All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Add support for the new ChatGPT API, including
gpt-3.5-turbo
model and the new Chat completions endpoint. Chat completions guide. - Add support for the new Whisper API, allowing Transcriptions and Translations, accepting a variety of formats (
m4a
,mp3
,mp4
,mpeg
,mpga
,wav
,webm
). Speech to text guide. - Add new
AudioTranscriptions
Handler, which transcribes audio into the input language text using the Whisper API. - Add new
AudioTranslations
Handler, which transcribes audio into english text using the Whisper API. - Add new
ChatCompletions
Handler, which creates a completion for one or more chat messages using the ChatGPT API.
- Clarify which models can be used in
\Tectalic\OpenAi\Models\Edits\CreateRequest::$model
when performing Edits. - Clarify that
\Tectalic\OpenAi\Models\Embeddings\CreateRequest:$input
can be a maximum of 8192 tokens (not 2048 tokens). - Clarify that
\Tectalic\OpenAi\Models\ImagesEdits\CreateImageRequest::$mask
is no longer a required field. - API version updated from 1.1.0 to 1.2.0.
- Add support for PHPUnit v9.6.x and v10.x.
- Remove
id
andmodel
required properties from theTectalic\OpenAi\Models\Edits\CreateResponse
model, as they are no longer returned by OpenAI's API. - Improve compatibility with the
php-http/discovery
package v1.15.0 and newer. - Use Fully Qualified Class Names for Examples in Readme.
- Update Copyright year.
- Fix
Response body parse failed
error when retrieving a Model response from theEdits::create()
handler and method. - Fix incorrect Error handling example in Readme.
- Use parameters defined outside endpoint methods.
- Encourage the use of
php-http/mock-client
for testing and mocking API responses. - Remove the
Tests\MockHttpClient
class, and use thephp-http/mock-client
package instead. - Make Handler and Model class names more readable.
- Use correct model type for nested models.
- Add support for DALL·E image generation.
- Add new
ImageGenerations
Handler, which creates an image given a prompt. - Add new
ImagesEdits
Handler, which creates an edited or extended image given an original image and a prompt. - Add new
ImagesVariations
Handler, which creates a variation of a given image.
- Improve Handler unit tests.
- API version updated from 1.0.6 to 1.1.0.
- Improve readme.
- Remove deprecated
Answers
handler and associated models. - Remove deprecated
Classifications
handler and associated models. - Remove deprecated
Engines
handler and associated models. - Remove deprecated
EnginesSearch
handler and associated models.
- Switch License.
- Add support for Moderation using a new
Moderations::create()
Handler class and Method. - Add usage information to response models:
Completions::create()
,Edits::create()
andEmbeddings::create()
.
- Define required properties for response models.
- Rename all nested response models.
- Change default value for
Tectalic\OpenAi\Models\FineTunes\CreateRequest::$prompt_loss_weight
. - 22 API Methods are now supported, grouped into 14 API Handlers.
- API version updated from 1.0.5 to 1.0.6.
- Don't run CI for tags.
- Use correct model type for nested models:
Tectalic\OpenAi\Models\FineTunes\CreateResponse
,Tectalic\OpenAi\Models\FineTunes\RetrieveResponse
andTectalic\OpenAi\Models\FineTunesCancel\CancelFineTuneResponse
.
- Initial release.