Custom http headers #93
Conversation
…e. Refactor URLRequestBuildable and its implementations to separate configuration concerns from URL building concerns.
… that the code better adheres to the Single Responsibility Principle.
…nly be used by the OpenAI class.
…o enable future extension and/or testing.
|
Kudos, SonarCloud Quality Gate passed!
|
|
Hi! Is there anything more I need to do in order to get this merged? |
|
Hey, @benvolioT! |
| // on the next pass in case that new data has completed the string to create valid (parsable) JSON, and so on. | ||
| // Note that while OpenAI's API doesn't appear to return partial string fragments, proxy service such as Helicone | ||
| // do seem to do so, making this logic necessary. | ||
| accumulatedData.append(stringContent) |
There was a problem hiding this comment.
Could you extract this change to a separate PR, please?
Looks like this is a separate fix, that doesn't apply to custom headers.
| } | ||
|
|
||
| extension OpenAI { | ||
| internal func generateHeaders() -> [String: String] { |
There was a problem hiding this comment.
I would change the naming to defaultHeaders to add the knowledge that it is going to be appended further.
What do you think?
Feat: Get token usage data for streamed chat completion response.
|
What
This pull request adds the ability to specify custom HTTP headers when integrating with OpenAI via proxy APIs such as Helicone. It also refactors OpenAi's performRequest() by encapsulating the logic for handling the session data task in its own class, URLSessionDataTaskManager. This should simplify the OpenAI class and make the code easier to manage. This pull request simplifies JSONRequest and MultipartFormDataRequest by abstracting out common code to remove repetition that previously existed in these structs. Finally, it simplifies the protocol URLRequestBuildable, moving configuration-driven parameter setting to the constructors of these struct instances, rather than in the call to that protocol's build() function.
Why
The motivation for this change was to allow users to integrate with OpenAI via the Helicone proxy. To do so, they not only need to point the OpenAI client at the Helicone host, but they also have to set an additional HTTP header on the requests. This pull request makes that possible.
Affected Areas
The affected areas of the library include: