vit_gpt_dart_api 6.0.5 changelog

6.0.5 #

Feat: RealtimeModel and its implementations support initial messages.
Feat: RealtimeModel now has "isSendingInitialMessages" stream.
Refac: Conversation model class now support initial messages in its constructor.
Refac: Message.createdAt is no longer final.

6.0.2 #

Feat: AudioRecorderModel now has resume() and pause() methods.

6.0.1 #

Feat: Message class now has "usage".

6.0.0 #

BREAKING: RealtimeModel no longer has "onUsage" stream, use "onResponse".
Feat: RealtimeModel now has "onResponse" stream, meant to provide the ai response and the usage in on single event.

5.2.0 #

Feat: Usage class present in the "response.done" event of OpenaiRealtimeRepository now has the cached tokens details.

5.1.0 #

Feat: Message now accepts a more flexible json in the "Message.fromMap" constructor.

5.0.1 #

Fix: All factories in DynamicFactory have a setter and getter.
Feat: Everytime a method tries to retrive/save something from/to local storage, the code will either abort or return a default value instead of throwing an exception.
Fix: "completion" from DynamicFactories is used in "createCompletionRepository" method.

5.0.0 #

BREAKING: Method was renamed in ThreadsModel from "sendMessage" to "createMessage".
BREAKING: Class VitGptConfiguration was renamed to VitGptDartConfiguration.
BREAKING: Removed function "setupFactories".
BREAKING: Renamed factories in DynamicFactories:
- simplePlayerFactory → simplePlayer.
- speeachToText → transcriber.
- completionFactory → completionWithAssistant.

4.2.0 #

Feat: Added "logLevel" in VitGptConfiguration to make easier to change the log level across the package.
Fix: Logs are no longer shown in profile and release mode by default.

4.1.0 #

Feat: TranscriptionEnd now includes "content" which is the accumulated text for the transcription.
Refac: Uses package "logger" to create logs. To customaze it, change on "VitGptConfiguration.logger".

4.0.0 #

BREAKING: Refactored RealtimeModel and therefore, OpenaiRealtimeRepository.
BREAKING: Refactored Message class to be consistent with other APIs.
BREAKING: SenderType enum is now called Role to be consistent with other APIs.
Feat: RealtimeModel now has token usage stream.

3.10.3 #

Fix: OpenaiRealtimeRepository fixed typo on event "conversation.item.created".

3.10.2 #

Fix: OpenaiRealtimeRepository correctly catches open ai events for ai text response.

3.10.1 #

Fix: When using openai realtime, sends the audio data to the stream in the correct order.

3.10.0 #

Feat: DynamicFactories now has public simple player factory setter.

3.9.0 #

Feat: Cancel ai response in RealtimeModel.

3.8.0 #

Feat: New stream "onRawAiAudio" on RealtimeModel.
Feat: New property "streamAiAudioAsText" on RealtimeModel.
Fix: "isUserSpeaking" is not properly updated on OpenaiRealtimeRepository.

3.7.0 #

Feat: Any error in the server messages in OpenaiRealtimeRepository is now handled.
Fix: Rate limits are now correctly parsed in OpenaiRealtimeRepository.

3.6.1 #

Fix: OpenaiRealtimeRepository now correctly handles base 64 data.

3.6.0 #

Feat: Method "getApiToken" now will try to fetch the api token from the http client if the local storage was not set.
Fix: OpenaiRealtimeRepository now produces the correct headers.

3.5.0 #

Feat: RealtimeModel now has the "isConnected" property.
Fix: OpenaiRealtimeRepository now correctly emits events as String instead of Map.

3.4.2 #

Fix: OpenaiRealtimeRepository is now correctly configured to use WebSocket connection instead of socket io.

3.4.1 #

Fix: Adjusted web socket headers in OpenaiRealtimeRepository.

3.4.0 #

Feat: RealtimeModel now has "getSocketHeaders" to improve customization for websocket.

3.3.2 #

Fix: send user audio data in OpenaiRealtimeRepository no longer sends invalid data.
Log: Added logs to OpenaiRealtimeRepository.

3.3.1 #

Feat: A LocalStorageModel is no longer required to set an open ai token.

3.3.0 #

Feat: Realtime model, repository and factory.
Feat: Completion factory in DynamicFactory.
Doc: Added example of completion repository.
BREAKING: Removed property "threadId" from Message.

3.2.3 #

Feat: updated dependencies

3.2.2 #

Fix: Conversation parse from map.

3.2.1 #

Fix: ThreadsModel dynamic factory propertly works now.

3.2.0 #

Feat: ThreadsRepository now accepts custom dio client on constructor.

3.1.1 #

Fix: CompletionRepository now correctly checks for empty previous messages.

3.1.0 #

Refac: "message" parameter is not a optional named parameter in ConversationRepository.
Feat: Added optional parameter "previousMessage" to "prompt" method in ConversationRepository.

3.0.0 #

BREAKING: ConversationRepository no longer has "onFirstMessageCreated" on "prompt" method.
Added optional parameter "previousMessages" to CompletionModel methods "fetch" and "fetchStream".
Added the getter "addsResponseAutomatically" to CompletionModel.
Removed constructor parameter "messages" from CompletionRepository in favor of "previousMessages" of "fetch" and "fetchStream" methods.
AssistantRepository now uses the API parameter "additional_messages", which eliminates the need to call another route separately to add the message to the thread.
Added "onMessageCreated" and "onMessageCreateError" callbacks to the "prompt" method from ConversationRepository.
The "prompt" method from ConversationRepository no longer waits for message creations on the thread and instead relies on the new callbacks to improve performance.

2.5.2 #

Fix: TranscriberRepository now correctly closes the stream of strings when the transcription is stopped.

2.5.1 #

Fix: AssistantRepository now correctly uses the http client given in the constructor.

2.5.0 #

Feat: added "assistantRepository" factory to DynamicFactories.

2.4.2 #

Fix: Stop speaker

2.4.1 #

Fixed createTranscriberRepository.

2.4.0 #

Renamed createListenerRepository to createTranscriberRepository.

2.3.0 #

Added "onMicVolumeChange" and "onSilenceChange" to TranscriberModel.

2.2.2 #

Added security checks on TranscriberRepository to prevent errors in case of miss use of methods.

2.2.1 #

Renamed ListenerRepository to TranscriberRepository.
FIX: TranscriberRepository now disposes its voice recorder.

2.2.0 #

Renamed ListenerModel to TranscribeModel.
Deprecated "transcribe" method in favor of TranscribeModel usage directly.

2.1.0 #

Feat: Added dynamic factory for ListenerModel (used on transcription).

2.0.1 #

Refac: 'model' is not longer required when creating a instance of Assistant.

2.0.0 #

Feat: Renamed completion onChunk to onJsonComplete to make it more clearer. Also added the same field to ConversationRepository.

1.8.6 #

Feat: onChunk callback for completion models.

1.8.5 #

FEAT: Dynamic factory for threads model

1.8.4 #

Fix: made SpeakerHandler logger private.

1.8.3 #

Added logs to SpeakerHandler.

1.8.2 #

Added safe guard to SpeakerHandler on "speakSentences" to prevent multiple timers.

1.8.1 #

Speaker handler no longer processes new chunks if it is stopped.

1.8.0 #

Volume stream support for Speaker Handler.

1.7.0 #

Max sentence delay in speaker handler and local storage.

1.6.0 #

Function for onPlay and onSentenceCompleted on speaker handler.

1.5.1 #

Fix: parse of json stream from open ai

1.5.0 #

Ability to catch and handle exception on the assistant repository

1.4.0 #

Ability to save thread title.

1.3.2 #

Fixed speaker omitting pontuation on sentences.

1.3.1 #

Silence detector with custom static variables.

1.3.0 #

Silence detector on recorder handler.

1.2.3 #

Ability to change speaker voice.

1.2.2 #

Fixed bug when loading non existent thread.

1.2.1 #

Allow custom tts handler.

1.2.0 #

Assistant list and run.

1.1.1 #

Configuration options for audio.

1.1.0 #

Ask assistant on thread.

1.0.0 #

Initial version.

vit_gpt_dart_api 6.0.5 vit_gpt_dart_api: ^6.0.5 copied to clipboard

Metadata