CHANGELOG.md

Changelog

[4.20.0]

• Add on_low_language_confidence property to LanguageDetectionOptions

  Controls behavior when language confidence is below threshold. Either "error" (default) or "fallback". When set to "fallback", the transcription will use the fallback language instead of erroring when confidence is low.

[4.8.0]

• Add multichannel property to TranscriptParams
• Add multichannel and audio_channels property to Transcript
• Add channel property to TranscriptWord, TranscriptUtterance, TranscriptSentence, and SentimentAnalysisResult

[4.7.1]

• Log a warning when a user tries to use API key authentication in the browser to connect to the real-time Streaming STT API.
• Update dependencies
• Use assembly.ai short URL for sample files

[4.7.0]

• Add language_confidence_threshold to Transcript, TranscriptParams, and TranscriptOptionalParams.

  The confidence threshold for the automatically detected language. An error will be returned if the language confidence is below this threshold.

• Add language_confidence to Transcript

  The confidence score for the detected language, between 0.0 (low confidence) and 1.0 (high confidence)

Using these new fields you can determine the confidence of the language detection model (enable by setting language_detection to true), and fail the transcript if it doesn't meet your desired threshold.

Learn more about the new automatic language detection model and feature improvements on our blog.

[4.6.2]

• Change RealtimeErrorType from enum to const object.
• Add RealtimeErrorTypeCodes which is a union of RealtimeErrorType values

[4.6.1]

• Remove conformer-2 from SpeechModel union type.
• Remove conformer-2 deprecation warning

[4.6.0]

• Add more TSDoc comments for RealtimeService documentation
• Add new LeMUR models
• Add TranscriptWebhookNotification which is a union of TranscriptReadyNotification or RedactedAudioNotification
• Add RedactedAudioNotification which represents the body of the PII redacted audio webhook notification.

[4.5.0]

• You can now retrieve previous LeMUR responses using client.lemur.getResponse<LemurTask>("YOUR_REQUEST_ID").
• LeMUR functions now return usage with the number of input_tokens and output_tokens.

[4.4.7]

• Rename TranscriptService.redactions function to TranscriptService.redactedAudio.
• Add TranscriptService.redactedAudioFile function.
• Add workerd export to fix cache issue with fetch on Cloudflare Workers.

[4.4.6]

• Fix Rollup exports so __SDK_VERSION__ is properly replaced with the version of the SDK.

[4.4.5]

• Add new PiiPolicy enum values

[4.4.4]

• Add an export that only includes the Streaming STT code. You can use the export
  • by importing assemblyai/streaming,
  • or by loading the assemblyai.streaming.umd.js file, or assemblyai.streaming.umd.min.js file in a script-tag.
• Add new EntityType enum values

[4.4.3] - 2024-05-09

• Add react-native exports that resolve to the browser version of the library.

[4.4.2] - 2024-05-03

Changed

• Caching is disabled for all HTTP request made by the SDK
• Accept data-URIs in client.files.upload(dataUri), client.transcripts.submit(audio: dataUri), client.transcripts.transcribe(audio: dataUri).
• Change how the WebSocket libraries are imported for better compatibility across frameworks and runtimes. The library no longer relies on a internal #ws import, and instead compiles the imports into the dist bundles. Browser builds will use the native WebSocket, other builds will use the ws package.

[4.4.1] - 2024-04-16

Changed

• Deprecate enableExtraSessionInformation parameter in CreateRealtimeTranscriberParams type

[4.4.0] - 2024-04-12

Added

• Add disablePartialTranscripts parameter to CreateRealtimeTranscriberParams
• Add enableExtraSessionInformation parameter to CreateRealtimeTranscriberParams
• Add session_information event to RealtimeTranscriber.on()

Changed

• ⚠️ Deprecate conformer-2 literal for TranscriptParams.speech_model property

Fixed

• Add missing status property to AutoHighlightsResult

[4.3.4] - 2024-04-02

Added

• SpeechModel.Best enum
• TranscriptListItem.error property

Changed

• Make PageDetails.prev_url nullable
• Rename Realtime to Streaming inside code documentation
• More inline code documentation

Fixed

• Rename SubstitutionPolicy literal "entity_type" to "entity_name"
• Fix the pagination example in "List transcripts" sample on README

[4.3.3] - 2024-03-18

Added

• GitHub action to generate API reference
• Generate API reference with Typedoc and host on GitHub Pages

Changed

• Add conformer-2 to SpeechModel type
• Change language_code field to accept any string
• Move from JSDoc to TSDoc
• Update ws to 8.13.0
• Update dev dependencies (no public facing changes)

[4.3.2] - 2024-03-08

Added

• Add audio_url property to TranscribeParams in addition to the audio property. You can use one or the other. audio_url only accepts a URL string.
• Add TranscriptReadyNotification type for the transcript webhook body.

Changed

• Update codebase to use TSDoc
• Update README.md with more samples

[4.3.0] - 2024-02-15

Added

• Add RealtimeTranscriber.configureEndUtteranceSilenceThreshold function
• Add RealtimeTranscriber.forceEndUtterance function
• Add end_utterance_silence_threshold property to CreateRealtimeTranscriberParams and RealtimeTranscriberParams types.

[4.2.3] - 2024-02-13

Added

• Add speech_model field to TranscriptParams and add SpeechModel type.

[4.2.2] - 2024-01-29

Changed

• Windows paths passed to client.transcripts.transcribe and client.transcripts.submit will work as expected.

[4.2.1] - 2024-01-23

Added

• Add answer_format to LemurActionItemsParams type

Changed

• Rename RealtimeService to RealtimeTranscriber, RealtimeServiceFactory to RealtimeTranscriberFactory, RealtimeTranscriberFactory.createService() to RealtimeTranscriberFactory.transcriber(). Deprecated aliases are provided for all old types and functions for backwards compatibility.
• Restrict the type for redact_pii_audio_quality from string to RedactPiiAudioQuality an enum string.

[4.2.0] - 2024-01-11

Added

• Add content_safety_confidence to TranscriptParams & TranscriptOptionalParams.

Changed

• The RealtimeService now sends audio as binary instead of a base64-encoded JSON object.

[4.1.0] - 2023-12-22

Added

• Add "anthropic/claude-2-1" to LemurModel type
• Add encoding option to the real-time service and factory. encoding can be "pcm_s16le" or "pcm_mulaw".
• "pcm_mulaw" is a newly supported audio encoding for the real-time service.

Changed

• Allow any string into final_model for LeMUR requests

[4.0.1] - 2023-12-14

Added

• Add "assemblyai/mistral-7b" to LemurModel type

Changed

• Update types with @example
• Update types with Format: uuid if applicable

[4.0.0] - 2023-12-08

Added

• Add node, deno, bun, browser, and workerd (Cloudflare Workers) exports to package.json. These exports are compatible versions of the SDK, with a few limitations in some cases. For more details, consult the SDK Compatibility document.
• Add dist/assemblyai.umd.js and dist/assemblyai.umd.min.js. You can reference these script files directly in the browser and the SDK will be available at the global assemblyai variable.

Changed

• RealtimeService.sendAudio accepts audio via type ArrayBufferLike.
• Breaking: RealtimeService.stream returns a WHATWG Streams Standard stream, instead of a Node stream. In the browser, the native web standard stream will be used.
• ws is used as the WebSocket client as before, but in the browser, the native WebSocket client is used.
• Rename Node SDK to JavaScript SDK as the SDK is compatible with more runtimes now.

[3.1.1] - 2023-11-21

Added

• Add client.transcripts.transcribe function to transcribe an audio file with polling until transcript status is completed or error. This function takes an audio option which can be an audio file URL, path, stream, or buffer.
• Add client.transcripts.submit function to queue a transcript. You can use client.transcripts.waitUntilReady to poll the transcript returned by submit. This function also takes an audio option which can be an audio file URL, path, stream, or buffer.

Changed

• Deprecated client.transcripts.create in favor of transcribe and submit, to be more consistent with other AssemblyAI SDKs.
• Renamed types
  • Renamed Parameters type suffix with Params type suffix
  • Renamed CreateTranscriptParameters to TranscriptParams
  • Renamed CreateTranscriptOptionalParameters to TranscriptOptionalParams.
• Added deprecated aliases for the forementioned types
• Improved type docs

[3.1.0] - 2023-11-16

Added

• Add AssemblyAI.transcripts.waitUntilReady function to wait until a transcript is ready, meaning status is completed or error.
• Add chars_per_caption parameter to AssemblyAI.transcripts.subtitles function.
• Add input_text property to LeMUR functions. Instead of using transcript_ids, you can use input_text to provide custom formatted transcripts as input to LeMUR.

Changed

• Change default timeout from 3 minutes to infinite (-1). Fixes #17

Fixed

• Correctly serialize the keywords for client.transcripts.wordSearch.
• Use more widely compatible syntax for wildcard exporting types. Fixes #18.

[3.0.1] - 2023-10-30

Changed

• The SDK uses fetch instead of Axios. This removes the Axios dependency. Axios relies on XMLHttpRequest which isn't supported in Cloudflare Workers, Deno, Bun, etc. By using fetch, the SDK is now more compatible on the forementioned runtimes.

Fixed

• The SDK uses relative imports instead of using path aliases, to make the library transpilable with tsc for consumers. Fixes #14.
• Added speaker property to the TranscriptUtterance type, and removed channel property.

[3.0.0] - 2023-10-24

Changed

• AssemblyAI.files.upload accepts streams and buffers, in addition to a string (path to file).

Removed

• Breaking: The module does not have a default export anymore, because of inconsistent functionality across module systems. Instead, use AssemblyAI as a named import like this: import { AssemblyAI } from 'assemblyai'.

[2.0.2] - 2023-10-13

Added

• AssemblyAI.transcripts.wordSearch searches for keywords in the transcript.
• AssemblyAI.lemur.purgeRequestData deletes data related to your LeMUR request.
• RealtimeService.stream creates a writable stream that you can write audio data to instead of using `RealtimeService.sendAudio``.

Fixed

• The AssemblyAI class would be exported as default named export instead in certain module systems.

[2.0.1] - 2023-10-10

Re-implement the Node SDK in TypeScript and add all AssemblyAI APIs.

Added

• Transcript API client
• LeMUR API client
• Real-time transcript client