assemblyai 2.0.2 → 3.0.1

package/dist/types/openapi.generated.d.ts

@@ -64,19 +64,24 @@ export type ContentSafetyLabelResult = {
     sentences_idx_end: number;
     /** @description The sentence index at which the section begins */
     sentences_idx_start: number;
+    /** @description The transcript of the section flagged by the Content Moderation model */
+    text: string;
+    /** @description Timestamp information for the section */
+    timestamp: Timestamp;
+};
+export type ContentSafetyLabelsResult = {
+    results: ContentSafetyLabelResult[];
     /** @description A summary of the Content Moderation severity results for the entire audio file */
     severity_score_summary: {
         [key: string]: SeverityScoreSummary;
     };
+    /** @description Will be either success, or unavailable in the rare case that the Content Moderation model failed. */
+    status: AudioIntelligenceModelStatus;
     /** @description A summary of the Content Moderation confidence results for the entire audio file */
     summary: {
         [key: string]: number;
     };
-    /** @description The transcript of the section flagged by the Content Moderation model */
-    text: string;
-    /** @description Timestamp information for the section */
-    timestamp: Timestamp;
-};
+} | null;
 export type CreateRealtimeTemporaryTokenParameters = {
     /** @description The amount of time until the token expires in seconds. */
     expires_in: number;
@@ -347,7 +352,7 @@ export type SentimentAnalysisResult = {
     end: number;
     /** @description The detected sentiment for the sentence, one of POSITIVE, NEUTRAL, NEGATIVE */
     sentiment: Sentiment;
-    /** @description The speaker of the sentence if Speaker Diarization is enabled, else null */
+    /** @description The speaker of the sentence if [Speaker Diarization](https://www.assemblyai.com/docs/models/speaker-diarization) is enabled, else null */
     speaker?: string | null;
     /** @description The starting time, in milliseconds, of the sentence */
     start: number;
@@ -391,7 +396,7 @@ export type Timestamp = {
     /** @description The start time in milliseconds */
     start: number;
 };
-/** @description THe result of the topic detection model. */
+/** @description The result of the topic detection model. */
 export type TopicDetectionResult = {
     labels?: {
         /** @description The IAB taxonomical label for the label of the detected topic, where > denotes supertopic/subtopic relationship */
@@ -448,11 +453,7 @@ export type Transcript = {
      * @description An array of results for the Content Moderation model, if it was enabled during the transcription request.
      * See [Content moderation](https://www.assemblyai.com/docs/Models/content_moderation) for more information.
      */
-    content_safety_labels?: {
-        results: ContentSafetyLabelResult[];
-        /** @description Will be either success, or unavailable in the rare case that the Content Safety Labels model failed. */
-        status: AudioIntelligenceModelStatus;
-    } | null;
+    content_safety_labels?: ContentSafetyLabelsResult;
     /** @description Customize how words are spelled and formatted using to and from values */
     custom_spelling?: TranscriptCustomSpelling[] | null;
     /** @description Whether custom topics was enabled in the transcription request, either true or false */
@@ -477,7 +478,7 @@ export type Transcript = {
     /** @description Enable [Topic Detection](https://www.assemblyai.com/docs/Models/iab_classification), can be true or false */
     iab_categories?: boolean | null;
     /**
-     * @description An array of results for the Topic Detection model, if it was enabled during the transcription request.
+     * @description The result of the Topic Detection model, if it was enabled during the transcription request.
      * See [Topic Detection](https://www.assemblyai.com/docs/Models/iab_classification) for more information.
      */
     iab_categories_result?: {
@@ -620,6 +621,27 @@ export type TranscriptListItem = {
     resource_url: string;
     status: TranscriptStatus;
 };
+export type TranscriptListParameters = {
+    /** @description Get transcripts that were created after this transcript ID */
+    after_id?: string;
+    /** @description Get transcripts that were created before this transcript ID */
+    before_id?: string;
+    /**
+     * Format: date
+     * @description Only get transcripts created on this date
+     */
+    created_on?: string;
+    /**
+     * Format: int64
+     * @description Maximum amount of transcripts to retrieve
+     * @default 10
+     */
+    limit?: number;
+    /** @description Filter by transcript status */
+    status?: TranscriptStatus;
+    /** @description Only get throttled transcripts, overrides the status filter */
+    throttled_only?: boolean;
+};
 export type TranscriptParagraph = {
     /** Format: double */
     confidence: number;
@@ -642,12 +664,20 @@ export type TranscriptSentence = {
  */
 export type TranscriptStatus = "queued" | "processing" | "completed" | "error";
 export type TranscriptUtterance = {
-    channel: string;
-    /** Format: double */
+    /**
+     * Format: double
+     * @description The confidence score for the transcript of this utterance
+     */
     confidence: number;
+    /** @description The ending time, in milliseconds, of the utterance in the audio file */
     end: number;
+    /** @description The speaker of this utterance, where each speaker is assigned a sequential capital letter - e.g. "A" for Speaker A, "B" for Speaker B, etc. */
+    speaker: string;
+    /** @description The starting time, in milliseconds, of the utterance in the audio file */
     start: number;
+    /** @description The text for this utterance */
     text: string;
+    /** @description The words in the utterance. */
     words: TranscriptWord[];
 };
 export type TranscriptWord = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "assemblyai",
-  "version": "2.0.2",
+  "version": "3.0.1",
   "description": "The AssemblyAI Node.js 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.",
   "main": "dist/index.js",
   "module": "dist/index.esm.js",
@@ -51,6 +51,7 @@
     "i": "^0.3.7",
     "jest": "^29.5.0",
     "jest-cli": "^29.5.0",
+    "jest-fetch-mock": "^3.0.3",
     "jest-junit": "^16.0.0",
     "jest-mock-extended": "^3.0.4",
     "jest-websocket-mock": "^2.4.1",
@@ -67,7 +68,6 @@
     "typescript": "^5.2.2"
   },
   "dependencies": {
-    "axios": "^1.4.0",
     "ws": "^8.13.0"
   }
 }

package/src/index.ts

@@ -1,7 +1,2 @@
-import * as services from "./services";
-import { AssemblyAI } from "./services";
-export * from "./services";
 export type * from "./types";
-export default AssemblyAI;
-class AssemblyAIExports extends AssemblyAI {}
-module.exports = Object.assign(AssemblyAIExports, services);
+export * from "./services";

package/src/services/base.ts

@@ -1,4 +1,5 @@
-import { AxiosInstance } from "axios";
+import { BaseServiceParams } from "..";
+import { Error as JsonError } from "..";
 
 /**
  * Base class for services that communicate with the API.
@@ -6,7 +7,47 @@ import { AxiosInstance } from "axios";
 export abstract class BaseService {
   /**
    * Create a new service.
-   * @param params The AxiosInstance to send HTTP requests to the API.
+   * @param params The parameters to use for the service.
    */
-  constructor(protected client: AxiosInstance) {}
+  constructor(private params: BaseServiceParams) {}
+  protected async fetch(
+    input: string,
+    init?: RequestInit | undefined
+  ): Promise<Response> {
+    init = init ?? {};
+    init.headers = init.headers ?? {};
+    init.headers = {
+      Authorization: this.params.apiKey,
+      "Content-Type": "application/json",
+      ...init.headers,
+    };
+    if (!input.startsWith("http")) input = this.params.baseUrl + input;
+
+    const response = await fetch(input, init);
+
+    if (response.status >= 400) {
+      let json: JsonError | undefined;
+      const text = await response.text();
+      if (text) {
+        try {
+          json = JSON.parse(text);
+        } catch {
+          /* empty */
+        }
+        if (json?.error) throw new Error(json.error);
+        throw new Error(text);
+      }
+      throw new Error(`HTTP Error: ${response.status} ${response.statusText}`);
+    }
+
+    return response;
+  }
+
+  protected async fetchJson<T>(
+    input: string,
+    init?: RequestInit | undefined
+  ): Promise<T> {
+    const response = await this.fetch(input, init);
+    return response.json() as Promise<T>;
+  }
 }

package/src/services/files/index.ts

@@ -1,22 +1,28 @@
-import { readFile } from "fs/promises";
-import { BaseService } from "@/services/base";
-import { UploadedFile } from "@/types";
+// import the fs module instead if specific named exports
+// to keep the assemblyai module more compatible. Some fs polyfills don't include `createReadStream`.
+import fs from "fs";
+import { BaseService } from "../base";
+import { UploadedFile, FileUploadParameters, FileUploadData } from "../..";
 
 export class FileService extends BaseService {
   /**
    * Upload a local file to AssemblyAI.
-   * @param path The local file to upload.
+   * @param input The local file path to upload, or a stream or buffer of the file to upload.
    * @return A promise that resolves to the uploaded file URL.
    */
-  async upload(path: string): Promise<string> {
-    const file = await readFile(path);
+  async upload(input: FileUploadParameters): Promise<string> {
+    let fileData: FileUploadData;
+    if (typeof input === "string") fileData = fs.createReadStream(input);
+    else fileData = input;
 
-    const { data } = await this.client.post<UploadedFile>("/v2/upload", file, {
+    const data = await this.fetchJson<UploadedFile>("/v2/upload", {
+      method: "POST",
+      body: fileData as BodyInit,
       headers: {
         "Content-Type": "application/octet-stream",
       },
-    });
-
+      duplex: "half",
+    } as RequestInit);
     return data.upload_url;
   }
 }

package/src/services/index.ts

@@ -1,10 +1,11 @@
-import { createAxiosClient } from "@/utils/axios";
-import { BaseServiceParams } from "@/types";
+import { BaseServiceParams } from "..";
 import { LemurService } from "./lemur";
 import { RealtimeService, RealtimeServiceFactory } from "./realtime";
 import { TranscriptService } from "./transcripts";
 import { FileService } from "./files";
 
+const defaultBaseUrl = "https://api.assemblyai.com";
+
 class AssemblyAI {
   /**
    * The files service.
@@ -31,12 +32,13 @@ class AssemblyAI {
    * @param params The parameters for the service, including the API key and base URL, if any.
    */
   constructor(params: BaseServiceParams) {
-    params.baseUrl = params.baseUrl || "https://api.assemblyai.com";
-    const client = createAxiosClient(params);
-    this.files = new FileService(client);
-    this.transcripts = new TranscriptService(client, this.files);
-    this.lemur = new LemurService(client);
-    this.realtime = new RealtimeServiceFactory(client, params);
+    params.baseUrl = params.baseUrl || defaultBaseUrl;
+    if (params.baseUrl && params.baseUrl.endsWith("/"))
+      params.baseUrl = params.baseUrl.slice(0, -1);
+    this.files = new FileService(params);
+    this.transcripts = new TranscriptService(params, this.files);
+    this.lemur = new LemurService(params);
+    this.realtime = new RealtimeServiceFactory(params);
   }
 }
 

package/src/services/lemur/index.ts

@@ -8,54 +8,55 @@ import {
   LemurActionItemsResponse,
   LemurTaskResponse,
   PurgeLemurRequestDataResponse,
-} from "@/types";
-import { BaseService } from "@/services/base";
+} from "../..";
+import { BaseService } from "../base";
 
 export class LemurService extends BaseService {
-  async summary(params: LemurSummaryParameters): Promise<LemurSummaryResponse> {
-    const { data } = await this.client.post<LemurSummaryResponse>(
-      "/lemur/v3/generate/summary",
-      params
-    );
-    return data;
+  summary(params: LemurSummaryParameters): Promise<LemurSummaryResponse> {
+    return this.fetchJson<LemurSummaryResponse>("/lemur/v3/generate/summary", {
+      method: "POST",
+      body: JSON.stringify(params),
+    });
   }
 
-  async questionAnswer(
+  questionAnswer(
     params: LemurQuestionAnswerParameters
   ): Promise<LemurQuestionAnswerResponse> {
-    const { data } = await this.client.post<LemurQuestionAnswerResponse>(
+    return this.fetchJson<LemurQuestionAnswerResponse>(
       "/lemur/v3/generate/question-answer",
-      params
+      {
+        method: "POST",
+        body: JSON.stringify(params),
+      }
     );
-    return data;
   }
 
-  async actionItems(
+  actionItems(
     params: LemurActionItemsParameters
   ): Promise<LemurActionItemsResponse> {
-    const { data } = await this.client.post<LemurActionItemsResponse>(
+    return this.fetchJson<LemurActionItemsResponse>(
       "/lemur/v3/generate/action-items",
-      params
+      {
+        method: "POST",
+        body: JSON.stringify(params),
+      }
     );
-    return data;
   }
 
-  async task(params: LemurTaskParameters): Promise<LemurTaskResponse> {
-    const { data } = await this.client.post<LemurTaskResponse>(
-      "/lemur/v3/generate/task",
-      params
-    );
-    return data;
+  task(params: LemurTaskParameters): Promise<LemurTaskResponse> {
+    return this.fetchJson<LemurTaskResponse>("/lemur/v3/generate/task", {
+      method: "POST",
+      body: JSON.stringify(params),
+    });
   }
 
   /**
    * Delete the data for a previously submitted LeMUR request.
    * @param id ID of the LeMUR request
    */
-  async purgeRequestData(id: string): Promise<PurgeLemurRequestDataResponse> {
-    const { data } = await this.client.delete<PurgeLemurRequestDataResponse>(
-      `/lemur/v3/${id}`
-    );
-    return data;
+  purgeRequestData(id: string): Promise<PurgeLemurRequestDataResponse> {
+    return this.fetchJson<PurgeLemurRequestDataResponse>(`/lemur/v3/${id}`, {
+      method: "DELETE",
+    });
   }
 }

package/src/services/realtime/factory.ts

@@ -3,30 +3,35 @@ import {
   RealtimeTokenParams,
   CreateRealtimeServiceParams,
   RealtimeServiceParams,
-} from "@/types";
-import { AxiosInstance } from "axios";
+  RealtimeTemporaryTokenResponse,
+} from "../..";
 import { RealtimeService } from "./service";
+import { BaseService } from "../base";
 
-export class RealtimeServiceFactory {
-  constructor(
-    private client: AxiosInstance,
-    private params: BaseServiceParams
-  ) {}
+export class RealtimeServiceFactory extends BaseService {
+  private rtFactoryParams: BaseServiceParams;
+  constructor(params: BaseServiceParams) {
+    super(params);
+    this.rtFactoryParams = params;
+  }
 
   createService(params?: CreateRealtimeServiceParams): RealtimeService {
-    if (!params) params = { apiKey: this.params.apiKey };
+    if (!params) params = { apiKey: this.rtFactoryParams.apiKey };
     else if (!("token" in params) && !params.apiKey) {
-      params.apiKey = this.params.apiKey;
+      params.apiKey = this.rtFactoryParams.apiKey;
     }
 
     return new RealtimeService(params as RealtimeServiceParams);
   }
 
   async createTemporaryToken(params: RealtimeTokenParams) {
-    const response = await this.client.post<{ token: string }>(
+    const data = await this.fetchJson<RealtimeTemporaryTokenResponse>(
       "/v2/realtime/token",
-      params
+      {
+        method: "POST",
+        body: JSON.stringify(params),
+      }
     );
-    return response.data.token;
+    return data.token;
   }
 }

package/src/services/realtime/index.ts

@@ -1,2 +1,2 @@
-export * from "@/services/realtime/factory";
-export * from "@/services/realtime/service";
+export * from "./factory";
+export * from "./service";

package/src/services/realtime/service.ts

@@ -8,13 +8,13 @@ import {
   PartialTranscript,
   FinalTranscript,
   SessionBeginsEventData,
-} from "@/types";
+} from "../..";
 import {
   RealtimeError,
   RealtimeErrorMessages,
   RealtimeErrorType,
-} from "@/utils/errors";
-import { Writable } from "stream";
+} from "../../utils/errors";
+import Stream from "stream";
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
 
@@ -32,7 +32,6 @@ export class RealtimeService {
     this.realtimeUrl = params.realtimeUrl ?? defaultRealtimeUrl;
     this.sampleRate = params.sampleRate ?? 16_000;
     this.wordBoost = params.wordBoost;
-    this.realtimeUrl = params.realtimeUrl ?? defaultRealtimeUrl;
     if ("apiKey" in params) this.apiKey = params.apiKey;
     if ("token" in params) this.token = params.token;
 
@@ -162,8 +161,8 @@ export class RealtimeService {
     this.socket.send(JSON.stringify(payload));
   }
 
-  stream(): Writable {
-    const stream = new Writable({
+  stream(): NodeJS.WritableStream {
+    const stream = new Stream.Writable({
       write: (chunk: Buffer, encoding, next) => {
         this.sendAudio(chunk);
         next();

package/src/services/transcripts/index.ts

@@ -1,4 +1,4 @@
-import { BaseService } from "@/services/base";
+import { BaseService } from "../base";
 import {
   ParagraphsResponse,
   SentencesResponse,
@@ -12,9 +12,10 @@ import {
   Retrieveable,
   SubtitleFormat,
   RedactedAudioResponse,
+  TranscriptListParameters,
   WordSearchResponse,
-} from "@/types";
-import { AxiosInstance } from "axios";
+  BaseServiceParams,
+} from "../..";
 import { FileService } from "../files";
 
 export class TranscriptService
@@ -25,8 +26,8 @@ export class TranscriptService
     Deletable<Transcript>,
     Listable<TranscriptList>
 {
-  constructor(client: AxiosInstance, private files: FileService) {
-    super(client);
+  constructor(params: BaseServiceParams, private files: FileService) {
+    super(params);
   }
 
   /**
@@ -45,13 +46,16 @@ export class TranscriptService
       params.audio_url = uploadUrl;
     }
 
-    const res = await this.client.post<Transcript>("/v2/transcript", params);
+    const data = await this.fetchJson<Transcript>("/v2/transcript", {
+      method: "POST",
+      body: JSON.stringify(params),
+    });
 
     if (options?.poll ?? true) {
-      return await this.poll(res.data.id, options);
+      return await this.poll(data.id, options);
     }
 
-    return res.data;
+    return data;
   }
 
   private async poll(
@@ -82,21 +86,29 @@ export class TranscriptService
    * @param id The identifier of the transcript.
    * @returns A promise that resolves to the transcript.
    */
-  async get(id: string): Promise<Transcript> {
-    const res = await this.client.get<Transcript>(`/v2/transcript/${id}`);
-    return res.data;
+  get(id: string): Promise<Transcript> {
+    return this.fetchJson<Transcript>(`/v2/transcript/${id}`);
   }
 
-  // TODO: add options overload to support list querystring parameters
   /**
-   * Retrieves a paged list of transcript listings.
-   * @param nextUrl The URL to retrieve the transcript list from. If not provided, the first page will be retrieved.
-   * @returns
+   * Retrieves a page of transcript listings.
+   * @param parameters The parameters to filter the transcript list by, or the URL to retrieve the transcript list from.
    */
-  async list(nextUrl?: string | null): Promise<TranscriptList> {
-    const { data } = await this.client.get<TranscriptList>(
-      nextUrl ?? "/v2/transcript"
-    );
+  async list(
+    parameters?: TranscriptListParameters | string
+  ): Promise<TranscriptList> {
+    let url = "/v2/transcript";
+    if (typeof parameters === "string") {
+      url = parameters;
+    } else if (parameters) {
+      url = `${url}?${new URLSearchParams(
+        Object.keys(parameters).map((key) => [
+          key,
+          parameters[key as keyof TranscriptListParameters]?.toString() || "",
+        ])
+      )}`;
+    }
+    const data = await this.fetchJson<TranscriptList>(url);
     for (const transcriptListItem of data.transcripts) {
       transcriptListItem.created = new Date(transcriptListItem.created);
       if (transcriptListItem.completed) {
@@ -104,7 +116,7 @@ export class TranscriptService
       }
     }
 
-    return data as unknown as TranscriptList;
+    return data;
   }
 
   /**
@@ -112,28 +124,21 @@ export class TranscriptService
    * @param id The identifier of the transcript.
    * @returns A promise that resolves to the transcript.
    */
-  async delete(id: string): Promise<Transcript> {
-    const res = await this.client.delete<Transcript>(`/v2/transcript/${id}`);
-    return res.data;
+  delete(id: string): Promise<Transcript> {
+    return this.fetchJson(`/v2/transcript/${id}`, { method: "DELETE" });
   }
 
   /**
    * Search through the transcript for a specific set of keywords.
    * You can search for individual words, numbers, or phrases containing up to five words or numbers.
    * @param id The identifier of the transcript.
-   * @param id Keywords to search for.
+   * @param words Keywords to search for.
    * @return A promise that resolves to the sentences.
    */
-  async wordSearch(id: string, words: string[]): Promise<WordSearchResponse> {
-    const { data } = await this.client.get<WordSearchResponse>(
-      `/v2/transcript/${id}/word-search`,
-      {
-        params: {
-          words: JSON.stringify(words),
-        },
-      }
+  wordSearch(id: string, words: string[]): Promise<WordSearchResponse> {
+    return this.fetchJson<WordSearchResponse>(
+      `/v2/transcript/${id}/word-search?words=${JSON.stringify(words)}`
     );
-    return data;
   }
 
   /**
@@ -141,11 +146,8 @@ export class TranscriptService
    * @param id The identifier of the transcript.
    * @return A promise that resolves to the sentences.
    */
-  async sentences(id: string): Promise<SentencesResponse> {
-    const { data } = await this.client.get<SentencesResponse>(
-      `/v2/transcript/${id}/sentences`
-    );
-    return data;
+  sentences(id: string): Promise<SentencesResponse> {
+    return this.fetchJson<SentencesResponse>(`/v2/transcript/${id}/sentences`);
   }
 
   /**
@@ -153,11 +155,10 @@ export class TranscriptService
    * @param id The identifier of the transcript.
    * @return A promise that resolves to the paragraphs.
    */
-  async paragraphs(id: string): Promise<ParagraphsResponse> {
-    const { data } = await this.client.get<ParagraphsResponse>(
+  paragraphs(id: string): Promise<ParagraphsResponse> {
+    return this.fetchJson<ParagraphsResponse>(
       `/v2/transcript/${id}/paragraphs`
     );
-    return data;
   }
 
   /**
@@ -167,11 +168,8 @@ export class TranscriptService
    * @return A promise that resolves to the subtitles text.
    */
   async subtitles(id: string, format: SubtitleFormat = "srt"): Promise<string> {
-    const { data } = await this.client.get<string>(
-      `/v2/transcript/${id}/${format}`
-    );
-
-    return data;
+    const response = await this.fetch(`/v2/transcript/${id}/${format}`);
+    return await response.text();
   }
 
   /**
@@ -179,11 +177,10 @@ export class TranscriptService
    * @param id The identifier of the transcript.
    * @return A promise that resolves to the subtitles text.
    */
-  async redactions(id: string): Promise<RedactedAudioResponse> {
-    const { data } = await this.client.get<RedactedAudioResponse>(
+  redactions(id: string): Promise<RedactedAudioResponse> {
+    return this.fetchJson<RedactedAudioResponse>(
       `/v2/transcript/${id}/redacted-audio`
     );
-    return data;
   }
 }
 

package/src/types/files/index.ts

@@ -0,0 +1,11 @@
+type FileUploadParameters = string | FileUploadData;
+type FileUploadData =
+  | NodeJS.ReadableStream
+  | ReadableStream
+  | Blob
+  | BufferSource
+  | ArrayBufferView
+  | ArrayBufferLike
+  | Uint8Array;
+
+export type { FileUploadParameters, FileUploadData };

package/src/types/index.ts

@@ -1,3 +1,4 @@
+export type * from "./files";
 export type * from "./transcripts";
 export type * from "./realtime";
 export type * from "./services";