So we have implemented a feature using Assistant API, but as we are refactoring the code we are planning to move to Response API. But currently response do not have concept of thread.
And I was going through this article - https://openai.com/index/new-tools-for-building-agents which says
We’re working to achieve full feature parity between the Assistants and the Responses API, including support for Assistant-like and Thread-like objects, and the Code Interpreter tool. Once this is complete, we plan to formally announce the deprecation of the Assistants API with a target sunset date in mid-2026.
So I wanted to ask will there be thread like concept in responses? and by when is the OpenAI team planning to release?
Please shed some light on this if anyone has any knowledge about this.
Thanks for the help in advance
1 Like
Responses API request has a field named previous response id.
Specifying this previous response id tells the API that the request is a response to a specific message.
When you include this in your messages, you are effectively creating a thread.
When you keep it empty, you create a new thread.
While it is true that you can reuse a previous response ID as an input for up to 30 days when using the Responses endpoint, that does not give the full effect of “thread”.
Assistants’ threads are manageable mutable objects. You can delete message from them. As a detraction, you cannot easily reuse them as a state for multiple API calls, as you place a message into a thread and then obtain an appended AI message that also becomes part of the thread.
Responses IDs are simply a state of all the messages as a result of an API call. There also is no parameter like Assistants’ turn threshold to limit your expense.
So, no, there is no threadlike function replication there. Hopefully, OpenAI are thinking hard about a stateful product a developer would actually want in their next go-around at this endpoint.
1 Like