        model: this.config.model,
+        voice: this.config.voice,
         input: text,
-        response_format: this.config.responseFormat as
-          | "mp3"
-          | "opus"
-          | "aac"
-          | "flac"
-          | "wav"
-          | "pcm",
+        response_format: this.config.responseFormat,
       });
 
       if (!response.body) {