Changelog for openai-2.0.0
2.0.0:
-
BREAKING CHANGE: Renamed
Item_InputMessage
toItem_Input_Message
inOpenAI.V1.Responses
for consistency with new constructorsThe
InputItem
data type in the Responses API has been updated to use a consistent naming scheme. If you are using the Responses API, you will need to update your code:-- Before (v1.2.0): Item_InputMessage { role = ..., content = ..., status = ... } -- After (v2.0.0): Item_Input_Message { role = ..., content = ..., status = ... }
-
Add support for function tool calls in Responses API
- New
InputItem
constructors:Item_Input_Function_Call
,Item_Input_Function_Call_Output
,Item_Input_Item_Reference
- Add flattened tool JSON representation for Responses API compatibility
- Export status constants:
statusIncomplete
,statusCompleted
- Export tool choice constants:
toolChoiceNoneText
,toolChoiceAutoText
,toolChoiceRequiredText
- New
-
Code quality improvements:
- Optimize
isFunctionField
usingHashSet
for O(1) lookups - Simplify
unflattenChoice
with guards - Extract magic strings as named constants
- Optimize
1.2.0:
/v1/responses
: Add support for Responses API- Add
Tool_Web_Search
1.1.1:
- Remove timeout on default
ClientEnv
/v1/chat/completions
: Add support for Search- Fix CreateSpeech JSON instances, add new voices and optional instructions field
- Correct
ToJSON
ofFileSearchResources
- New example app for tool-calling and chat-loop
1.1.0:
-
BREAKING CHANGE: Fix details representations for various types [#44] [#45] [#50] [#51]
A few details-related fields were fixed to match the behavior of the OpenAI API.
-
Add support for
reasoning_effort
parameter in chat completions
1.0.1:
- Include
README
- Include usage example
1.0.0:
- Initial release