Is support for OpenAI /v1/responses API planned?

[nblog]

Hi,

I’d like to ask if there are any plans to support the OpenAI /v1/responses API endpoint in addition to /v1/chat/completions? This would help integrations with the newest OpenAI APIs and tooling. If it is on the roadmap, is there an expected timeline?

Thanks!

[nblog]

Thanks for taking a look. To clarify, this is not a model-version or request-configuration issue, and it is not about Anthropic SDK behavior.

My question is specifically about whether ai-sdk-cpp plans to add first-class support for OpenAI's /v1/responses endpoint, in addition to the current Chat Completions path.

I also did a quick source-level check of the current OpenAI streaming implementation, and there seems to be a related capability gap around streaming tool calls:

• OpenAIClient currently configures OpenAI completions as /v1/chat/completions and stream_text() streams against that endpoint.
• OpenAIStreamImpl::parse_sse_line() only handles choices[0].delta.content and finish events. It does not inspect or accumulate choices[0].delta.tool_calls.
• Non-streaming tool calls are supported only on the full response path, where OpenAIResponseParser::parse_success_completion_response() parses choices[0].message.tool_calls.

So, as far as I can tell, both streaming tool-call shapes are currently unsupported:

1. a complete delta.tool_calls payload arriving in a single SSE chunk; and
2. the normal OpenAI incremental format where function.arguments arrives as partial string fragments across multiple chunks and needs to be assembled by tool-call index / id.

That is one of the reasons I am asking about /v1/responses support and the broader streaming event model. If there is already a supported way to receive streaming tool calls through stream_text(), could you point me to the relevant API or example?

Source references from current main at the time of checking:

• src/providers/openai/openai_client.cpp: /v1/chat/completions and stream_text() endpoint setup
• src/providers/openai/openai_stream.cpp: parse_sse_line() only emits text deltas from delta.content
• src/providers/openai/openai_response_parser.cpp: non-streaming message.tool_calls parsing

[daveCode-dot]

@nblog Apologies — that prior comment was posted in error from an experimental auto-queue and has been deleted. You're right that it didn't address the actual question about /v1/responses API support. Sorry for the noise.