[Chore!] Add error verification on callback from stream

[letschers]

I was trying to implement a similar feature like the "Stop Generating" from ChatGPT and I noticed that there's no mechanism to stop the stream call coming from the OpenAI. I had the problem where even if I return from my callback, it will be called again and again. The most simple way I found was to return an error from the callback and validate this error.

This is a breaking change for every application that uses this lib (but also pretty simple to refactor), but I guess is important as much as necessary; if my callback is returning an error and anything is being done from the clientside, the server will keep generating chunks and wasting tokens.

I made in only one stream function to see what you guys think about it, but if it's relevant I'm willing to change the remaining stream methods

[pullrequest]

✅ This pull request was sent to the PullRequest network.

---

@letschers you can click here to see the review status or cancel the code review job.

[pullrequest]

PullRequest Breakdown

Reviewable lines of change

+ 5
- 3

100% Go

Type of change

Minor Update - These changes appear to be a minor update to existing functionality and features.

[pullrequest]

The change is simple and straightforward with the explanation from the author. But, as stated by the author, this is a breaking change that need to be taking care of by any other code using this

Xiaoyong W

---

Reviewed with ❤️ by PullRequest

[letschers]

Hey guys, any news?

[tylermann]

This change makes sense to me.

I could see a non-breaking change variant where we allow something like this:

onData = func (resp *ChatCompletionStreamResponse) {
    resp.Stop(errors.New('some error))
}

Although I think your proposed change here is more idiomatic and it is a fairly easy refactor as you mentioned. So lets just go with this.

Thanks for the contribution and sorry for the delay to getting around to this.