Client Laravel Package

Added

• Add updateAttributes method to VectorStoresFiles (#626)
• Add support for chat completion audio modality (#629)
• Add Containers API (#636)
• Add model field to CreateResponse for embeddings (#634)

Fixed

• Fix streamed response tool call without arguments in VLLM (#623)
• Fix support for code_interpreter_call on retrieve response (#632)
• Fix JSON encoding to preserve UTF-8 characters using JSON_UNESCAPED_UNICODE (#628)

Added

• Add helper method output_text to Responses API (#579)
• Add support for response cancel in Responses API (#588)
• Add support for Image Generation Tool to Responses API (#594)
• Add streaming for Image Generation in Responses API (#602)
• Add support for Code Interpreter to Responses API (#610)
• Add support for Remote MCP Tool in Responses API (#601)
• Add realtime ephemeral tokens (#591)
• Add streaming support for Audio transcriptions (#603)
• Add realtime key documentation (#597)

Fixed

• Add missing index to CreateStreamedResponseToolCall (#562)
• Make $parameters optional in thread create() method (#577)
• Support Azure model.list endpoint (#599)
• Add reasoning_effort support to Assistants (#606)
• Use proper parameter order in Response constructor (#615)
• Support reused prompts/instructions in Responses API (#616)

Changed

• Use proper header notation for Responses API docs (#596)
• Collapse legacy/deprecated sections by default in docs (#609)

Deprecated

• mark Assistants API as deprecated (#607)
• mark Completions API as legacy (#608)

Added

• Add support for Responses API (#541)

Fixed

• Add Throwable type support to ClientFake responses (#576)

Changed

• Removed PHP 8.1 support

Added

• Add logprobs to Chat Response (#533)
• Add ResponseHasMetaInformationContract contract to ThreadRunStepResponse (#523)
• Add support for 'attributes' on vector store files (#551)
• Add OpenAI compatibility support for Google Gemini (#502)
• Add compatibility for Aliyun LLM APIs (#530)
• Add ability to pass arguments to files list request (#557)
• Add search vector store functionality (#559)
• Add Image Response usage (#571)
• Add category applied input types to moderation response (#572)
• Add support for annotations in Chat response (Web Search) (#564)
• Add test coverage for assistants streaming and related functionality (#444)

Changed

• Update GitHub Action deprecations & opt into Dependabot (#544)
• Draw attention away from deprecated completions endpoints in docs (#548)

Fixed

• Fix type definition for responses in ClientFake::addResponses method (#382)
• Fix Content retrieval in HttpTransport with custom HttpClient (#343)
• Fix correct completion endpoint when logprobs missing (#550)
• Fix chat completion choices to allow responses without logprobs field (#554)
• Fix support for streaming of non-OpenAI models that return "ping" (#556)
• Fix OpenRouter token usage response (#560)
• Fix Gemini list models (#567)

Added

• Add http status to ErrorException (#487)
• Add cached_usage to CreateResponseUsage for Chat (#494)
• Add moderation categories (Illicit*) (#495)

Fixed

• Fix missing parameters on FineTuning RetrieveJobResponse (#496)
• Fix nullable description on Assistants Tool Function (#484)
• Fix attachment key missing on ThreadMessageResponse (#471)

Added

• Add thread.run.incomplete to ThreadRunStreamResponse (#421)
• Add withProject to configure the project for the client (#377)
• Add fake b64_json to support mocking (#462)

Fixed

• Fix image url content type to use url instead of file_id (#422)
• Fix type error on VectorStoresFilesTestResources (#460)
• Fix vector store cancel method (#435)

Added

• Add support for Assistants API v2 and Vector Stores endpoints (#420)

Docs

• Add vector store endpoints documentation (#420)

Added

• Support for usage stream option on chat endpoint (#398)

Fixed

• Missing output parameter on streamed code interpreter output (#406)

Added

• Add support for Assistants API v2 and Vector Stores endpoint (#405)

Added

• Add support for Batches endpoint (#403)

Added

• Assistants: add streaming support (#367)

Added

• Audio: add support for timestamp_granularities (#374)

Fixed

• Fix default fake data for meta information (#332)

Added

• ThreadRun: Add "usage" property to the response (#330)

Fixed

• ThreadRunStep: "content" missing in response if result has not been submitted (#319)
• Files: "bytes" in retrieve response is may null (#325)

Added

• Add support for Assistants and Threads endpoint (#271)
• Add stream support for Text To Speech (#235)
• Add test resources for Assistants and Threads (#279)

Changed

• Remove thread messages delete endpoint (#309)

Fixed

• Handle x-request-id in meta information (#283)
• Handle meta information from azure headers (#307)
• Add missing default system_fingerprint to chat create response fixture (#308)
• Convert headers to lower case before creation meta information (#306)

Docs

• Remove threads list endpoint from README.md (#275)
• Clarify assistants files docs (#278)
• Fix image creation example (#297)
• Fix outdated links (#299)
• Add troubleshooting section and explain how to configure HTTP client timeouts

Removed

• Remove list() from Threads resource

Added

• Add support for Assistants and Threads endpoint (#271)

Fixed

• instruction on ThreadRunResponse may be nullable

Added

• Add revised_prompt property to CreateResponseData on the image create endpoint (#257)

Docs

• Fix model in one of the examples

Added

• Add RetrieveJobResponseError and batch_size, learning_rate_multiplier parameters on RetrieveJobResponseHyperparameters for fine-tuning endpoint (#255)

Added

• Add support for Assistants and Threads endpoint (#243)

Added

• Add support for GTP-4 vision on the chat completion endpoint (#241)

Added

• Add support for tool calls on the chat completion endpoint (#239)

Added

• Add support for the audio speech endpoint (#237)

Changed

• Update Models endpoint response object to the latest API changes (#235)

Docs

• Update FineTuning job id names (#230)
• Use Chat resource as the primary example

Fixed

• nEpochs on RetrieveJobResponseHyperparameters may be string
• processingMs ond MetaInformationOpenAI may be null (#218)

Added

• Add "has_more" to fine-tuning jobs and events list responses (#206)

Changed

• Add parameters to the fine-tuning jobs list request to filter the results (#206)

Fixed

• error_code may be int

Fixed

• Missing openai-version header from Azure

Fixed

• Typo in class name MetaInformationOpenAI

Added

• Add support for the fine-tuning API (#199)
• Provide access to header / meta information for all responses (#195)

Changed

• Mark FineTunes resource as deprecated
• Mark Edits resource as deprecated
• Add missing moderation enums (#178)

Fixed

• Chat completion create response with function calling on Azure (#184)
• Breaking change on OpenAI API regarding "transient" field in Audio translations (#168)
• Docs: fix OpenAI URL

Fixed

• Breaking change on OpenAI API regarding "transient" field in Audio (#160)

Changed

• Error handling: use error code as exception message if error message is empty (#150)

Fixed

• Error handling: Catch error in stream responses (#150)
• Error handling: Handle errors where message is an array (#150)

Fixed

• Chat/CreateResponse faking with function_call (#145)

Added

• Add support for function calling in the Chat Completions API (#144)

Fixed

• Exception handling for server error with non default content type header (#134)
• Faking embedding responses for multidimensional vectors (#131)

Added

• Add support for psr/http-message ^2.0 (#130)

Fixed

• fix: stream broken after checking for errors (regression of #113)

Added

• Support for HTTP base uri (#106)

Changed

• unify exception handling between HTTP client implementations (#113)

Fixed

• fix toArray() on CreateStreamedResponseDelta to match the original API response (#108)

Docs

• explain usage for "OpenAI on Azure" (#109)

Added

• Testing support (#71)

Changed

• Trim ApiKey before sending it to the API (#101)

Fixed

• Nullable fields on error response (#102)

Added

• Stream support (#84)

Changed

• Removed dependency for guzzlehttp/guzzle and use PSR-18 client discovery instead (#75)
• Add Client factory which allows for a custom HTTP client
• Client factory further accepts custom HTTP headers, query parameters and API URI

Fixed

• status_details can be a string in file responses. Affects Files and FineTunes resources (#68)

Added

• Audio resource to turn audio into text powered by whisper-1 (#62)

Added

• Chat resource aka ChatGPT powered by gpt-3.5-turbo (#60)

Fixed

• Nullable finish_reason on Completions CreateResponse (#52, 545e0ab)

Fixed

• Missing events on FineTunes RetrieveResponse (#41)

Changed

• OpenAI::client() first argument changed from apiToken to apiKey (#25)

Fixed

• Getting contents from Guzzle's response causing issues with middleware (#33)

Fixed

• FineTunes create response: batch_size, learning_rate and fine_tuned_model are nullable (#16)
• File responses: add missing fields status and status_details

Added

• Add images() resource to interact with DALL-E

Fixed

• Parse completions create response with logprobs correctly

Added

• First version