feat: mid-stream language_codes updates via updateConfiguration (v4.35.3)

CHANGELOG.md

@@ -1,5 +1,17 @@
 # Changelog
 
+## [4.35.3]
+
+- Allow `language_codes` in `updateConfiguration()` — re-steer the transcription language mid-stream without reconnecting; pass `[]` to clear steering and restore the model's default multilingual code-switching (Universal-3.5 Pro Streaming only)
+
+## [4.35.2]
+
+- Add `opus` and `ogg_opus` streaming `encoding` values — accepts compressed Opus audio (raw packet-per-message or Ogg-encapsulated). `sampleRate` is ignored for both; the Opus stream is self-describing
+
+## [4.35.1]
+
+- Add `languageCodes` streaming parameter — steers transcription toward a set of languages (Universal-3.5 Pro Streaming only). Recommended replacement for `languageCode`, which is now deprecated but still supported
+
 ## [4.34.6]
 
 - Add `keepAlive()` method to `StreamingTranscriber` — sends a `KeepAlive` message to reset the server's inactivity timer when `inactivityTimeout` is configured

README.md

@@ -129,7 +129,7 @@ const audioFile = "https://assembly.ai/sports_injuries.mp3";
 
 const params = {
   audio: audioFile,
-  speech_models: ["universal-3-pro", "universal-2"],
+  speech_models: ["universal-3-5-pro", "universal-2"],
   language_detection: true,
 };
 
@@ -151,7 +151,7 @@ If you don't want to wait until the transcript is ready, you can use `submit`:
 ```js
 let transcript = await client.transcripts.submit({
   audio: "https://assembly.ai/espn.m4a",
-  speech_models: ["universal-3-pro", "universal-2"],
+  speech_models: ["universal-3-5-pro", "universal-2"],
   language_detection: true,
 });
 ```
@@ -167,7 +167,7 @@ When you create a transcript, you can either pass in a URL to an audio file or u
 // Upload a file via local path and transcribe
 let transcript = await client.transcripts.transcribe({
   audio: "./news.mp4",
-  speech_models: ["universal-3-pro", "universal-2"],
+  speech_models: ["universal-3-5-pro", "universal-2"],
   language_detection: true,
 });
 ```
@@ -182,7 +182,7 @@ If you don't want to wait until the transcript is ready, you can use `submit`:
 ```js
 let transcript = await client.transcripts.submit({
   audio: "./news.mp4",
-  speech_models: ["universal-3-pro", "universal-2"],
+  speech_models: ["universal-3-5-pro", "universal-2"],
   language_detection: true,
 });
 ```
@@ -206,7 +206,7 @@ const audioFile = "https://assembly.ai/wildfires.mp3";
 
 const params = {
   audio: audioFile,
-  speech_models: ["universal-3-pro", "universal-2"],
+  speech_models: ["universal-3-5-pro", "universal-2"],
   language_detection: true,
   speaker_labels: true,
 };
@@ -321,7 +321,6 @@ Create the streaming transcriber.
 const transcriber = client.streaming.transcriber({
   speechModel: "universal-3-5-pro",
   sampleRate: 16_000,
-  formatTurns: true,
 });
 ```
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "assemblyai",
-  "version": "4.35.0",
+  "version": "4.35.3",
   "description": "The AssemblyAI JavaScript SDK provides an easy-to-use interface for interacting with the AssemblyAI API, which supports async and real-time transcription, as well as the latest LeMUR models.",
   "engines": {
     "node": ">=18"

src/exports/streaming.ts

@@ -1,7 +1,7 @@
 export * from "../types/asyncapi.generated";
 export * from "../types/realtime";
 export * from "../types/helpers";
-export * from "../types/streaming/dual-channel";
+export * from "../types/streaming";
 export * from "../services/realtime/service";
 export * from "../services/streaming/service";
 export * from "../services/streaming/factory";

src/services/streaming/service.ts

@@ -313,9 +313,19 @@ export class StreamingTranscriber {
     }
 
     if (this.params.languageCode !== undefined) {
+      console.warn(
+        "[Deprecation Warning] `languageCode` is deprecated and will be removed in a future release. Please use `languageCodes` instead.",
+      );
       searchParams.set("language_code", this.params.languageCode);
     }
 
+    if (this.params.languageCodes !== undefined) {
+      searchParams.set(
+        "language_codes",
+        JSON.stringify(this.params.languageCodes),
+      );
+    }
+
     if (this.params.languageDetection !== undefined) {
       searchParams.set(
         "language_detection",

src/types/asyncapi.generated.ts

@@ -25,7 +25,7 @@ export type AudioData = ArrayBufferLike;
  * The encoding of the audio data
  * @defaultValue "pcm_s16"le
  */
-export type AudioEncoding = "pcm_s16le" | "pcm_mulaw";
+export type AudioEncoding = "pcm_s16le" | "pcm_mulaw" | "opus" | "ogg_opus";
 
 /**
  * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.

src/types/streaming/index.ts

@@ -109,7 +109,19 @@ export type StreamingTranscriberParams = {
   prompt?: string;
   agentContext?: string;
   speechModel?: StreamingSpeechModel;
+  /**
+   * @deprecated Use `languageCodes` instead (pass a single-element array, e.g. `["es"]`,
+   * for the same behavior). Still supported for backward compatibility.
+   */
   languageCode?: string;
+  /**
+   * Recommended way to select languages. Steers transcription toward a set of
+   * languages by biasing output toward them on a per-token basis while still
+   * allowing native code-switching among them. Pass the languages you expect
+   * (e.g. `["en", "es"]`), or a single-element array (e.g. `["es"]`) for a
+   * monolingual session. Universal-3.5 Pro Streaming only.
+   */
+  languageCodes?: string[];
   languageDetection?: boolean;
   domain?: StreamingDomain;
   inactivityTimeout?: number;
@@ -359,6 +371,12 @@ export type StreamingUpdateConfiguration = {
   filter_profanity?: boolean;
   interruption_delay?: number;
   turn_left_pad_ms?: number;
+  /**
+   * Steer transcription toward a set of languages mid-stream. Pass an empty
+   * array (`[]`) to clear steering and restore the model's default
+   * multilingual code-switching. Universal-3.5 Pro Streaming only.
+   */
+  language_codes?: string[];
 };
 
 export type StreamingForceEndpoint = {

tests/unit/streaming.test.ts

@@ -231,6 +231,26 @@ describe("streaming", () => {
     );
   });
 
+  it("should include language_codes in updateConfiguration message", async () => {
+    rt.updateConfiguration({ language_codes: ["en", "es"] });
+    await expect(server).toReceiveMessage(
+      JSON.stringify({
+        type: "UpdateConfiguration",
+        language_codes: ["en", "es"],
+      }),
+    );
+  });
+
+  it("should send an empty language_codes array in updateConfiguration to clear steering", async () => {
+    rt.updateConfiguration({ language_codes: [] });
+    await expect(server).toReceiveMessage(
+      JSON.stringify({
+        type: "UpdateConfiguration",
+        language_codes: [],
+      }),
+    );
+  });
+
   it("should send KeepAlive message on keepAlive()", async () => {
     rt.keepAlive();
     await expect(server).toReceiveMessage(
@@ -347,6 +367,47 @@ describe("streaming", () => {
     await connect(rt, server);
   });
 
+  it("should include language_codes in connection URL", async () => {
+    await cleanup();
+    WS.clean();
+
+    const languageCodes = ["en", "es"];
+    const wsUrl =
+      `${websocketBaseUrl}?token=123&sample_rate=16000&speech_model=universal-3-5-pro` +
+      `&language_codes=${encodeURIComponent(JSON.stringify(languageCodes))}`;
+    server = new WS(wsUrl);
+    rt = new StreamingTranscriber({
+      websocketBaseUrl,
+      token: "123",
+      sampleRate: 16_000,
+      speechModel: "universal-3-5-pro",
+      languageCodes: ["en", "es"],
+    });
+    onOpen = jest.fn();
+    rt.on("open", onOpen);
+    await connect(rt, server);
+  });
+
+  it.each(["opus", "ogg_opus"] as const)(
+    "should include %s encoding in connection URL",
+    async (encoding) => {
+      await cleanup();
+      WS.clean();
+
+      const wsUrl = `${websocketBaseUrl}?token=123&sample_rate=16000&encoding=${encoding}`;
+      server = new WS(wsUrl);
+      rt = new StreamingTranscriber({
+        websocketBaseUrl,
+        token: "123",
+        sampleRate: 16_000,
+        encoding,
+      });
+      onOpen = jest.fn();
+      rt.on("open", onOpen);
+      await connect(rt, server);
+    },
+  );
+
   it("should include whisper-rt speech model in connection URL", async () => {
     await cleanup();
     WS.clean();