memery
memery copied to clipboard
deepfates
Search over large image datasets with natural language and computer vision!
The streamlit GUI should keep the db and CLIP model loaded while the folder doesn't change. This probably necessitates writing its own Flow, I guess? Maybe with native widgets for...
It'd be nice to install this without having to downgrade to torch 1.7.1 (as well as torchvision and some others). It seems like this is being forced by the `clip-by-openai`...
Annoy takes a long time to build, maybe there's something else out there that doesn't
Memery serve will break if you don't give a filepath argument. It's optional through typer but I think the streamlit gui is configured to require it.
Open in local file explorer, with appropriate operation for different operating systems
Related to #13, in the sense that this issue is made worse by the indexing process not being resumable. When indexing a large directory of various types of files (with...
This is helpful when debugging, and encouraging when searches take a long time. But it's also messy with a bunch of `print` statements going all over the place, so if...
I have code for a pyMDE mapping of the CLIP embeddings already, just need to refactor it to work with the new API for queryFlow might want to make improvements...
Better practice to pass around typed objects. Especially if we hope to have a distributed server backend ( #7 )
Eventually it would be ideal to have a hosted/self-hosted server that communicates with thin clients on various devices. This may be accomplished with something like anvil.works or jina.ai but it...