What problem would this solve?
I'd like to implement some fancy features such as on-pic OCR (using NPU) and custom panels. However, they would probably hurt the product philosophy of fast and lightness, and may lead to an over-customized edition just for few.
Especially for OCR backends, RapidOCR is too large and complicated and may not needed by many, while AI infering backends on the Windows platform is divided by several manufactors, causing no universal implemention. Windows Foundry Local does not solve this problem as well.
Proposed solution
I wonder if this project could provide an open interface, and let some relatively huge and ugly things be plugged in manually.
Alternatives considered
No response
Additional context
No response
What problem would this solve?
I'd like to implement some fancy features such as on-pic OCR (using NPU) and custom panels. However, they would probably hurt the product philosophy of fast and lightness, and may lead to an over-customized edition just for few.
Especially for OCR backends, RapidOCR is too large and complicated and may not needed by many, while AI infering backends on the Windows platform is divided by several manufactors, causing no universal implemention. Windows Foundry Local does not solve this problem as well.
Proposed solution
I wonder if this project could provide an open interface, and let some relatively huge and ugly things be plugged in manually.
Alternatives considered
No response
Additional context
No response