WCR APIs: Tell us what investments you want to see!

[connorMicrosoft]

Hey folks,

Connor from Microsoft here. Which WCR APIs are you using the most? What are your thoughts on how we should prioritize improvements to the APIs?

Thanks!

[kmgallahan]

The underlying ML models required for these APIs currently require your device to be running the latest Windows 11 Insider Preview Build on the Dev Channel. Additionally, these APIs require your device to be a Copilot+ PC.

99.99% of devs can't use the APIs, so resolving that would help.

[purudpd]

Here are some of the things I believe are very much needed 😊

• Bring your own model, both Image and / or Text. Allow developers to bring their models and use them directly through the WCR.
• Support for vector embedding models.
• Powerful speech to text API and text to speech APIs.
• Add support for newer LLM tech, function calling, JSON Mode, longer context etc.

[Marv51]

As you must hear all the time, either a CPU/GPU fallback or a cloud fallback is needed for me to adopt these APIs for any real features in the next years.

[CakeRepository]

Function calling here