Added
- Support for OpenAI embedding models, including batch inputs, custom dimensions, token usage, and result metadata (#34).
- Support for fine-tuned OpenAI models by deriving capabilities from their underlying base models (#17).
- Support for editing and refining generated images using reference images and OpenAI’s image-editing endpoint (#29).
- A
TokenLimitReachedExceptionwhen OpenAI responses are incomplete because the maximum output-token limit was reached (#10).
Changed
- Made sampling-option capabilities model-aware for reasoning models, added Responses API support for log probabilities, and reject incompatible sampling and reasoning configurations before sending a request (#40).
- Bumped WordPress tested-up-to version 7.1 (#43).
Fixed
- Function-call name handling for names that do not meet OpenAI’s naming requirements, while preserving the original PHP AI Client function names returned to callers (#31).
New Contributors
- @rmens made their first contribution in #17
- @itzmekhokan made their first contribution in #29
- @henryperkins made their first contribution in #40
- @chubes4 made their first contribution in #34
Full Changelog: 1.0.3...1.1.0.
View all items closed in the milestone.
Contributors
henryperkins, rmens, and 2 other contributors