Hi there,

First of all, thank you for building OmicVerse — it's an incredibly comprehensive and well-designed toolkit that has been very helpful to the community. The unified API design and extensive documentation are really impressive.

I'm the developer of FlashDeconv, a spatial deconvolution method we recently published on bioRxiv. I was wondering if you would be open to integrating it into OmicVerse's spatial module.

Brief introduction:

• Processes 1 million spots in ~3 minutes on CPU (no GPU required)
• O(N) linear scaling for both time and memory
• Capable of detecting rare cell types via leverage-score sampling

If you're interested, I'd be happy to do all the implementation work myself, including writing the wrapper, documentation, and tests following your contribution guidelines.

Totally understand if it doesn't fit the project's scope — just wanted to ask first before submitting a PR.

Thanks again for your great work!