Incorporation of different embedding and OCR models

[slowlead]

Due to security concerns, I have been trying to use an embeddinggemma-300m embedding model and tesseract (https://github.com/tesseract-ocr/tesseract) for OCR. Unfortunately, this has led to quite a few problems in implementation that are well beyond my skillset. Having them as baked in options would provide good options to the included models.