CHANGELOG.md

Changelog

v5.0.1

• Add exports for AudioManagerOptions and AudioOutputsServicePlayOptions.

v5.0.0

BREAKING

• Added AudioManager class to handle audio input and output.
• Removed Microphone and Speaker classes. Replaced with AudioManager methods.
• Use AudioContext API for better browser compatibility.

v4.0.5

• Updated types SpeechRecognitionStartEvent, SpeechRecognitionResponse, SpeechRecognitionParameters with traceId.
• Add orientation type to ImageLayer.

v4.0.3

• Speech recognition now uses the correct sample rate in Firefox.

v4.0.2

• Stopping the microphone now stops the underlying MediaStreamTrack, removing the red 'listening' icon.

v4.0.1

• Fixed an issue where speech recognition was using an incorrect sample rate.

v4.0.0

• BREAKING: This packages now exports ES Module only.
• BREAKING: An implementation of fetch is no longer included with this library. Consumers of this library should ensure their environment supports fetch.
• BREAKING: This library now relies on having URLSearchParams available in the environment.
• BREAKING: This library now relies on having BigInt available in the environment.
• BREAKING: playthroughIds and conversationIds have been changed from number to string type everywhere in this SDK, and renamed to playthroughUuid and conversationUuid.
• BREAKING: api.createPlaythroughToken, api.createConversation and api.createCharacterConversation all now return an object instead of a scalar, to facilitate any future changes to these methods.
• BREAKING: memories in getPlaythroughInfo and in the message event now return saveValues as JSON instead of only as strings. For example, counter memories are now actually numbers and boolean memories are now actually booleans.
• BREAKING: getMessageHistory has been removed, and getEventHistory added as a more fully-featured alternative with much greater support for filtering, and can return all event types.
• setMemory now accepts any JSON values instead of only strings.
• BREAKING: Speech recognition stream now uses common objects to start up, deliver results and stop, regardless of which downstream service is selected.
• Add start and stop events to the speech recognition stream.

v3.9.0

• Added result event to Microphone so clients can subscribe to raw SpeechRecognition events.
• recognise and recognise-interim now emit the text of the last result instead of the first result in the SpeechRecognition event if continuous is true.

v3.8.0

• Multiple memories can now be set at once using the setMemory call.

v3.7.0

• Add support for forkPlaythrough API. This enables a player to upgrade to the latest published version from their old playthrough, copying across memories and emotions into the new playthrough, and returning the new token. Note that conversations are not carried across.

v3.6.1

• problem events scoped to a conversation can now be listened to via conversation.on("problem", ...)

v3.6.0

• It's now possible to specify multiple supported speech encodings in speechConfig by passing an array instead of a string. Charisma will use the first encoding that the voice synthesis service supports.
• Added experimental support for intermediate client events. These events can be sent to Charisma to prevent characters from talking if the player is still speaking or typing. This can only be enabled for a story by getting in touch at hello@charisma.ai.

v3.5.0

• Added support for Decentraland.

v3.4.2

• SDK info is now also sent upon reconnection to the room.

v3.4.1

• package.json now references correct emitted types location.

v3.4.0

• Added languageCode option to createPlaythroughToken, to play Charisma stories in languages other than English.
• Added SDK info to joining a room, for Charisma to track which SDK versions are in use.

v3.3.0

• It's now possible to subscribe to events that are sent from other players, such as other players' messages. This can be done by adding a subscriber to a conversation to listen for the corresponding event, e.g. conversation.on("reply", () => { /* remote player's reply */ }). These handlers will not be fired for messages sent from the local connected client, only for remote clients.
• Add missing graphId: number to MessagePathItem type.
• Updated dependencies.

v3.2.0

• Added startGraphId and startGraphReferenceId to StartEvent to start from a specific graph ID.
• Added experimental pause and play methods to Playthrough.

v3.1.0

• Support for action node/event.
• SpeechRecognitionStopOptions is now exported.

v3.0.0

There is a new emotion engine in Charisma! As a result...

• message.characterMoods has been removed and replaced with message.emotions. This contains each character's current mood and relationship with the player, and any active feeling effects.
• setMood has been removed. We may add an equivalent API for the new emotion engine in the future. Let us know about your use case if this interests you!

v2.3.0

• Microphone.stopListening() now accepts an options parameter with a single option waitForLastResult. If set to true, then the recognise will be called a final time with the result of the audio captured so far. If false, the operation will be aborted, so no additional recognise event will occur.

v2.2.0

• Speaker.play() now accepts an options parameter as its second parameter instead of a boolean value (which used to represent interrupt). This change is backwards compatible, but the old boolean way is deprecated and will be removed in the next major release.
  • options contains two parameters: trackId and interrupt. trackId can be used to interrupt only a particular track, for example, to prevent a character talking over themselves. interrupt can now be configured to all (interrupt all playing audio), track (interrupt the specified trackId if playing), or none (don't interrupt any audio).

v2.1.0

• Adds the option to pass an apiKey to use for authentication for playthrough token creation. This is now the recommended way to authenticate as API keys do not expire (unless regenerated) and are more secure than the userToken. userToken should no longer be used.

v2.0.0

This release makes several breaking changes. The main change is replacing socket.io with colyseus.js.

• Replaces socket.io with colyseus.js.
  • Due to how Colyseus serializes data, audio is now an ArrayBuffer instead of an object with the data property.
• API methods and the Playthrough constructor now accept a baseUrl option, which is used in preference to globalBaseUrl. globalBaseUrl is now set with setGlobalBaseUrl instead of setBaseUrl.
• API methods are now individually exported instead of being static methods on the Charisma class, as well as being exported under a bracket api object.
• Improved the implementation of Microphone.
• Replace multiple connection events from Charisma (connect, disconnect etc) with single connection-status event.
• The Charisma class has been renamed to Playthrough.
• The cleanup function has been renamed to disconnect.

v1.10.0

• Change imageLayers field to an array of object, each including url, resizeMode and points.

v1.9.1

• Add isImpactShareable and impactImageUrl fields to impacts, and fix the type of impact ids to be strings.

v1.9.0

• BREAKING CHANGE: eventIds are now emitted as strings. Please upgrade to this version to continue using the reconnection "catch-up" logic (though everything else should work).

v1.8.1

• Speaker will no longer try to play audio if the context's state is not running. This resolves an issue where the user has not granted permission for the audio context to play sound, and so the play(...) promise never resolves.

v1.8.0

• Reconnecting will now fetch and emit messages that were emitted from the server after the last received message.
• Add impacts field to GetPlaythroughResult type.

v1.7.0

• Pass through more events: reconnect, reconnecting, disconnect and problem.
• Added types for new panel message, and added bubble-related types onto the media key.
• Adjusted setMemory type to accept null.
• Removes scene-complete event and stopOnSceneComplete option.
• Adds episode-complete event. The chat engine automatically stops on episode end if the episode is started by an app user.
• Adds restartFromEpisodeId and restartFromEpisodeIndex methods and removes restartFromScene method.

v1.6.1

• GET requests no longer try to add a body to the request.

v1.6.0

• Add start and stop events to Speaker.

v1.5.0

• Adds support for use in Node.js.

v1.4.0

• Microphone.startListening(timeout) now has a timeout parameter to automatically stop the microphone after timeout milliseconds.
  • Microphone.resetTimeout(timeout) will reset the timeout to timeout milliseconds.
• Microphone now emits start and stop events, particularly useful in conjuction with timeout.

v1.3.0

• Add an interrupt option to Speaker to ensure replies don't overlap.
• Breaking: Target ES2018; drop support for pre-ES2018 environments.

v1.2.0

• Add restartFromScene method to SDK. This can be used to reset the playthrough to the state it was in at the beginning of a particular scene.
• Exports more types and adjusts message types to include eventId, timestamp and memories.

v1.1.0

• Add resume event to SDK. This can be used to resume a conversation from where it left off.

v1.0.5

• Use webkitAudioContext for Speaker on Safari.

v1.0.4

• Export the Impact type.

v1.0.3

• impacts are now objects containing their ID as well as the impact string.

v1.0.2

• Rename setStopOnSceneEnd to setStopOnSceneComplete to ensure consistency with the event name.

v1.0.1

• Fix createPlaythroughToken throwing an error when both version and userToken are not provided.

v1.0.0

• Initial stable release.
• Completely overhauls the SDK API, please see the README for more details on how to use the newer, conversation-based API.

Past major versions

v0.9.2

• Pass data (containing impacts) through on scene complete event.

v0.9.1

• Pass stopOnSceneComplete through to the CharismaInstance.

v0.9.0

• Add stopOnSceneComplete option to prevent automatically continuing between scenes.

v0.8.3

• Add media field onto the character message type.

v0.8.2

• Add tapToContinue to message history type.

v0.8.1

• Add timestamp to messages returned from getMessageHistory.
• Improved type for getMessageHistory.

v0.8.0

• Can now specify playthroughToken to re-use a playthrough instead of creating a new one when connecting.
• Can now fetch message history of the playthrough using charisma.getMessageHistory.

v0.7.3

• Fix IMessageCharacter.speech type.

v0.7.2

• ISynthesisConfig and some additional types are now exported.

v0.7.1

• speech now takes a config object as well as a boolean. It can specify the audio encoding to use and whether it returns the raw audio data, or a link to an audio file.

v0.7.0

• BREAKING: The reply event has been renamed to message, and now has a type field on the payload to distinguish between character and media events. Other fields have been refactored, such as character and speech. Please consult src/types.ts to find the new message format.
• A new tap event is available for the client to send.

v0.6.0

• Accidentally published version, but never tagged as latest.

v0.5.1

• Fix broken 0.5.0 publish.

v0.5.0

• Removed browser field from package.json. Consumers can use the UMD directly from unpkg.
• Removed actIndex as it is no longer supported.

v0.4.2

• Buffer set-memory events until status: 'ready' is received.

v0.4.1

• actIndex and sceneIndex can now be set on the start event to start from a specific story scene.

v0.4.0

• BREAKING: UMD name changed from Charisma to CharismaSDK. The ES/CJS builds now don't bundle their dependencies.
• Added setMemory method to directly set a memory.
• Fixed all ID types to be number, not string.

v0.3.1

• Passing no version to the connect method now results in using the latest published version, rather than the draft version.

v0.3.0

• Package renamed (rescoped) to @charisma-ai/sdk.

v0.2.0

• The debug option has been replaced with the version option, which defaults to undefined (the latest published story version).

v0.1.2

• The microphone now stops listening when a reply with endStory set to true is emitted.

v0.1.1

• AudioContext is now created on-demand rather than on initialisation.

v0.1.0

• Socket.io now forces websockets, skipping the long-polling check.

v0.0.4

• Fixed issue where audio was not working on Safari.

v0.0.3

• Microphone keeps better track of whether to resume speech recognition after speaking.

v0.0.2

• Support for recording speech-to-text via Chrome SpeechRecognition.
  • New events recognise-interim and recognise.
  • Speech recognition is paused while the audio is played.

v0.0.1

• Initial release.
• Support for reply and start client events, and reply, start-typing and stop-typing server events.
• Support for playing text-to-speech audio.