User Profile

I'm working with the new speech recognition APIs in iOS 17 and have encountered some confusion regarding the use of URLs in SFSpeechLanguageModel.prepareCustomLanguageModel and the SFSpeechLanguageModel.Configuration. In the SFSpeechLanguageModel.Configuration initializer, I provide a URL that points to a custom language model .bin file. However, there's also a URL parameter in the prepareCustomLanguageModel method. I'm unclear about the purpose of this second URL and how it differs from the one in the configuration. To add to the confusion, the documentation for these new APIs is not fully fleshed out at this point. I've tried injecting both .bin files (for the custom language model and the one for prepareCustomLanguageModel) into the same URL, but the results haven't clarified their distinct roles. In experiments I conducted, I checked the confidence level of recognized phrases from the same audio file with and without the custom language model .bin file. Surprisingly, the confidence levels remained the same in both scenarios, leading me to question if the custom model is being utilized correctly. Has anyone else worked with these new APIs and can provide clarity on: The distinct roles of the URLs in SFSpeechLanguageModel.Configuration and prepareCustomLanguageModel. Why there might be no noticeable difference in confidence levels when using a custom language model. Any insights or experiences with these new aspects of the iOS 17 speech recognition API would be greatly appreciated.

Hi, At this moment Xcode Cloud builds our SDK scheme. What we are interested is running a post-xcodebuild script that will run our custom: "xcodebuild -create-xcframework...." which in turn, will produce the product (file) we want to download and use. Of course, locally the script works great but we would like to integrate this process in Xcode Cloud CI. The way I see a good solution will be to have the product of the script available for download in the Artifact section inside Xcode Cloud. But, if it can't be done, how can we provide a link to the file after the machine has finished to run the script or does it mean that once the machine has finished the build it deletes everything. What about Xcode Cloud API ? is there a solution there ? Perhaps a another workaround would be to add another code to the script that will upload the script product/file to, for example, S3 so we would have access to it ? Thank you.