feat(regen): add Flux TTS batch (REST) endpoint and agent latency report (#519)

GregHolmes · fern-api[bot] · web-flow · commit a27336e24257 · 2026-07-14T10:52:57.000-04:00
* chore: initialize SDK regeneration branch

* chore: unfreeze files pending regen

* SDK regeneration

* chore: re-apply manual patches after regen

* test: add Flux TTS batch (REST) wire coverage

---------

Co-authored-by: fern-api[bot] &lt;115122769+fern-api[bot]@users.noreply.github.com&gt;
diff --git a/.fern/metadata.json b/.fern/metadata.json
@@ -11,8 +11,8 @@
         "shouldGenerateWebsocketClients": true,
         "enableForwardCompatibleEnums": true
     },
-    "originGitCommit": "335af96251466afae7a9f71badb65c630a48626e",
+    "originGitCommit": "14c0ca94fcd0279d99570389d7b688dc3a18ea41",
     "originGitCommitIsDirty": true,
     "invokedBy": "manual",
-    "sdkVersion": "5.5.0"
+    "sdkVersion": "5.5.1"
 }
diff --git a/.fernignore b/.fernignore
@@ -27,6 +27,7 @@ tests/unit/error-handling.test.ts
 tests/unit/multiple-keyterms.test.ts
 tests/unit/nodejs-version-compatibility.test.ts
 tests/unit/regen-constraints.test.ts
+tests/unit/speak-v2-batch.test.ts
 tests/unit/transport-factory.test.ts
 tests/unit/websocket-reconnection.test.ts
 tests/unit/websocket-wrappers.test.ts
diff --git a/package.json b/package.json
@@ -463,6 +463,17 @@
             },
             "default": "./dist/cjs/api/resources/speak/resources/v1/resources/audio/exports.js"
         },
+        "./speak/v2/audio": {
+            "import": {
+                "types": "./dist/esm/api/resources/speak/resources/v2/resources/audio/exports.d.mts",
+                "default": "./dist/esm/api/resources/speak/resources/v2/resources/audio/exports.mjs"
+            },
+            "require": {
+                "types": "./dist/cjs/api/resources/speak/resources/v2/resources/audio/exports.d.ts",
+                "default": "./dist/cjs/api/resources/speak/resources/v2/resources/audio/exports.js"
+            },
+            "default": "./dist/cjs/api/resources/speak/resources/v2/resources/audio/exports.js"
+        },
         "./voiceAgent": {
             "import": {
                 "types": "./dist/esm/api/resources/voiceAgent/exports.d.mts",
diff --git a/reference.md b/reference.md
@@ -2930,6 +2930,73 @@ await client.speak.v1.audio.generate({
 </dl>
 
 
+</dd>
+</dl>
+</details>
+
+## Speak V2 Audio
+<details><summary><code>client.speak.v2.audio.<a href="/src/api/resources/speak/resources/v2/resources/audio/client/Client.ts">generate</a>({ ...params }) -> core.BinaryResponse</code></summary>
+<dl>
+<dd>
+
+#### 📝 Description
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+Synthesize a complete block of text into a single audio response using Deepgram's Flux TTS batch (REST) API. Use this for pre-rendering fixed audio (IVR prompts, notifications, narration) where the whole text is known up front and you don't need incremental playback or interruption.
+</dd>
+</dl>
+</dd>
+</dl>
+
+#### 🔌 Usage
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+```typescript
+await client.speak.v2.audio.generate({
+    model: "model",
+    text: "text"
+});
+
+```
+</dd>
+</dl>
+</dd>
+</dl>
+
+#### ⚙️ Parameters
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+**request:** `Deepgram.speak.v2.SpeakV2Request` 
+    
+</dd>
+</dl>
+
+<dl>
+<dd>
+
+**requestOptions:** `AudioClient.RequestOptions` 
+    
+</dd>
+</dl>
+</dd>
+</dl>
+
+
 </dd>
 </dl>
 </details>
diff --git a/src/api/resources/agent/resources/v1/client/Socket.ts b/src/api/resources/agent/resources/v1/client/Socket.ts
@@ -21,6 +21,7 @@ export declare namespace V1Socket {
         | Deepgram.agent.AgentV1ConversationText
         | Deepgram.agent.AgentV1UserStartedSpeaking
         | Deepgram.agent.AgentV1AgentThinking
+        | Deepgram.agent.AgentV1LatencyReport
         | Deepgram.agent.AgentV1FunctionCallRequest
         | Deepgram.agent.AgentV1AgentStartedSpeaking
         | Deepgram.agent.AgentV1AgentAudioDone
diff --git a/src/api/resources/agent/resources/v1/types/AgentV1InjectAgentMessage.ts b/src/api/resources/agent/resources/v1/types/AgentV1InjectAgentMessage.ts
@@ -10,6 +10,7 @@ export interface AgentV1InjectAgentMessage {
      *
      * * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
      * * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
+     * * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
      */
     behavior?: AgentV1InjectAgentMessage.Behavior | undefined;
 }
@@ -20,10 +21,12 @@ export namespace AgentV1InjectAgentMessage {
      *
      * * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
      * * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
+     * * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
      */
     export const Behavior = {
         Default: "default",
         Queue: "queue",
+        Interrupt: "interrupt",
     } as const;
     export type Behavior = (typeof Behavior)[keyof typeof Behavior] | string;
 }
diff --git a/src/api/resources/agent/resources/v1/types/AgentV1LatencyReport.ts b/src/api/resources/agent/resources/v1/types/AgentV1LatencyReport.ts
@@ -0,0 +1,20 @@
+// This file was auto-generated by Fern from our API Definition.
+
+export interface AgentV1LatencyReport {
+    /** Message type identifier for the latency report */
+    type: "LatencyReport";
+    /** Speech-to-text: time from audio received to transcript produced, in seconds */
+    stt_latency?: number | undefined;
+    /** Time to first token of any type (text, tool call, or thinking), in seconds */
+    ttt_token_latency?: number | undefined;
+    /** Time to first text token from the LLM, in seconds */
+    ttt_text_latency?: number | undefined;
+    /** Time to first tool-call token from the LLM, in seconds */
+    ttt_tool_latency?: number | undefined;
+    /** Time to first thinking token from the LLM, in seconds */
+    ttt_thinking_latency?: number | undefined;
+    /** Text-to-speech: time from first text token to first audio byte, in seconds */
+    tts_latency?: number | undefined;
+    /** End-to-end: time from user utterance end to first audio byte, in seconds */
+    total_latency?: number | undefined;
+}
diff --git a/src/api/resources/agent/resources/v1/types/index.ts b/src/api/resources/agent/resources/v1/types/index.ts
@@ -9,6 +9,7 @@ export * from "./AgentV1InjectAgentMessage.js";
 export * from "./AgentV1InjectionRefused.js";
 export * from "./AgentV1InjectUserMessage.js";
 export * from "./AgentV1KeepAlive.js";
+export * from "./AgentV1LatencyReport.js";
 export * from "./AgentV1ListenUpdated.js";
 export * from "./AgentV1PromptUpdated.js";
 export * from "./AgentV1ReceiveFunctionCallResponse.js";
diff --git a/src/api/resources/speak/resources/v2/client/Client.ts b/src/api/resources/speak/resources/v2/client/Client.ts
@@ -1,12 +1,13 @@
 // This file was auto-generated by Fern from our API Definition.
 
 import type { BaseClientOptions } from "../../../../../../BaseClient.js";
-import { type NormalizedClientOptions, normalizeClientOptions } from "../../../../../../BaseClient.js";
+import { type NormalizedClientOptionsWithAuth, normalizeClientOptionsWithAuth } from "../../../../../../BaseClient.js";
 import { mergeHeaders, mergeOnlyDefinedHeaders } from "../../../../../../core/headers.js";
 import * as core from "../../../../../../core/index.js";
 import { toJson } from "../../../../../../core/json.js";
 import * as environments from "../../../../../../environments.js";
 import type * as Deepgram from "../../../../../index.js";
+import { AudioClient } from "../resources/audio/client/Client.js";
 import { V2Socket } from "./Socket.js";
 
 export declare namespace V2Client {
@@ -37,10 +38,15 @@ export declare namespace V2Client {
 }
 
 export class V2Client {
-    protected readonly _options: NormalizedClientOptions<V2Client.Options>;
+    protected readonly _options: NormalizedClientOptionsWithAuth<V2Client.Options>;
+    protected _audio: AudioClient | undefined;
 
     constructor(options: V2Client.Options = {}) {
-        this._options = normalizeClientOptions(options);
+        this._options = normalizeClientOptionsWithAuth(options);
+    }
+
+    public get audio(): AudioClient {
+        return (this._audio ??= new AudioClient(this._options));
     }
 
     public async connect(args: V2Client.ConnectArgs): Promise<V2Socket> {
diff --git a/src/api/resources/speak/resources/v2/exports.ts b/src/api/resources/speak/resources/v2/exports.ts
@@ -3,3 +3,4 @@
 export { V2Client } from "./client/Client.js";
 export * from "./client/index.js";
 export { V2Socket } from "./client/Socket.js";
+export * from "./resources/index.js";
diff --git a/src/api/resources/speak/resources/v2/index.ts b/src/api/resources/speak/resources/v2/index.ts
@@ -1,2 +1,3 @@
 export * from "./client/index.js";
+export * from "./resources/index.js";
 export * from "./types/index.js";
diff --git a/src/api/resources/speak/resources/v2/resources/audio/client/Client.ts b/src/api/resources/speak/resources/v2/resources/audio/client/Client.ts
@@ -0,0 +1,120 @@
+// This file was auto-generated by Fern from our API Definition.
+
+import type { BaseClientOptions, BaseRequestOptions } from "../../../../../../../../BaseClient.js";
+import {
+    type NormalizedClientOptionsWithAuth,
+    normalizeClientOptionsWithAuth,
+} from "../../../../../../../../BaseClient.js";
+import { mergeHeaders } from "../../../../../../../../core/headers.js";
+import * as core from "../../../../../../../../core/index.js";
+import * as environments from "../../../../../../../../environments.js";
+import { handleNonStatusCodeError } from "../../../../../../../../errors/handleNonStatusCodeError.js";
+import * as errors from "../../../../../../../../errors/index.js";
+import * as Deepgram from "../../../../../../../index.js";
+
+export declare namespace AudioClient {
+    export type Options = BaseClientOptions;
+
+    export interface RequestOptions extends BaseRequestOptions {}
+}
+
+export class AudioClient {
+    protected readonly _options: NormalizedClientOptionsWithAuth<AudioClient.Options>;
+
+    constructor(options: AudioClient.Options = {}) {
+        this._options = normalizeClientOptionsWithAuth(options);
+    }
+
+    /**
+     * Synthesize a complete block of text into a single audio response using Deepgram's Flux TTS batch (REST) API. Use this for pre-rendering fixed audio (IVR prompts, notifications, narration) where the whole text is known up front and you don't need incremental playback or interruption.
+     *
+     * @throws {@link Deepgram.BadRequestError}
+     */
+    public generate(
+        request: Deepgram.speak.v2.SpeakV2Request,
+        requestOptions?: AudioClient.RequestOptions,
+    ): core.HttpResponsePromise<core.BinaryResponse> {
+        return core.HttpResponsePromise.fromPromise(this.__generate(request, requestOptions));
+    }
+
+    private async __generate(
+        request: Deepgram.speak.v2.SpeakV2Request,
+        requestOptions?: AudioClient.RequestOptions,
+    ): Promise<core.WithRawResponse<core.BinaryResponse>> {
+        const {
+            callback,
+            callback_method: callbackMethod,
+            mip_opt_out: mipOptOut,
+            tag,
+            bit_rate: bitRate,
+            container,
+            encoding,
+            model,
+            sample_rate: sampleRate,
+            priority,
+            ..._body
+        } = request;
+        const _queryParams: Record<string, unknown> = {
+            callback,
+            callback_method: callbackMethod != null ? callbackMethod : undefined,
+            mip_opt_out: mipOptOut,
+            tag,
+            bit_rate: bitRate,
+            container: container != null ? container : undefined,
+            encoding: encoding != null ? encoding : undefined,
+            model,
+            sample_rate: sampleRate,
+            priority: priority != null ? priority : undefined,
+        };
+        const _authRequest: core.AuthRequest = await this._options.authProvider.getAuthRequest();
+        const _headers: core.Fetcher.Args["headers"] = mergeHeaders(
+            _authRequest.headers,
+            this._options?.headers,
+            requestOptions?.headers,
+        );
+        const _response = await (this._options.fetcher ?? core.fetcher)<core.BinaryResponse>({
+            url: core.url.join(
+                (await core.Supplier.get(this._options.baseUrl)) ??
+                    (
+                        (await core.Supplier.get(this._options.environment)) ??
+                        environments.DeepgramEnvironment.Production
+                    ).base,
+                "v2/speak",
+            ),
+            method: "POST",
+            headers: _headers,
+            contentType: "application/json",
+            queryString: core.url
+                .queryBuilder()
+                .addMany(_queryParams)
+                .mergeAdditional(requestOptions?.queryParams)
+                .build(),
+            requestType: "json",
+            body: _body,
+            responseType: "binary-response",
+            timeoutMs: (requestOptions?.timeoutInSeconds ?? this._options?.timeoutInSeconds ?? 60) * 1000,
+            maxRetries: requestOptions?.maxRetries ?? this._options?.maxRetries,
+            abortSignal: requestOptions?.abortSignal,
+            fetchFn: this._options?.fetch,
+            logging: this._options.logging,
+        });
+        if (_response.ok) {
+            return { data: _response.body, rawResponse: _response.rawResponse };
+        }
+
+        if (_response.error.reason === "status-code") {
+            switch (_response.error.statusCode) {
+                case 400:
+                    throw new Deepgram.BadRequestError(_response.error.body as unknown, _response.rawResponse);
+                default:
+                    throw new errors.DeepgramError({
+                        statusCode: _response.error.statusCode,
+                        body: _response.error.body,
+                        rawResponse: _response.rawResponse,
+                    });
+            }
+        }
+
+        return handleNonStatusCodeError(_response.error, _response.rawResponse, "POST", "/v2/speak");
+    }
+}
diff --git a/src/api/resources/speak/resources/v2/resources/audio/client/index.ts b/src/api/resources/speak/resources/v2/resources/audio/client/index.ts
@@ -0,0 +1 @@
+export * from "./requests/index.js";
diff --git a/src/api/resources/speak/resources/v2/resources/audio/client/requests/SpeakV2Request.ts b/src/api/resources/speak/resources/v2/resources/audio/client/requests/SpeakV2Request.ts
@@ -0,0 +1,35 @@
+// This file was auto-generated by Fern from our API Definition.
+
+import type * as Deepgram from "../../../../../../../../index.js";
+
+/**
+ * @example
+ *     {
+ *         model: "model",
+ *         text: "text"
+ *     }
+ */
+export interface SpeakV2Request {
+    /** URL to which we'll make the callback request */
+    callback?: string;
+    /** HTTP method by which the callback request will be made */
+    callback_method?: Deepgram.speak.v2.AudioGenerateRequestCallbackMethod;
+    /** Opts out requests from the Deepgram Model Improvement Program. Refer to our Docs for pricing impacts before setting this to true. https://dpgr.am/deepgram-mip */
+    mip_opt_out?: boolean;
+    /** Label your requests for the purpose of identification during usage reporting */
+    tag?: string | string[];
+    /** The bitrate of the audio in bits per second. Choose from predefined ranges or specific values based on the encoding type. */
+    bit_rate?: number;
+    /** Container specifies the file format wrapper for the output audio. The available options depend on the encoding type. */
+    container?: Deepgram.speak.v2.AudioGenerateRequestContainer;
+    /** Encoding allows you to specify the expected encoding of your audio output */
+    encoding?: Deepgram.speak.v2.AudioGenerateRequestEncoding;
+    /** Flux TTS model used to synthesize the submitted text, in the form `flux-{voice}-{language}` (for example, `flux-alexis-en`). Required; unlike the v1 (Aura) endpoint there is no default and only flux models are accepted. English-only at launch. The voice catalog is not yet finalized, so this is typed as a free-form string rather than a closed enum; it can be locked to an enum (matching STT Flux `ListenV2Model`) once the GA catalog is confirmed. */
+    model: string;
+    /** Sample Rate specifies the sample rate for the output audio. Based on the encoding, different sample rates are supported. For some encodings, the sample rate is not configurable */
+    sample_rate?: number;
+    /** Processing priority for asynchronous (callback) requests. The only supported value is low. */
+    priority?: "low";
+    /** The text content to be converted to speech. The server normalizes and preprocesses the text (e.g. stripping inline controls) before synthesis. */
+    text: string;
+}
diff --git a/src/api/resources/speak/resources/v2/resources/audio/client/requests/index.ts b/src/api/resources/speak/resources/v2/resources/audio/client/requests/index.ts
@@ -0,0 +1 @@
+export type { SpeakV2Request } from "./SpeakV2Request.js";
diff --git a/src/api/resources/speak/resources/v2/resources/audio/exports.ts b/src/api/resources/speak/resources/v2/resources/audio/exports.ts
@@ -0,0 +1,4 @@
+// This file was auto-generated by Fern from our API Definition.
+
+export { AudioClient } from "./client/Client.js";
+export * from "./client/index.js";
diff --git a/src/api/resources/speak/resources/v2/resources/audio/index.ts b/src/api/resources/speak/resources/v2/resources/audio/index.ts
@@ -0,0 +1,2 @@
+export * from "./client/index.js";
+export * from "./types/index.js";
diff --git a/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestCallbackMethod.ts b/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestCallbackMethod.ts
@@ -0,0 +1,9 @@
+// This file was auto-generated by Fern from our API Definition.
+
+export const AudioGenerateRequestCallbackMethod = {
+    Post: "POST",
+    Put: "PUT",
+} as const;
+export type AudioGenerateRequestCallbackMethod =
+    | (typeof AudioGenerateRequestCallbackMethod)[keyof typeof AudioGenerateRequestCallbackMethod]
+    | string;
diff --git a/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestContainer.ts b/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestContainer.ts
diff --git a/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestEncoding.ts b/src/api/resources/speak/resources/v2/resources/audio/types/AudioGenerateRequestEncoding.ts
diff --git a/src/api/resources/speak/resources/v2/resources/audio/types/index.ts b/src/api/resources/speak/resources/v2/resources/audio/types/index.ts
diff --git a/src/api/resources/speak/resources/v2/resources/index.ts b/src/api/resources/speak/resources/v2/resources/index.ts
diff --git a/src/api/types/ListenV2Keyterm.ts b/src/api/types/ListenV2Keyterm.ts
diff --git a/src/api/types/SpeakV2AcceptedResponse.ts b/src/api/types/SpeakV2AcceptedResponse.ts
diff --git a/src/api/types/SpeakV2Response.ts b/src/api/types/SpeakV2Response.ts
diff --git a/src/api/types/index.ts b/src/api/types/index.ts
diff --git a/tests/unit/speak-v2-batch.test.ts b/tests/unit/speak-v2-batch.test.ts

Original file line number	Diff line number	Diff line change
`@@ -10,6 +10,7 @@ export interface AgentV1InjectAgentMessage {`
`10`	`10`	`*`
`11`	`11`	* * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
`12`	`12`	* * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
	`13`	+ * * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
`13`	`14`	`*/`
`14`	`15`	`behavior?: AgentV1InjectAgentMessage.Behavior \| undefined;`
`15`	`16`	`}`
`@@ -20,10 +21,12 @@ export namespace AgentV1InjectAgentMessage {`
`20`	`21`	`*`
`21`	`22`	* * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
`22`	`23`	* * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
	`24`	+ * * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
`23`	`25`	`*/`
`24`	`26`	`export const Behavior = {`
`25`	`27`	`Default: "default",`
`26`	`28`	`Queue: "queue",`
	`29`	`+ Interrupt: "interrupt",`
`27`	`30`	`} as const;`
`28`	`31`	`export type Behavior = (typeof Behavior)[keyof typeof Behavior] \| string;`
`29`	`32`	`}`