A voice in the wilderness: Advocating for the Assistants API

API Feedback
1

It’s widely speculated that the Assistants API was sidelined due to its perceived complexity and lack of flexibility, particularly concerning agent functionalities. While I’m not privy to OpenAI’s internal decision-making process that led to the ‘splitting’ of its features into the Response API and the Agents SDK, I can’t help but feel a missed opportunity.

In my view, evolving the Assistants API by integrating dynamic functionality exchange with completions context would have offered a unique approach to agent workflows and RAG, distinguishing it from the rest of the market. However, OpenAI has chosen to abandon this path, delegating further development to Responses and Agents.

Although the Agents SDK is relatively straightforward to use (despite its seemingly premature ‘production’ release), I believe the Assistants API held far greater potential. Ultimately, this is, as the title of my post suggests, a voice crying in the wilderness.

2 Likes
2

:cry: No need to get dramatic :smile:

Broadly, the functionality is carrying over:

:rocket:

1 Like
3

Okay, I’ll dial back the ‘drama’ a bit :). Anyway, intention is one thing, and how it translates into reality is another… we’ll trust in the good work of the people at OpenAI.

4

In any case I’d argue it is better to use Completions and get much better control over token expenditure. But perhaps they will address that.

1 Like
5

I don’t see any drama here. I see an obscure reference to a NASA engineer’s recommendation that nearly got him fired, but when the idea was considered, it saved a billion dollars and made the space program safer.