Text-embeddings

[haesleinhuepf]

In order to know which slides are related to each other and group slides by content, it might be useful to explore text-embeddings. To minimize costs, I propose to try mxbai-embed-large and ollama-python.

We could reuse infrastructure developed in #8 to turn PDFs into images, then images into text, and text into embeddings. The embedding could then be explored interactively using a notebook similar to this one.

[lea-33]

Thanks! I am working on it. I just wanted to add that in the example notebook the code for the stackview.sliceplot is somehow not working correctly as you don't actually see the plot! But it worked after i tried it myself, so it should work actually.

[haesleinhuepf]

Don't worry about the interactive plot. This doesn't render on the github website :-)

[lea-33]

Don't worry about the interactive plot. This doesn't render on the github website :-)

Alright! I also tried to do it in the Notebook I downloaded but still I can't really select the data points like it is supposed to be. I only can draw a circle but the data points are not chosen as a selection afterwards.. Have you ever encountered that problem?